Search
1000 results for “invest_in_me_research”
-
PepsiCo
Despite decades of promises to end deforestation for palm oil PepsiCo (owner of crisp brands Frito-Lay, Cheetos and Doritos along with hundreds of other snack food brands) have continued sourcing palm oil that strongly linked to ecocide, indigenous landgrabbing, and the habitat destruction of the rarest animals on earth.
All of these animals are on a knife-edge of survival. It is for this reason, we boycott PepsiCo and its sub-brands. Find out about their forest destroying activities below and what you can do to stop them by using your wallet as a weapon. it’s the #Boycott4Wildlife
Take action
1. Share this page to Twitter
2. See PepsiCo’s palm oil deforestation from the past year
3. Boycott sub-brands of PepsiCo
4. Read reports about PepsiCo
5. Sign a petition about PepsiCo and palm oil
6. Boycott other brands using so-called “sustainable” palm oil
return to menu ↑
Take Action: Share to BlueSky & Twitter
Crisp and drink giant #Pepsi runs quirky ad campaigns enticing zoomers and millennials into a lifetime of #obesity and #diabetes. Yet few people know PepsiCo are linked to #indigenous #landgrabbing for #palmoil Take action! #Boycottpalmoil @palmoildetect https://wp.me/scFhgU-pepsico
Share to BlueSky Share to TwitterNext time you snack AVOID #Cheetos #Doritos #Lays crisps and #MountainDew 🍟🥤 because violence against #indigenous people for #palmoil comes as an unwanted freebie in snacks owned by #Pepsi. Take action! #BoycottPalmOil #Boycott4Wildlife @palmoildetect https://wp.me/scFhgU-pepsico
Share to BlueSky Share to Twitterreturn to menu ↑
View PepsiCo’s palm oil deforestation for the past year
Data courtesy of Palm Watch, a multidisciplinary research initiative by the University of Chicago.
Look Up PepsiCo on PalmWatchreturn to menu ↑
Take Action: Boycott These PepsiCo Sub-Brands
- Pepsi
- Lays
- Mountain Dew
- Doritos
- Gatorade
- Tropicana
- Quaker Oats
- Lipton
- Starbucks
- Aquafina
- Ruffles
- Cheetos
- Brisk
- Tostitos
- Frittos
- Diet Pepsi
- Diet Mountain Dew
- Sierra Mist
- 7Up
- Mirinda
- Walkers
- Pepsi Black
- Pure Leaf
- Bubly
- Naked
- Soda Stream
- Kevita
- Lifewtr
- Sierra Mist
- Stubborn Soda
- Rold Gold
- Miss Vickie’s
- Red Rock Deli
- Cracker Jack
- Nut Harvest
- Life
- Matador
- Quaker Chewy Granola Bars
- Santitas
- Funyuns
- Cap’n Crunch
- Rice-a-Roni and Pasta
- Roni Quaker Rice Crisps & Rice Cakes
- Maui Style
- Sabritones
- Munchies
- Munchos
- Grandma’s
- Aunt Jemima
- Izze Propel
- O.N.E
- Sobe Elixirs & Teas
- Mug Root Bear
- Stacy’s
- Bare Snacks
- Sabra
- Smart50
- Fritos
- Near East
- Sun Chips
- Smartfood
- Off the Eaten Path
- Simply
return to menu ↑
Take Action: Read Reports About PepsiCo
2021 report by Pusaka and others
Report by Pusaka, Walhi, and Forest Peoples Programme finds that household names including Nestlé, PepsiCo, Wilmar and Unilever and associated global financial institutions and investors continue to ‘turn a blind eye’ to human rights abuses in their palm oil supply chains.
Despite these very serious, long term and well documented human rights abuses and environmental damage, on the ground, major downstream companies continue to invest in, or source products from these plantations.
Read report2021 BBC Investigation
A 2021 joint BBC/Gecko Project and Mongabay Investigation found that Nestlé, Kellogg’s, Unilever, Johnson & Johnson, and PepsiCo have sourced palm oil from Indonesian companies linked to human rights abuses and have failed to pass on millions in profits to smallholder ‘plasma’ farmers.
Read more2021 Chain Reaction report
A Chain Reaction Report from 2021 showed that they have caused 100,000ha of deforestation in their palm oil supply chain since 2016.
Read morehttps://twitter.com/RAN/status/928326415098851328?s=20&t=de1q2YEmTjbSg2f_cJh83w
https://twitter.com/SumOfUs/status/1343240455664885763?s=20&t=ob6lQlwrIBkYsAjtaQG5iw
PepsiCo: Ties to illegal deforestation
Read full reportreturn to menu ↑
Sign petition: Tell PepsiCo stop destroying rainforests for palm oil!
PepsiCo’s profit-first palm oil policy is still destroying rainforests.
Meanwhile, PepsiCo keeps on promising that it’s working towards a truly sustainable palm oil policy, making commitments to human rights and zero deforestation. But this new report leaves no doubt: this whole time, PepsiCo’s palm oil promises have been nothing but smoke and mirrors.
Sign the Eko petitionreturn to menu ↑
Take Action: Boycott Other Brands
Brands Brands Brands Brands Brands Brands Brands Brands BrandsHere are some other ways you can help by using your wallet as a weapon and joining the #Boycott4Wildlife
Why join the #Boycott4Wildlife?
Greenwashing Tactic #4: Fake Labels
The Counterpunch: Consumer Solutions To Fight Extinction
Palm Oil Detectives is 100% self-funded
Palm Oil Detectives is completely self-funded by its creator. All hosting and website fees and investigations into brands are self-funded by the creator of this online movement. If you like what I am doing, you and would like me to help meet costs, please send Palm Oil Detectives a thanks on Ko-Fi.
Say thanks on Ko-Fi#Borneo #Boycott4wildlife #BoycottPalmOil #brandBoycotts #Cheetos #deforestation #diabetes #Doritos #greenwashing #indigenous #landgrabbing #Lays #Malaysia #MountainDew #News #obesity #PalmOil #palmOilBiofuel #palmOilDeforestation #palmoil #Pepsi #PepsiCo #RSPOGreenwashing
-
Accommodations: Individualized Responses to Structural Design Problems
The accommodations for natural human variation should be mutual.
Yet on a programmatic basis, disability policy and other social programs remain enmeshed, even at their best, in accommodation models, where specific proven needs or deficits generate specific individualized responses. What might it look like to shift our framing of the social safety net to a universal model?
I Shouldn’t Have to Dehumanize My Son to Get Him Support | The Nation
This captures an aspect of accommodation models that really frustrate us. They encourage individualized responses to structural design problems.
Structural Redesign
We do not need more after-the-fact accommodations awkwardly bolted on exclusionary systems.
Inclusion Must Be Global, Decolonized, Culturally and Linguistically Diverse, and Anti-Normative
Systems built for the most complex communicators are not niche solutions—they are blueprints for broad accessibility, exposing what existing systems routinely ignore (Hamraie 2017). We do not need more after-the-fact accommodations awkwardly bolted on exclusionary systems. What is required is radical redesign: systems that begin at the margins and work inward, rather than centering an imagined “average” user (Costanza-Chock 2020; Srinivasan 2025a). This means building from the outset for people with intersecting sensory, motor, and communication needs, as well as for those excluded by language, geography, or normative assumptions (Peña 2019; Srinivasan 2025a; Bal et al. 2016)
This has not been a call for one-size-fits-all designs but for adaptive systems that are customizable, flexible, and co-created with the very communities they claim to include. Structural redesign is not about eliminating all accommodations, but embedding them so deeply that they become expected rather than exceptional. For the most marginalized, inclusion will not come from incremental adjustments but from radical solutions and systemic overhauls. It requires collective commitment to radical acceptance—to presuming competence, to trusting diverse ways of knowing and being, to valuing both independence and interdependence, and to designing environments that recognize all people as having dignity and deserving to belong. As disability justice movements have long shown, design that starts at the edges moves everyone forward (Berne 2015; Hamraie 2017; Pineda 2020).
Inclusion Must Be Global, Decolonized, Culturally and Linguistically Diverse, and Anti-Normative
Instead of designing by default for “proven needs” well-known in disability and neurodiversity communities, accommodations models require individual episodes of forced intimacy, repeated over and over and over for the rest of your life. We should treat each episode of forced intimacy as a stress case that puts our designs to the test of real life.
Edge Cases and Stress Cases
Embodied experience isn’t an edge case. It’s a blueprint.
Our industry tends to call these edge cases-things that affect an insignificant number of users. But the term itself is telling, as information designer and programmer Evan Hensleigh puts it: “Edge cases define the boundaries of who and what you care about” (http://bkaprt.com/dfrl/00-01/). They demarcate the border between the people you’re willing to help and the ones you’re comfortable marginalizing.
That’s why we’ve chosen to look at these not as edge cases, but as stress cases: the moments that put our design and content choices to the test of real life.
It’s a test we haven’t passed yet. When faced with users in distress or crisis, too many of the experiences we build fall apart in ways large and small.
https://www.amazon.com/Design-Real-Life-Sara-Wachter-Boettcher/dp/1937557405
Instead of treating stress situations as fringe concerns, it’s time we move them to the center of our conversations-to start with our most vulnerable, distracted, and stressed-out users, and then work our way outward. The reasoning is simple: when we make things for people at their worst, they’ll work that much better when people are at their best.
School IEPs are a treasure trove of stress cases and structural problems currently treated individually. Let’s design for pluralism instead of putting us through a soul-chipping accommodations process that, at best, patches over bad design driven by “artificial economies of scarcity”.
Artificial Economies of Scarcity
The many forms of difference. Adaptive Behavior Assessment (ABAS-3), Adult ADHD Self-report Scale (ASRS-v1.1), and Behavior Rating Inventory Executive Function (BRIEF 2) forms spread across a wooden tableDisability systems rely on artificial economies of scarcity. Programs are underfunded, so caregivers, teachers, social workers, and disabled people themselves are all pushed to project their needs as necessary and virtuous.
What you can’t know unless you have #disability is how all the paperwork chips away at your soul. Every box you tick, every sentence about your “impairment” and “needs” becomes part of the narrative of your identity…
Bascom tells me that experiences like ours happen because disability service systems are never designed to support people with disabilities but are “about managing access to scarce resources. We start with the assumption that these resources are limited, so you have to prove over and over again that you need them more than anyone else. If we as a society invested more resources in supporting people with disabilities, we could redesign our systems accordingly.”
I Shouldn’t Have to Dehumanize My Son to Get Him Support | The Nation
The accommodations for natural human variation should be mutual.
Accessibility is a collective process! Invest in care, and design for real life. “The accommodations for natural human variation should be mutual.”
oh, there it is: neurodivergent people shouldn’t have to apply for ‘reasonable’ accommodations when the accommodations for natural human variation should be mutual
The medical model is essentially individualist.
The medical model is essentially individualist.
The prevalence of deficit-based thinking has the further consequence of focusing attention directly on the individual and away from social and environmental factors that might in fact play a significant role in shaping autistic lives (Engel, 1977).
Annual Research Review: Shifting from ‘normal science’ to neurodiversity in autism science
Instead of designing for our undeniable interdependence and mutuality, we are suffocated by a medical model steeped in deficit model thinking.
https://www.amazon.com/Inclusive-Education-Autistic-Children-Classroom/dp/1785923218
Terzi (2005, p.446), for example, is of the view that the medical model as played out in educational environments results in ‘perspectives emphasising individual limitations’ rather than the ways in which the organisation and design of schools might create those very difficulties in the first instance.
Wood, Rebecca. Inclusive Education for Autistic Children (p. 38). Jessica Kingsley Publishers. Kindle Edition.
Investigation and Humiliation
Y’all know the file, right? This has been the thing that had been following me since I started special education. Those things are thick and deep. KGB got nothing on special ed.
You have the right to food moneyProviding of course youDon't mind a littleInvestigation, humiliationAnd if you cross your fingersRehabilitationKnow your rightsThese are your rights-- Know Your Rights
#accessibility #accommodation #care #design #designForRealLife #disability #forcedIntimacy #inclusion #stressCases #structuralIdeology
-
Is Resin Printing Worth the Mess? Brutally Honest Breakdown for First-Timers
1,526 words, 8 minutes read time.
If you’ve been lurking in the shadows of 3D printing forums or scanning YouTube for the next big step in your printing game, chances are you’ve stumbled on resin printing. It’s that tantalizing tech that promises jaw-dropping detail, surfaces so smooth they make FDM prints look like sandpaper, and the kind of precision that makes miniatures and prototypes scream quality. But here’s the real talk: resin printing comes with a mess and a handful of headaches that many first-timers don’t see coming. So, is it worth diving into the resin pool, or should you stick to good ol’ filament? Let’s rip off the band-aid and get gritty on the truth of resin printing.
What Is Resin Printing? A Quick Overview
Before we dissect the good, the bad, and the ugly, it’s important to get clear on what resin printing actually is. Unlike FDM printers that melt and extrude plastic filament layer by layer, resin printers use a vat of liquid photopolymer resin cured by light. The most common types you’ll hear about are SLA (Stereolithography), MSLA (Masked Stereolithography), and DLP (Digital Light Processing). All use UV light to harden the resin in very thin, precise layers, which is why the level of detail you get is miles ahead of filament printing.
Resin printing is a fundamentally different beast—it’s more about light chemistry than hot plastic. That difference brings massive rewards in detail and finish, but also a totally different workflow that can feel like stepping into an alien lab if you’re used to FDM.
The Good: Why Resin Printing Rocks
Let’s start with the shine—resin printing delivers insane detail and surface smoothness that filament printers can’t touch. For guys who are into tabletop gaming, collectibles, or prototyping tiny mechanical parts, resin prints can capture the crisp edges and subtle curves you thought only existed in CAD renders. The resolution is typically measured in microns, not millimeters, which means you can pick out textures on a miniature’s armor or the intricate lattice on a prototype bracket with pinpoint accuracy.
Beyond beauty, resin prints can be incredibly strong and functional, depending on the resin you use. There are tough engineering resins, flexible ones, and even biocompatible varieties for dental or medical applications. This versatility means resin printing is carving out a solid place not just with hobbyists, but with businesses looking for rapid, high-fidelity prototyping without resorting to expensive CNC or injection molding.
Another bonus is how fast resin printers can spit out parts. Sure, you’re still building layer by layer, but curing a whole layer at once rather than tracing it with a nozzle often means speedier prints for small, detailed objects. When you want quality and speed in the same package, resin printing has your back.
The Bad: The Mess and Headaches of Resin Printing
Here’s where things get real. The downside to resin printing is all about the mess and the safety headaches that come with working with liquid resin. This stuff isn’t your run-of-the-mill filament spool you toss in and forget. Resin is a toxic, smelly chemical cocktail that demands respect and careful handling. Direct skin contact can cause irritation or allergic reactions, and the fumes aren’t something you want lingering in your man cave.
The post-processing is a chore you won’t escape. Once your print is done, you need to wash it, usually in isopropyl alcohol, to strip off uncured resin. Then, you have to cure it under UV light to harden it fully. This washing and curing routine isn’t just another step; it can take as long as the print itself and involves dealing with flammable liquids and sticky resin sludge.
Disposal is another headache. You can’t just pour leftover resin or used alcohol down the drain without risking environmental damage and local code violations. You’ll need to research how to properly cure and dispose of waste resin, which adds another layer of complexity for the newbie.
On top of that, the resin printer itself demands careful cleaning and maintenance. The vats and FEP films (the thin transparent layers at the bottom of resin trays) wear out and need replacing, and any spills can quickly turn your workspace into a nightmare. Without proper ventilation and protective gear like nitrile gloves and safety glasses, you’re flirting with respiratory irritation and skin problems.
Equipment and Setup: What You’ll Need to Manage the Mess
If you’re thinking resin printing sounds awesome but want to avoid turning your garage into a toxic swamp, prepping the right setup is non-negotiable. First up, safety gear isn’t optional — gloves, a respirator or mask rated for organic vapors, and eye protection are your frontline defense. You’ll also want a well-ventilated space or ideally, a dedicated room with a fume extractor. Trust me, the resin smell sticks around and gets old fast.
Next, post-processing tools like an ultrasonic cleaner or a good wash station can save you time and hassle. UV curing stations are essential to finish prints properly—while sunlight can do the job, it’s slow and inconsistent. Some budget printers come with UV lights built-in, but many require a separate device.
Your workspace should be easy to clean and resistant to resin spills. Plastic trays, disposable paper towels, and dedicated resin containers will save your sanity. The resin itself can be messy—be prepared for drips and splashes, especially when pouring and cleaning.
Maintenance and Ongoing Costs
Unlike filament printers where the ongoing costs are mostly filament and maybe a new nozzle now and then, resin printing carries a heavier price tag over time. Resin is more expensive per liter than filament, and waste from failed prints or washing can add up quickly. Consumables like replacement vats, FEP films, gloves, and isopropyl alcohol add to the tally.
Plus, the time cost isn’t trivial. Post-processing can double your total print time, especially if you’re meticulous about cleaning and curing. And neglecting maintenance or safety can lead to poor print quality or health issues.
First-Timer Tips: How to Survive and Thrive
If you’re still here and seriously thinking about dipping your toes into resin printing, here’s some hard-earned advice. Start small with cheap resins and basic printers before dropping serious cash. Never skip safety protocols—those gloves and goggles exist for a reason.
Plan your post-processing workflow before your first print. Set up a dedicated cleaning area, and always have proper waste disposal methods ready. Expect a learning curve; don’t get discouraged by early fails or messy spills. Clean resin off your tools and surfaces immediately; once it cures, it’s a nightmare to remove.
One of the biggest rookie mistakes is rushing prints or post-processing to save time. Resin printing rewards patience and precision. Follow manufacturer instructions closely, experiment with settings gradually, and join forums or communities to swap tips.
Is It Worth It? The Final Verdict
So, is resin printing worth the mess? The honest answer is: it depends. If you crave ultra-high detail, smooth surfaces, and can handle a bit of chemistry lab discipline, resin printing opens doors that filament can’t. Miniature painters, jewelers, model makers, and prototype developers will appreciate the leaps in quality and speed.
However, if you’re sensitive to chemicals, don’t want to invest in extra gear or spend significant time on post-processing, resin might not be your best first choice. FDM printing still rocks for durability, ease, and low cost.
The tech is evolving, and newer resins and machines are getting safer and less messy, but it’s still a commitment. Understanding the risks, costs, and workflow upfront will help you decide if this next-level tech deserves a spot in your printing arsenal.
Conclusion
Resin printing isn’t just a step up from filament; it’s a whole new game with different rules. It demands respect for the chemicals, time for cleanup, and patience to master. But the payoff—mind-blowing detail and finish—makes it an addiction for those who love pushing 3D printing’s limits.
If you’re ready to take the plunge or want to share your resin printing war stories, drop a comment below or reach out directly. And don’t forget to subscribe to our newsletter for more raw, honest 3D printing insights. This community’s all about keeping it real and getting the most out of our gear.
D. Bryan King
Sources
- All3DP: Resin 3D Printing Pros and Cons
- 3D Insider: What Is Resin 3D Printing?
- Formlabs: Resin vs FDM 3D Printing
- MakerBot: Resin 3D Printing Guide
- Creality: How to Clean a Resin 3D Printer
- MatterHackers: Safely Using Resin 3D Printers
- 3D Hubs: FDM vs Resin Printing
- Digital Trends: How Resin 3D Printing Works
- Prusa Printers: How to Clean Resin Prints
- Liebert Pub: Safety Considerations for Resin 3D Printing
- Hackaday: Resin 3D Printing – How Messy Is It?
- MakeUseOf: Is Resin 3D Printing Worth It?
- Tom’s Guide: Best Resin 3D Printers of 2024
- 3DPrint.com: Minimizing Resin Printing Mess
- YouTube: Resin 3D Printing Setup and Cleanup Tips
Disclaimer:
The views and opinions expressed in this post are solely those of the author. The information provided is based on personal research, experience, and understanding of the subject matter at the time of writing. Readers should consult relevant experts or authorities for specific guidance related to their unique situations.
Related Posts
Rate this:
#3dPrinterResinWaste #3dPrintingResinProsAndCons #bestResinPrinters #cleaningResinPrints #isResinPrintingWorthIt #resin3dPrintCleanup #resin3dPrintCuring #resin3dPrinterParts #resin3DPrinterSetup #resin3dPrinting #resin3dPrintingCost #resin3dPrintingDetail #resin3dPrintingEquipment #resin3dPrintingGuide #resin3dPrintingIsopropylAlcohol #resin3dPrintingMistakes #resin3dPrintingQuality #resin3dPrintingSpeed #resin3dPrintingToxicity #resin3dPrintingUVLight #resinPrinterCleaning #resinPrinterMaintenance #resinPrinterSafetyGear #resinPrintingAdvantages #resinPrintingAlternatives #resinPrintingBeginners #resinPrintingCons #resinPrintingConsumables #resinPrintingDetails #resinPrintingDisadvantages #resinPrintingDisposal #resinPrintingDurability #resinPrintingEssentials #resinPrintingForBeginners #resinPrintingForPrototyping #resinPrintingFumes #resinPrintingGloves #resinPrintingHealthPrecautions #resinPrintingHealthRisks #resinPrintingMess #resinPrintingMiniatures #resinPrintingPostCure #resinPrintingPostProcessing #resinPrintingPros #resinPrintingResinTypes #resinPrintingReviews #resinPrintingSafety #resinPrintingSetupTips #resinPrintingSmell #resinPrintingTechnology #resinPrintingTips #resinPrintingTipsAndTricks #resinPrintingVentilation #resinPrintingWasteManagement #resinPrintingWorkflow #resinPrintingWorkspace #resinVsFdm #toxicResinDangers #UVCuringResinPrints
-
Turning Saffron into Slop – Treylya Safran yn Skomblans
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
Note: Artificial Intelligence (AI) has come to be synonymous with Generative AI (GenAI) and with Large Language Models (LLMs), such as ChatGPT, in common parlance. Unless explicitly stated, I use the terms interchangeably.
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow*, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
LOW-RESOURCES AND LINGUISTIC TYPOLOGY
Simply adding a language to an AI model leads to a spike in poor-quality articles, drowning out quality writing by humans. AI has “industrialized the acts of destruction—which affect vulnerable languages most, since AI translations are typically far less reliable for them.”1 Wikipedia editors from varied languages evidence that machine translation tools have made it easier than ever before to create shoddy articles in minoritised languages, causing massive damage in minutes. AI leads to non-speakers producing much longer, truthier rubbish, Sámi computational linguistics expert Trond Trosterud notes: “the problem [is] that they are armed with Google Translate. Earlier they were armed only with dictionaries.”1
Kernewek, like all but 60 of the world’s roughly 7,000 languages, is designated “low-resource”, meaning it lacks sufficient data to train a machine.2 It is tempting, therefore, to assume that the solution is to provide more data. However, training an LLM requires petabytes of text, audio and video—manually categorised and in a machine-readable format—a vast trove that Kernewek simply does not have.3 Professor Will Lamb, Chair of Gaelic Ethnology and Linguistics at Edinburgh University, speaks of “millions of work hours devoted to just one aspect” of a working AI.4
Even if ChatGPT is trained on another language than English, the time and labour required may make it largely unviable. Current assessments of the performance of ChatGPT for different languages have shown that it performs worse in all tasks.5
Prof. Lina Dencik, Data Justice Lab
Furthermore, the amount of data and work is not the only barrier; at issue is the nature of the language itself. Microsoft has found that languages such as Breton—and thus Kernewek—cause a high rate of errors distinct from the size of their dataset, due to grammatical features, such as mutation, not present in well-sourced languages. As such, they remain poor without significant additional work.6 Essentially, simply adding more Kernewek may not help. Thus, engaging with AI is, for Kernewek, to tie ourselves to slop.
Noten: Skians Kreftus (SK) re dheuth ha bos kesstyr gans SK Dinythus (SKDin) ha gans Patronyow Yeth Bras (PYB), kepar ha ChatGPT, yn lavar kemmyn. Marnas bos menegys yn kler, my a us an termys yn keschanjyadow.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow*, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
ASNODHOW ISL HA TIPOLOGIETH YETHEL
Keworra yeth yn sempel orth patron SK a led orth spik yn erthyglow drog aga kwalita, ow peudhi skrif a gwalita gans tus. SK re wrug “diwysyansegi an aktys diswrians—hag a nas yethow goliadow an moyha, drefen bos treylyansow SK lieskweyth le lel yn tipek ragdha.”1 Golegydhyon Wikipedia a yethow divers a re dustuni re wrug medhelweyth-treylya y wul bos esya dell veu bythkweth kyns gwruthyl erthyglow pilyek yn yethow lyharivhes, ow kawsya damach kowrek yn mynysennow. SK a led orth digowsoryon owth askorra atal lieskweyth hirra ha gwirekka, konnyk yethonieth reknansek Sámi Trond Trosterud a not: “an kudyn [yw] aga bos ervys gans Google Translate. A-varra nyns ens ervys marnas gans gerlyvrow.”1
Kernewek, kepar hag oll marnas 60 a ogas lowr 7,000 yeth a’n bys, yw klassys avel “isel y asnodhow”, ow styrya nag eus dhodho kedhlow lowr dhe drenya jynn.2 Rakhenna, dynyek yw desevos bos an assoylyans profya moy a gedhlow. Byttegyns, res yw petavaytys a dekst, son ha gwydhyow—klassys dre leuv hag yn furvas redyadow gans jynn— dhe drenya PYB, tresorva efan nag eus dhe Gernewek yn sempel.3 Y kews Professor Will Lamb, Kaderyer Ethnologieth ha Yethonieth Wodhalek orth Pennskol Karedin, a “vilvilyow a ourys ober sakrys orth unn wedh hepken” a SK owth oberi.4
Hogen mars yw ChatGPT trenys war yeth a-der Sowsnek, an termyn hag ober yw res a styr y vos martesen anhewul dre vras. Arvreusyansow a-lemmyn a berformyans a ChatGPT rag yethow dyffrans re dhiskwedhas y perform gweth yn oberennow oll.5
Prof. Lina Dencik, Data Justice Lab
Pella, nyns yw an myns a gedhlow hag ober an unsel lett; a vern yw natur an yeth y honan. Microsoft re drovyas y kaws yethow kepar ha Bretonek—hag ytho Kernewek—kevradh ughel a wallow diblans a vraster aga sett kedhlow, drefen nasyow gramasek, kepar ha treylyansow, nag usi kevys yn yethow ughel aga asnodhow. Yndella, i a bes orth bos drog heb meur a ober keworransel.6 Yn essensek, possybyl yw ny wra keworra moy a Gernewek yn sempel gweres. Yndelma, oberi gans SK yw, rag Kernewek, omgelmi orth skomblans.
CORNISH UNDER CAPITALISM
But surely we can improve things over time? It will take a lot of help from AI companies, but it will be worth it. Sadly, Gabriel Nicholas, a research fellow at the Center for Democracy and Technology, has found that once a tech company has established basic capabilities for a language, they pat themselves on the back and move on.7
Big tech companies are just that: companies. They exist to make a profit. Unfortunately, a market dominated by big languages gives them no incentive to invest in improvements for small ones.
All of the speech technology, smart homes and voice interaction systems used today are the products of commercial research. To put it bluntly, they exist to either make money from your data, to sell you more goods and services, or to influence your thinking. None of this AI exists for the public good. […] Unless there is a strong enough economic argument, don’t expect big companies to rush into producing Welsh, Gaelic or Cornish speech systems.8
Prof. Ian McLoughlin, University of Kent
Should they decide that a Kernewek AI is a viable profit-making enterprise, our situation may even be worse than abandonment. As Dr. Fintan Mallory remarks, the dominant means of profit for privately-funded AI enterprises is to convert their tools into surveillance devices.9 As Kernewek is currently one of the UK’s only languages which is not currently easily surveillable, this poses a huge risk to Kernewek activism and the fight for self-determination in a state that seeks to criminalise dissent.
While we’re on the subject of Kernewek and its position under capitalism, let’s consider the human cost. I lost my 13-year career in language to AI as soon as English output became viable enough to excuse not paying a human. In the unlikely instance that we achieve an AI that can produce quality Kernewek, why would anyone bother paying speakers? The idea of AI sucking all the life out of my heritage language when we are struggling to survive as-is is appalling.
Simply put, profit is antithetical to people. While AI is the new favourite toy of profit, it will be antithetical to people. And a language is its people.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Combinations of characters on a screen mean nothing without agency and intention.10
Ross Perlin, Endangered Language Alliance
While language is not unique to humans, it is one of the chief parts of being human. It cannot be reduced to mere data, but is a highly social process.11 We all know how synthetic customer support via robot sounds or how AI fails to pick up nuance. As Dr. Mallory comments, “Language [is] something more like the soul of a community. You can’t store this in a machine. You can’t solve a human problem like linguicide with a view of language that removes the human component.”12
AI cannot comprehend Kernewek or any other language. It is a stochastic parrot: predicting what word is likely to follow the previous one.13 It cannot understand us. It cannot intend anything. If it tells you it feels delighted to help you, it is lying. I want our community to grow, but one hundred ‘Cornish-speaking’ computers do not add to it. One human does—bringing ideas and hopes and fears and foibles—and I do not think the Kernewek ‘speaking’ computers will add even one human to our community.
Worse, if it does, there is evidence from Microsoft to suggest that the use of GenAI on language tasks, even once a week, impairs cognitive ability to learn, leading to decreased engagement with the topic, overreliance on the technology and hobbled skills in independent problem-solving.14 By using AI tools to ‘teach’ a learner Kernewek, we may in fact be impairing their ability to learn the language at all without this crutch. We will make regurgitators in place of speakers.
Perlin also emphasises the human element, saying that when we hold community central to our languages, as we do, the stochastic parrot can feel like a violation.15 At the moment, I can tell when someone is using AI ‘Kernewek’ to me. The idea that one day I will not know when an outsider—someone I would welcome if they took up a book or a class—is puppeting my ancestors’ jaws and speaking through them is ghoulish. It has the instant sting of colonialism, of appropriation when one could appreciate, of parroting when one could join our chorus.
Hawai’ian scholar Ha‘alilio Solomon agrees: “It is painful, because it reminds us of all the times that our culture and language has been appropriated. We have been fighting tooth and nail in an uphill climb for language revitalization.[…] People are going to think that this is an accurate representation of the Hawaiian language.”16
TRUST AND COMMUNITY FEELING
The anti-machine backlash has long been simmering but is now seemingly breaking to the surface.17
NBC NEWS
The explosion of insults for AI itself (clanker, tinskin, toaster), its output (slop, dross, brainrot) and its users (slopper, groksucker, botlicker, second-hand thinker)—as well as others more clearly based on real-world slurs than I am comfortable to include—tells a tale of the general attitude of distrust and disgust towards the technology and its use on anglophone and other majority language internet.18 While the attitude among tech bros and corporates remains bombastic, for the general public AI is “becoming interchangeable with things that sort of suck.”19
Further, it’s not just majority languages with this negative view of AI as taint. A quick sampling of social media comments and likes regarding AI and Scottish Gaelic by Professor Lamb showed a split of 54% negative, 33% positive and 13% neutral. (Lamb, 2024) The sentiment of the top-rated negative comment was that AI is harmful and the second-highest that AI should be kept away from heritage languages.
What are we telling our descendants? That our language and culture isn’t worth the personal effort? That’s how I might read it, if I were them.20
Kernewek survey respondent
Kernewek paints an even starker picture, especially among younger and more technologically-savvy learners and speakers. A survey on Cornish Discord and Whatsapp found that 65% felt AI would be bad (11.5%) or very bad (53%) for the language. When asked what the community response should be to AI, 46% said we should prevent it and 27% avoid it, with only over-60s thinking that we should work with it.20
31% of respondents said using AI in Kernewek would cause them to feel estranged from the language, while 54% said that they would feel strongly estranged and 23% a little estranged from any organisation, resource or teacher using AI.
The response from those who gave their knowledge of AI as either “expert” or “good” was particularly damning. Everyone in this group responded that AI would be harmful for the language, that the use of AI would estrange them from a source strongly and that we should prevent the use of AI for Kernewek.
IDENTITY, AUTHENTICITY AND DIVERSITY
Aristotelis Ioannis Paschalidis, writing for UNESCO, was not speaking specifically about minoritised languages when he asked this, but the question resonates even more strongly for us: “How much loss of identity is one willing to sacrifice for efficiency?”21
Identity is of paramount importance to Kernewek speakers. Ute Wimmer’s study Reversing Language Shift: the Case of Cornish identified the language’s “function as a symbol of national identity” as the second highest motive (66%**) among speakers and learners, beaten only by Cornish culture (80%).22 This would seem cause for celebration, but when AI is added to the mix, it becomes a risk. Vincent Koc of Hyperlink states that AI can “inadvertently contribute to the dilution of language and cultural identity.”23
He also identifies that automating language learning or generation “may diminish the richness and authenticity that comes from human speakers who carry cultural histories in their speech.” Indeed, four studies by the University of Southern California have shown that using LLMs to assist writing “is linked to notable declines in linguistic diversity and may interfere with the societal and psychological insights language provides.”24
This is in English, one of the richest and largest languages in the world. Imagine the possible impact on a smaller language like Kernewek—with less documentation, less data, a tiny speakerbase and basically no money—and on its many language varieties and orthographies. Particular to the Kernewek context, Late speakers are already struggling to be seen as valid under the dominance of Middle. Do we think AI knows the difference? Thoughtlessly, it will either mix everything together, confusing everyone, or it will use Middle to overwhelm Late.
Generative AI-driven content creation, by favoring standardized languages, risks the disappearance of regional dialects.25
Barcelona supercomputing Center ….
Not only are varieties at risk; AI threatens to drown Kernewek as a whole. Perlin agrees that the linguistic flattening that occurred over centuries in English could manifest overnight in a minoritised language with AI at the helm—as it would be, being able to effortlessly outstrip human Kernewek. He raises concerns of LLMs freezing a language in place and even defining what it means to know the language, especially with low numbers of native speakers.26
Garbage translations multiply online like fake news. Native speakers of the languages in question are bypassed as being “too hard to find,” compared with automated methods of vetting that are completely disconnected from real-life communication. While larger and more powerful language communities may be able to hold the bots to account and even make strategic use of them, it is all too easy to imagine [a minority language] being overwhelmed.26
Ross Perlin, Endangered Language Alliance
Uncontrolled and in the hands of tech giants, synthetic Kernewek will outnumber and outmanoeuvre human Kernewek.
DATA SOVEREIGNTY AND COLONIALISM
Indigenous data sovereignty is the right of [an indigenous nation] to govern the collection, ownership, and application of its own data.27
Native Nations Institute
There are, however, indigenous cultures that are working on a more equitable relationship with AI. Tech without the giant requires resources, but it allows communities to retain data sovereignty over the cultural asset that is their language. Te Mana Raraunga, the Māori Data Sovereignty Network, has created a list of principles for the creation, use and sharing of Māori data, prioritising the need to enhance control for current and future Māori.
They raise a key point that should be considered carefully by stewards of linguistic and cultural knowledge: “Data from us, and about us and our resources, are valuable assets. Once control of it is lost, it is difficult to regain.”28 Decisions must not be taken lightly or hastily; we can always say “yes” if we have previously said “no” to a particular dataset’s use, but can never say “no” if we have already said “yes”.
The AI field, like any other space, is occupied by people who are set in their ways and unintentionally have a very colonial perspective.29
Michael Running Wolf, First Languages AI Reality
This is vital in the context of the potential control of Kernewek data by powerful external corporations. Capitalist extractivism has long been a bane on societies in the imperial periphery and our Cornish society is no different, having faced centuries of its wealth and natural resources being stripped and sold by and large for the profit of those outside Cornwall.
The book Indigenous Data Sovereignty and Policy notes that current data relations can be seen as “a continuation of the processes and underlying belief systems of extraction, exploitation, accumulation and dispossession that have been visited on Indigenous populations through historical colonialism.”30 This extractive understanding of information is, they note, not disrupted but rather replicated by paying people for their data.
Ultimately, our language must not lie in outside hands governed by proprietary principles that do not allow us sufficient sovereignty over one of our most valuable natural resources: our language. We must have open data principles, not bow to corporate control. We must steer and steward the use of our data, rather than expose it to use against our interests and for the pockets of big tech.
Rather than approaching language preservation as a technical problem, I think indigenous communities need to be politically empowered, whether that be funding from governments or legal protections to use their languages.31
Dr. Fintan Mallory, Durham University
We must prioritise language-as-community and seek open, equitable and ethical use of our language, heritage and other cultural assets. We must avoid thinking of AI as the magic that it promises and invest in basic research, driven by our own community. Corporations will not save us and, indeed, may do us great harm.
NO CORNISH ON A DEAD PLANET
Global capitalism and governments […] are addicted to ‘free’ market ideology over the wellbeing of communities, people and the planet.32
Cymdeithas yr Iaith Maniffesto 2022
Honestly, most takedowns of AI would have hit this point already. It’s one of the main arguments against Generative AI, but in case you’re not familiar with it, we will briefly look over the main points.
Water used in cooling AI data centers must be drinkable water. AI guzzles this water. The University of California has reported that “global water demand from AI could reach 4.2-6.6 billion cubic meters by 2027. That exceeds 50 percent of the UK’s annual water use in 2023.”33 All this while the Global Commission on the Economics of Water has declared “a rapidly accelerating water crisis” to which Kernewek should not be contributing.34
We have become utterly dependent on private technologies manufactured and controlled by a handful of opaque companies [who] appear mostly indifferent to the social consequences of their activities and only invest minimally if obliged by government regulations to enhance their public image.35
Iker Erdocia, Dublin City University
AI requires vast quantities of hardware at the cost of mining rare earth minerals. These are difficult to extract and purify and come with heavy environmental and social costs. They are often extracted from mines in countries with poorer environmental and labour protections. Reset states that “communities living near these mines, often indigenous or minority groups, regularly face land degradation, water contamination and human rights abuses. Much of this can be directly linked to the AI hardware.”36 When the hardware inevitably cooks and is useless, it is then thrown out as e-waste into poor communities. The potential advancement of Kernewek must not come at the expense of our sister indigenous and minority communities.
Training an also AI requires huge amounts of energy, soon perhaps as much as a small country37 and has an enormous carbon footprint.38 What is clear is that—through water usage, extractive industry, energy consumption and carbon footprint—AI is bad news for the struggling environment of the planet we live on and there is no Cornish on a dead planet.
MAKING AI AN EX-PARROT
Rather than making minority languages more accessible, AI is now creating an ever expanding minefield for students and speakers of those languages to navigate.39
mit technology review
We have heard of the vast improbability of getting AI to be able to mimic Kernewek in light of the costs in data, work, time and technology. We have considered the likely choice of cold negligence or surveillance product and the importance of data sovereignty. We have read about the effects on the livelihoods of Cornish speakers, as well as the the catastrophic costs to the environment and indigenous peoples.
We have learned that linguistic flattening by AI impoverishes its subjects and how AI may decide for us how our language must operate. We have seen the inescapability of language as human and the risks of creating ‘learners’ who cannot learn and ‘speakers’ who cannot speak. We have seen the dangers to reputation and trust for any organisation who would shovel what is seen as ‘slop’.
We have heard why giving in to the juggernaut of AI would be a mistake for Kernewek and how our community does not support our laying down of the shield. Instead, we must fight. We must make Kernewek a space as free of slop as possible, we must educate botlickers into ethical and effective language learning and use, we must avoid second-hand thinking.
We must make our language a no AI zone, a network of reliable humans and their human creations, built on authenticity, community, effort and trust: a Kernewek for the people, of the people and by the people.
KERNEWEK YN-DANN GEVALAV
Mes yn sur y hyllyn ni gwellhe taklow dres termyn? Y fydh res meur a weres a gompanis SK, mes y talvia dhyn. Yn trist, Gabriel Nicholas, kesvroder hwithrans orth an Center for Democracy and Technology, re drovyas pan wrug kompani tek fondya gallosow selyek rag unn yeth, i a omgeslowenha yn ughel hag ena movya yn-rag.7
Kompanis tek bras yw yndella poran: kompanis. Ymons i ena rag gwaynya budh. Y’n gwettha prys, ny wra marghas rewlys gans yethow bras ri kentryn dhe gevarghewi yn gwellhe rag an re byghan.
Oll a’n deknegieth kows, chiow konnyk ha systemow ynterweythres lev usys hedhyw yw an askorrasow a hwithrans kenwerthel. Dhe vos sogh, yth yns i po rag dendyl arghans a’th kedhlow, po gwertha gwara ha gonisyow, po delenwel dha dybyansow. Nyns yw tra vyth a’n SK ma rag an les kemmyn. […] Mar nag eus argyans erbysek krev lowr, na wra gwaytya kompanis bras dhe fyski dhe askorra systemow kows Kembrek, Godhalek po Kernewek.8
Prof. Ian McLoughlin, pennskol kint
Ha mars ervirons bos SK Kernewek aventur a yll gwaynya budh, possybyl yw bos agan studh gweth ages dell via gans forsakyans. Dell lever Dr. Fintan Mallor, an fordh vrassa a waynya budh rag kompanis SK arghesys yn privedh yw kedreylya aga thoulys yn devisyow aspians.9 Drefen bos Kernewek onan a’n yethow boghes y’n RU nag yw aspiadow yn es y’n eur ma, hemm yw peryl kowrek rag gweythresieth Kernewek ha’gan strif a-barth omdhetermyans yn stat a vynn galweythegi dissent.
Ha ni ow tochya Kernewek ha’y savla yn-dann gevalav, gwren ni mires orth an kost denel. My a gellis ow soodh 13 bloodh yn yethow dhe SK kettooth ha dell veu eskorrans Sowsnek hewul lowr dhe askusya sevel orth tyli den. Y’n kas diwirhaval may kevyn SK hag a yll askorra Kernewek da, prag y hwrussa nebonan omankombra ow pe kowser? An tybyans a SK ow tenna oll an bewnans a’m taves ertach ha ni ow kwynnel dhe dreusvewa dell on yw skruthus.
Yn sempel, budh yw gorthenebel orth tus. Hedre vo SK an degen nowydh flamm a vudh, y fydh gorthenebel orth tus. Ha yeth yw hy thus.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Nyns eus styr dhe gesunyansow a lytherennow war skrin heb dewis ha heb mynnas.10
Ross Perlin, Endangered Language Alliance
Kyn nag yw yeth dibarow dhe dhensys, onan a’n rannow chif a vos denel yw. Ny yll bos lehes dhe gedhlow hepken, mes yth yw argerdh sosyel dres eghen.11 Ni oll a wor py mar synthesek y sen skoodhyans prener der SK po fatel yll SK fyllel orth konvedhes arliwyow. Dell gampol Dr. Mallory, “Yeth [yw] neppyth moy kepar hag enev a gemeneth. Ny yllir gwitha hemma yn jynn. Ny yllir assoylya kudyn denel kepar ha yethladhans gans gwel a yeth hag a remov an gerann denel.”12
Ny yll SK konvedhes Kernewek po taves vyth aral. Papynjay chonsus yw: y targan py ger yw gwirhaval wosa an huni kyns.13 Ny yll agan konvedhes. Ny yll mynnes tra vyth. Mar kwra derivas orthis y vos pes da dha weres, gow yw. My a vynn agan kemeneth dhe devi, mes ny wra kans jynn-amontya a yll ‘kewsel Kernewek’ keworra orti. Y hwra unn den—ow tri tybyansow ha govenegow hag ownow ha gwanderyow—ha ny dybav y hwra an jynnys-amontya kernwegorek keworra unn den hogen orth agan kemeneth.
Gwettha, mar kwra, yma dustuni a-dhyworth Microsoft hag a brof y hwra an devnydh a SKDin war oberennow yeth, unweyth an seythen hogen, aperya gallos godhvosel a dhyski, ow ledya orth omworrans lehes gans an desten, gorfydhyans y’n deknegieth ha sleyneth sprallys a assoylya kudynnow yn anserghek.14 Der usya toulys SK dhe ‘dhyski’ Kernewek, possybyl yw ni dhe shyndya gallos dyski an yeth vytholl heb an kroch ma. Ni a wra gul mimyoryon yn le Kernewegoryon.
Ynwedh Perlin a boslev an elven dhenel, ow leverel pan wren ni synsi kemeneth avel kres agan yethow, dell wren, an papynjay chonsus a yll bos klewys kepar ha defolyans.15 Y’n eur ma, my a aswon pan eus nebonan owth usya ‘Kernewek’ SK dhymm. An tybyans ny wrav vy unn jydh godhvos pan eus estren—nebonan a wrussen vy dynerghi mar pe lyver po klass ganses—ow popettya diwawen ow hengerens ha kewsel dresta yw bedhrosus. Yma dhe’n dra an wan dhistowgh a drevesigeth, a berghenegyans pan yllir gwerthveurhe, a bapynjaya pan yllir junya agan kesgan.
Unver yw skolheyk Hawai’i henwys Noah Ha‘alilio Solomon: “Ankensi yw, drefen ni dhe vos kofhes a’n prysyow oll re beu agan gonisogeth ha yeth perghenegys. Ni re beu owth omladh dre dhens hag ewines yn batel gales a-barth dasvewheans yeth.[…] Y hwra pobel krysi bos hemma representyans ewn a’n yeth a Hawai’i.”16
TREST HAG OMGLEWANS AN GEMENETH
Hir re beu an kil-lash gorthjynn ow kovryjyon mes lemmyn yma va ow terri an arenep dell hevel.17
NBC NEWS
Tardh an arvedhennow rag SK y honan (clanker, tinskin, toaster), y askorras (slop, dross, brainrot) ha’y usyoryon (slopper, groksucker, botlicker, second-hand thinker)—keffrys hag erel selys moy yn kler war geryow kas gwir dell ov attes gans aga heworra—a re hwedhel a stons ollgemmyn a wogrys ha divlases war-tu hag an deknegieth ha’y devnydh war an kesrosweyth Sowsnek ha yethow bras erel.18 Kynth yw an stons yn-mysk gwesyon dek ha korforeth hwath gwresek, rag an boblek gemmyn y hwra SK “dos ha bos keschanjyadow gans taklow tamm kawgh.”19
Pella, nyns yw marnas yethow moyhariv gans an gwel negedhek ma a SK avel podrek. Sampel uskis a gampollow media sosyel ha meusi ow tochya SK ha Godhalek Alban gans Professor Lamb a dhiskwedhas fals a 54% negedhek, 33% posedhek ha 13% heptu. (Lamb, 2024) Sentiment an kampol negedhek an moyha talvesys o bos SK dregynnus hag an nessa y talvia dhyn lettya SK rag kestav gans tavosow ertach.
Pyth eson ni ow leverel orth agan diyskynysi? Ny dal agan yeth ha gonisogeth an strivyans personel? Hemm yw martesen fatel wrussen vy y redya, a pen vy i.20
Gorthebydh sondyans Kernewek
Kernewek a baynt aven moy serth, yn arbennik gans dyskoryon ha kowsoryon yowynka ha moy skentel gans tek. Sondyans war Discord ha Whatsapp Kernewek a drovyas bos 65% a grysis y fia SK drog (11.5%) po pur dhrog (53%) rag an yeth. Pan veu govynnys pyth a dal bos gorthyp an gemeneth orth SK, 46% a leveris y kodh y hedhi ha 27% y woheles, gans an dus moy ha 60 bloodh hepken ow tybi y kodh oberi ganso.20
31% a worthebydhyon an sondyans a leveris y hwrussa an devnydh a SK yn Kernewek aga fellhe a’n yeth, hag ynwedh 54% a leveris y fiens i pellhes yn krev ha 23% pellhes tamm a by kowethas, asnodh po dyskador pynag ow tevnydhya SK.
An gorthyp a’n re a leveris bos aga godhvos a SK po “konnyk” po “da” o dampnus yn arbennik. Pubonan y’n bagas ma a worthebis y fia SK dregynnus rag Kernewek, y hwrussa an devnydh a SK gans pennfenten aga fellhe a’n bennfenten na yn krev hag y kodh dhyn hedhi an devnydh a SK rag Kernewek.
HONANIETH, LELDER HA DIVERSETH
Nyns esa Aristotelis Ioannis Paschalidis, ow skrifa a-barth UNESCO, ow kewsel yn komparek a-dro dhe yethow lyharivhes pan wrug ev y wovyn, mes an govyn a dhassen yn kreffa ragon: “Pygemmys koll a honanieth a vynnir sakrifia rag effeythuster?”21
Honanieth yw a’n moyha bri rag Kernewegoryon. Studhyans Ute Wimmer Reversing Language Shift: the Case of Cornish a henow “gweythres [an yeth] avel arwodh a honanieth kenedhlek” avel an nessa ughella skila (66%**) yn-mysk kowsoryon ha dyskoryon, fethys gans gonisogeth Kernow (80%) hepken.22 Yth havalsa hemma bos acheson solempnyans, mes pan vo SK keworrys, y teu ha bos peryl. Vincent Koc a Hyperlink a lever y hyll SK “kevri dre wall orth an gwannheans a yeth ha honanieth wonisogethel”.23
Ev a aswon ynwedh y hallsa awtomategi dyski po dinythi yeth “lehe an rychedh ha lelder hag a dheu a gowsoryon dhenel neb a dheg istoriow gonisogethel y’ga hows”. Yn hwir, peswar studhyans gwrys gans Pennskol Kaliforni Soth re dhiskwedhas bos devnydhya PYB dhe weres gans skrifa “kelmys orth dyfygyansow nosedhek yn diverseth yethel hag y hyll mellya gans an konvedhes brysoniethel ha kowethasel yw proviys gans yeth.”24
Ha hemm yw yn Sowsnek, onan a’n yethow an ryccha ha brassa y’n bys. Dismyk an effeyth war yeth byghanna kepar ha Kernewek—gans le a dhogvennans, le a gedhlow, sel kowsoryon munys hag ogas hag arghans mann—ha war y lies orgraf hag eghen yeth. Yn arbennik yn gettesten Kernewek, seulabrys yma kowsoryon Diwedhes ow strivya dhe vos gwelys avel vas gans gwartheyvans Kres. A dybyn y hwor SK an dyffrans? Heb preder, y hwra po kemyska puptra warbarth, ow sowdheni pubonan, po devnydhya Kres dhe fetha Diwedhes.
An gwruthyl a dhalgh herdhys gans SK Dinythus, dre favera yethow savonegys, a argyl an vansyans a rannyethow ranndiryel.25
Kresen woramontyorieth Barcelona
Nyns yw eghennow hepken yn peryl; SK a wodros beudhi Kernewek yn tien. Akordys yw Perlin y hallsa an platheans yethel a hwarva dres kansbledhynnyow yn Sowsnek hwarvos dres nos yn yeth lyharivhes gans SK orth an fronnow—dell via, ow pos gallosek a bassya Kernewek denel heb assay. Ev a venek prederow yn kever PYB ow rewi yeth yn hy le ha hogen ow settya pyth yw an styr a wodhvos an yeth, yn arbennik gans niverow munys a gowsoryon deythyek.26
Treylyansow leun a atal a liesha warlinen kepar ha nowodhow fug. Kowsoryon deythyek a’n yethow ma yw passyes avel bos “re gales dhe drovya”, komparys orth fordhow awtomategys a surheans kwalita hag yw disjunys yn tien a geskomunyans y’n bys gwir. Kynth yw possybyl rag kemenethow yeth brassa ha moy gallosek synsi an bottys ma dhe akont ha’ga devnydhya yn stratejek hogen, re es yw dismygi [yeth lyhariv] ow pos reverthys.26
Ross Perlin, Endangered Language Alliance
Heb kontrol hag yn diwla an gewri deknegieth, Kernewek synthesek a wra gornivera ha gorthrabellhe Kernewek denel.
SOVRANEDH KEDHLOW HA KOLONEGIETH
Sovranedh kedhlow teythyek yw an gwir gans [kenedhel teythyek] a woverna an kuntel, perghenogeth ha gweytha a’y hedhlow hy honan.27
Native Nations Institute
Byttegyns, yma gonisogethow teythyek hag usi owth oberi war geskowethyans moy ewnhynsek gans SK. Tek heb an kowr a res asnodhow, mes y as kemenethow gwitha sovranedh kedhlow war an gerthen wonisogethel hag yw aga yeth. Te Mana Raraunga, Rosweyth Sovranedh Kedhlow Māori, re wrug rol a bennrewlys rag an gwruthyl, devnydhya ha kevrenna a gedhlow Māori, ow ragwirhe an edhom a grefhe maystri rag Māori a-lemmyn hag a dheu.
I a venek poynt posek hag a dalvia bos konsidrys gans rach gans stywards a skians yethel ha gonisogethel: “Kedhlow ahanan, a-dro dhyn ha’gan asnodhow, yw kerthennow a bris. Pan vo maystri kellys, kales yw y dhaskemeres.”28 Ny dal gul erviransow yn skav po yn uskis; y hyllyn pupprys leverel “ea” mar kwrussyn leverel “na” kyns orth us sett kedhlow, mes ny yllyn nevra leverel “na” mar kwrussyn leverel “ea” seulabrys.
An desten SK, kepar ha pub le aral, yw leun a dus hag yw settys y’ga maneryow ha gans gwel pur drevesigel yn tidowl.29
Michael Running Wolf, First Languages AI Reality
Hemm yw pur bosek y’n gettesten a’n kontrol possybyl a gedhlow Kernewek gans korforethow gallosek a-ves. Estenegieth jatelydhek re beu molleth war gowethasow y’n amal emperourethek ha nyns yw kowethas Kernewek dyffrans, wosa enebi kansvledhynnyow a’y rychys hag asnodhow naturek ow pos destryppys ha gwerthys dre vras gans budh tus yn-mes a Gernow.
An lyver henwys Indigenous Data Sovereignty and Policy a verk lemmyn y hyllir gweles perthynyansow kedhlow avel “pesyans a’n argerdhow ha systemow-krysi isworwedhek a estennans, drogusyans, kuntellyans ha diberghenogeth re beu gwrys war boblansow Teythyek dres trevesigeth istorek.”30 An konvedhes estennek ma a gedhlow yw, dell verkons, hevelebys a-der goderrys gans tyli pobel rag aga hedhlow.
Wostiwedh, res yw ma na vo agan yeth gorrys yn diwla a-ves routys gans pennrewlys perghenogel na as dhyn sovranedh lowr a onan a’gan asnodhow naturel an moyha posek: agan yeth. Res yw dhyn kavos pennrewlys kedhlow ygor, a-der plegya orth kontrol korforethel. Res yw dhyn lewya ha gidya an devnydh a’gan kedhlow, a-der y usya erbynn agan lesow ha rag pocketys tek bras.
A-der drehedhes an arwithans a davosow avel kudyn teknegiethel, my a dyb bos res dhe gemenethow teythyek bos reythhes yn politek, po der arghasans a wovernansow po dre dhifresyansow laghel dhe dhevnydhya aga yethow.31
Dr. Fintan Mallory, Pennskol Durham
Res yw dhyn ragwirhe yeth-avel-kemeneth ha hwilas devnydh ygor, ewnhynsek hag ethegel a’gan kerthennow yeth, ertach ha gonisogethel. Res yw dhyn goheles tybi a SK avel an hus mayth ambos ha kevarghewi yn hwithrans selyek, lewys gans agan kemeneth. Ny wra korforethow agan selwel ha, hogen, i a yll agan shyndya.
NYNS EUS KERNEWEK WAR BLANET MAROW
Governansow ha kevalav ollvysel […] yw omres dhe ideologieth marghas ‘rydh’ moy es dell yns omres dhe sewena kemenethow, pobel ha’n planet.32
Cymdeithas yr Iaith Maniffesto 2022
An brassa rann a vreusyansow a SK a wrussa meneges hemma seulabrys. Onan a’n argyansow brassa yw erbynn SK Dinythus, mes rag own bos ankoth dhis, ni a wra mires orth an chif boyntys.
Res yw bos evadow an dowr goyeynhe pub kresen kedhlow SK. Y kollenk an dowr ma. Pennskol Kaliforni re dherivas “y hallsa demond dowr ollvysel SK hedhes 4.2-6.6 bilvil metrow kubek erbynn 2027. Henn yw moy es 50 kansran a us dowr bledhynnyek an RU yn 2023.”33 Y kettermyn, an Desedhek Ollvysel Erbysieth Dowr a dheklaryas “barras dowr ow tardha yn uskis” ma na dal Kernewek kevri dhodho.34
Ni re dheuth ha bos yn hwir omres dhe deknegiethow privedh gwrys ha kontrolys gans dornas a gompanis diskler [hag] a hevel bos mygyl dre vras orth an sewyansow sosyel a’ga gwriansow ha kevri yn ispoyntel marnas mars yns i konstrinys gans rewlys an wovernans dhe wellhe aga imach poblek.35
Iker Erdocia, Pennskol Sita Dulyn
Yma edhom dhe SK a vynsow kowrek a galesweyth orth kost palas monyow tanow. Kales yw estenna ha purhe an re ma hag yma kostow kerghynedhel ha sosyel poos. Estennys yns i yn fenowgh a hwelyow yn powyow gans difresyansow lakka rag lavur ha’n kerghynnedh. Reset a lever “yn fenowgh y hwra kemenethow yw trigys yn ogas dhe’n hwelyow, yn fenowgh bagasow lyhariv po teythyek, enebi gwethheans an tir, defolyans an dowr hag abusyans gwiryow denel. Meur a hemma a yll bos kelmys yn tidro orth an galesweyth SK.”36 Pan yw an galesweyth kegys yn sertan hag euver, ena tewlys yw avel e-wast yn kemenethow boghosek. Res yw nyns yw an avonsyans possybyl a Gernewek orth kost agan kemenethow hwor lyhariv ha teythyek.
Ynwedh res yw myns hujes a nerth rag trenya SK, yn skon martesen an keth myns ha pow byghan37 hag yma ol troos karbon kowrek.38 Kler yw—der usadow dowr, diwysyans estennek, konsumyans nerth hag ol troos karbon—bos SK yeyn nowodhow rag kerghynnedh ow strivya a’n planet mayth on ni trigys warnodho ha nyns eus Kernewek war blanet marow.
GUL DHE SK BOS EKS-PAPYNJAY
A-der gul dhe yethow lyhariv bos moy hedhadow, lemmyn yma SK ow kwruthyl tardhek pupprys owth omlesa rag studhyoryon ha kowsoryon a’n yethow ma dhe wolya.39
mit technology review
Ni re glewas a’n anwirhevelepter efan a wul dhe SK gallos mimya Kernewek yn golow an kostys yn kedhlow, ober, termyn ha teknegieth. Ni re gonsidras lycklod an dewisynter dispresyans yeyn po askorras-aspia ha’n posekter a sovranedh kedhlow. Ni re redyas a-dro dhe’n effeythyow war vewnansow Kernewegoryon, keffrys ha’n kostys katastrofek rag an kerghynnedh ha poblow teythyek.
Ni re dhyskas y hwra platheans yethel gans SK boghosekhe y destennow ha fatel yll SK martesen ervira a’gan parth fatel godh dh’agan yeth oberi. Ni re welas an anwoheladewder a yeth avel denel ha’n peryllyow a wul ‘dyskoryon’ na yll dyski ha ‘kowsoryon’ na yll kewsel. Ni re welas an peryllyow orth bri ha fydhyans rag kowethasow a wrussa palas an pyth hag yw gwelys avel ‘skomblans’.
Ni re glewas prag y fia omblegya orth an jagganat a SK error rag Kernewek ha dell na vynn agan kemeneth skoodhya gorra an skoos a-dhyworthyn. Yn y le, res yw dhyn batalyas. Res yw dhyn gul dhe Gernewek bos spas mar rydh a skomblans dell yll bos, res yw adhyski orth botlapyoryon yn dyski ha devnydh yeth yn ethegel hag yn effeythus, res yw goheles tybi wortaswerth.
Res yw dhyn gul dh’agan yeth bos parth heb SK, rosweyth a dus fydhyadow ha’ga gwriansow denel, drehevys war lelder, kemeneth, assay ha trest: Kernewek hag yw a-barth an bobel, a’n bobel ha gans an bobel.
Niwlen Ster
Notennow
* A prime example is the laughably-unaffordable restaurant RenMor, which The Headland Hotel thinks is a version of “Re’n Mor”, which they believe means “by the sea” as in “next to the sea” but actually means “by the sea!” like saying “by Zeus!”. This is both hilarious and enraging.
** A figure perhaps lower than it should be if you consider that many of the “emotional motives” which were not counted in this category, such as “I’m Cornish, what better reason do you need?”, do also refer to identity.
FENTENNOW
1. Judah, J. (2025) How AI and Wikipedia have sent vulnerable languages into a doom spiral, MIT Technology Review.
2. Ackermann, A. (2023) When AI doesn’t speak your language, Coda.
3. Crichton, D. (2024) AI and the Death of Human Languages, Lux.
4. Lamb, W. (2024). Could Artificial Intelligence save Scottish Gaelic?, The University of Edinburgh.
5. Dencik, L. (/2025) AI Inequalities: Minority Languages, TUC Cymru.
6. Joshi, P., Santy, S., Budhiraja, A., Bali, K., & Microsoft Research, India. (2020). The State and Fate of Linguistic Diversity and Inclusion in the NLP World. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
7. Ackermann, A. (op cit)
8. McLoughlin, I. (2018) How to teach AI to speak Welsh (and other minority languages), The Conversation.
9. Mallory, F. (2025) RISE UP Panel Discussion & Q&A: What AI Can and Cannot Do for Minoritised Languages, YouTube.
10. Perlin, R. (2024) AI Won’t Protect Endangered Languages, The Dial.
11. RISE UP (2025) #4 RISE UP Event Summary: What AI Can and Cannot Do For Minoritised Languages, RISE UP.
12. Mallory, F. (2024) European Day of Languages: Will lesser spoken languages soon only be kept alive by AI technology? Durham University.
13. Bender, E., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
14. Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025) The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Microsoft.
15. Perlin, R. (op cit)
16. Judah, J. (op cit)
17. Abbruzzese, J., & Wile, R. (2025) Is an AI backlash brewing? What ‘clanker’ says about growing frustrations with emerging tech, NBC News.
18. Webster, K. (2025) Why Using ChatGPT at Work Could Hurt Your Reputation, Inc. Magazine.
19. Herrman, J. (2024) Is That AI? Or Does It Just Suck?, Intelligencer.
20. Wilson, L. (2025) Skians Kreftus ha Kernewek/Artificial Intelligence and Cornish
21. Paschalidis, A. I. (2025) AI and the great linguistic flattening, UNESCO.
22. Wimmer, U. (2010). Reversing Language Shift: the Case of Cornish. Cornish Language Board, p. 113
23. Koc, V. (2025) Generative AI and Large Language Models in Language Preservation: Opportunities and Challenges, ResearchGate.
24. Sourati, Z., Karimi-Malekabadi, F., & Ozcan, M. (2025) The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models, ResearchGate.
25. Melero, M. (2024) The Future of Language (and Cultural) Diversity in the Age of AI, CLARIN.
26. Perlin, R. (op cit)
27. Russo Carroll, S., Rodriguez Lonebear, D., & Martinez, A. (2017). Data Governance for Native Nation Rebuilding, Native Nations Institute.
28. Te Mana Raraunga. (2018). Frequently Asked Questions, Te Mana Raraunga.
29. Ackermann, A. (op cit)
30. Walter, M., Kukutai, T., Carroll, S. R., & Rodriguez-Lonebear, D. (Eds.). (2020). Indigenous Data Sovereignty and Policy. Taylor & Francis, p. 24
31. Mallory, F. (op cit)
32. Cymdeithas yr Iaith (2022) Cymru Rydd, Cymru Werdd, Cymru Gymraeg., p. 27
33. O’Sullivan, L. (2025). How AI’s Failure on Linguistic Diversity is Deepening Global Inequality, RESET – Digital for Good.
34. Harvey, F. (2024). Global water crisis leaves half of world food production at risk in next 25 years, The Guardian.
35. Erdocia, I., Migge, B., & Schneider, B. (2024). Language is not a data set—Why overcoming ideologies of dataism is more important than ever in the age of AI. Journal of Sociolinguistics, 28(5), p. 23
36. O’Sullivan, L. (op cit)
37. Erdenesanaa, D. (2023) A.I. Could Soon Need as Much Electricity as an Entire Country, The New York Times
38. Heikkilä, M. (2022) We’re getting a better idea of AI’s true carbon footprint, MIT Technology Review.
39. Judah, J. (op cit)#4 #AI #ArtificialIntelligence #Breus #Cornish #Cornwall #data #generativeAI #history #jynn #kedhlow #Kernewek #Kernow #Kernowek #LLM #machine #PYB #SK #SKDinythus #SkiansKreftus #Sordya
-
Turning Saffron into Slop – Treylya Safran yn Skomblans
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
Note: Artificial Intelligence (AI) has come to be synonymous with Generative AI (GenAI) and with Large Language Models (LLMs), such as ChatGPT, in common parlance. Unless explicitly stated, I use the terms interchangeably.
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow*, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
LOW-RESOURCES AND LINGUISTIC TYPOLOGY
Simply adding a language to an AI model leads to a spike in poor-quality articles, drowning out quality writing by humans. AI has “industrialized the acts of destruction—which affect vulnerable languages most, since AI translations are typically far less reliable for them.”1 Wikipedia editors from varied languages evidence that machine translation tools have made it easier than ever before to create shoddy articles in minoritised languages, causing massive damage in minutes. AI leads to non-speakers producing much longer, truthier rubbish, Sámi computational linguistics expert Trond Trosterud notes: “the problem [is] that they are armed with Google Translate. Earlier they were armed only with dictionaries.”1
Kernewek, like all but 60 of the world’s roughly 7,000 languages, is designated “low-resource”, meaning it lacks sufficient data to train a machine.2 It is tempting, therefore, to assume that the solution is to provide more data. However, training an LLM requires petabytes of text, audio and video—manually categorised and in a machine-readable format—a vast trove that Kernewek simply does not have.3 Professor Will Lamb, Chair of Gaelic Ethnology and Linguistics at Edinburgh University, speaks of “millions of work hours devoted to just one aspect” of a working AI.4
Even if ChatGPT is trained on another language than English, the time and labour required may make it largely unviable. Current assessments of the performance of ChatGPT for different languages have shown that it performs worse in all tasks.5
Prof. Lina Dencik, Data Justice Lab
Furthermore, the amount of data and work is not the only barrier; at issue is the nature of the language itself. Microsoft has found that languages such as Breton—and thus Kernewek—cause a high rate of errors distinct from the size of their dataset, due to grammatical features, such as mutation, not present in well-sourced languages. As such, they remain poor without significant additional work.6 Essentially, simply adding more Kernewek may not help. Thus, engaging with AI is, for Kernewek, to tie ourselves to slop.
Noten: Skians Kreftus (SK) re dheuth ha bos kesstyr gans SK Dinythus (SKDin) ha gans Patronyow Yeth Bras (PYB), kepar ha ChatGPT, yn lavar kemmyn. Marnas bos menegys yn kler, my a us an termys yn keschanjyadow.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow*, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
ASNODHOW ISL HA TIPOLOGIETH YETHEL
Keworra yeth yn sempel orth patron SK a led orth spik yn erthyglow drog aga kwalita, ow peudhi skrif a gwalita gans tus. SK re wrug “diwysyansegi an aktys diswrians—hag a nas yethow goliadow an moyha, drefen bos treylyansow SK lieskweyth le lel yn tipek ragdha.”1 Golegydhyon Wikipedia a yethow divers a re dustuni re wrug medhelweyth-treylya y wul bos esya dell veu bythkweth kyns gwruthyl erthyglow pilyek yn yethow lyharivhes, ow kawsya damach kowrek yn mynysennow. SK a led orth digowsoryon owth askorra atal lieskweyth hirra ha gwirekka, konnyk yethonieth reknansek Sámi Trond Trosterud a not: “an kudyn [yw] aga bos ervys gans Google Translate. A-varra nyns ens ervys marnas gans gerlyvrow.”1
Kernewek, kepar hag oll marnas 60 a ogas lowr 7,000 yeth a’n bys, yw klassys avel “isel y asnodhow”, ow styrya nag eus dhodho kedhlow lowr dhe drenya jynn.2 Rakhenna, dynyek yw desevos bos an assoylyans profya moy a gedhlow. Byttegyns, res yw petavaytys a dekst, son ha gwydhyow—klassys dre leuv hag yn furvas redyadow gans jynn— dhe drenya PYB, tresorva efan nag eus dhe Gernewek yn sempel.3 Y kews Professor Will Lamb, Kaderyer Ethnologieth ha Yethonieth Wodhalek orth Pennskol Karedin, a “vilvilyow a ourys ober sakrys orth unn wedh hepken” a SK owth oberi.4
Hogen mars yw ChatGPT trenys war yeth a-der Sowsnek, an termyn hag ober yw res a styr y vos martesen anhewul dre vras. Arvreusyansow a-lemmyn a berformyans a ChatGPT rag yethow dyffrans re dhiskwedhas y perform gweth yn oberennow oll.5
Prof. Lina Dencik, Data Justice Lab
Pella, nyns yw an myns a gedhlow hag ober an unsel lett; a vern yw natur an yeth y honan. Microsoft re drovyas y kaws yethow kepar ha Bretonek—hag ytho Kernewek—kevradh ughel a wallow diblans a vraster aga sett kedhlow, drefen nasyow gramasek, kepar ha treylyansow, nag usi kevys yn yethow ughel aga asnodhow. Yndella, i a bes orth bos drog heb meur a ober keworransel.6 Yn essensek, possybyl yw ny wra keworra moy a Gernewek yn sempel gweres. Yndelma, oberi gans SK yw, rag Kernewek, omgelmi orth skomblans.
CORNISH UNDER CAPITALISM
But surely we can improve things over time? It will take a lot of help from AI companies, but it will be worth it. Sadly, Gabriel Nicholas, a research fellow at the Center for Democracy and Technology, has found that once a tech company has established basic capabilities for a language, they pat themselves on the back and move on.7
Big tech companies are just that: companies. They exist to make a profit. Unfortunately, a market dominated by big languages gives them no incentive to invest in improvements for small ones.
All of the speech technology, smart homes and voice interaction systems used today are the products of commercial research. To put it bluntly, they exist to either make money from your data, to sell you more goods and services, or to influence your thinking. None of this AI exists for the public good. […] Unless there is a strong enough economic argument, don’t expect big companies to rush into producing Welsh, Gaelic or Cornish speech systems.8
Prof. Ian McLoughlin, University of Kent
Should they decide that a Kernewek AI is a viable profit-making enterprise, our situation may even be worse than abandonment. As Dr. Fintan Mallory remarks, the dominant means of profit for privately-funded AI enterprises is to convert their tools into surveillance devices.9 As Kernewek is currently one of the UK’s only languages which is not currently easily surveillable, this poses a huge risk to Kernewek activism and the fight for self-determination in a state that seeks to criminalise dissent.
While we’re on the subject of Kernewek and its position under capitalism, let’s consider the human cost. I lost my 13-year career in language to AI as soon as English output became viable enough to excuse not paying a human. In the unlikely instance that we achieve an AI that can produce quality Kernewek, why would anyone bother paying speakers? The idea of AI sucking all the life out of my heritage language when we are struggling to survive as-is is appalling.
Simply put, profit is antithetical to people. While AI is the new favourite toy of profit, it will be antithetical to people. And a language is its people.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Combinations of characters on a screen mean nothing without agency and intention.10
Ross Perlin, Endangered Language Alliance
While language is not unique to humans, it is one of the chief parts of being human. It cannot be reduced to mere data, but is a highly social process.11 We all know how synthetic customer support via robot sounds or how AI fails to pick up nuance. As Dr. Mallory comments, “Language [is] something more like the soul of a community. You can’t store this in a machine. You can’t solve a human problem like linguicide with a view of language that removes the human component.”12
AI cannot comprehend Kernewek or any other language. It is a stochastic parrot: predicting what word is likely to follow the previous one.13 It cannot understand us. It cannot intend anything. If it tells you it feels delighted to help you, it is lying. I want our community to grow, but one hundred ‘Cornish-speaking’ computers do not add to it. One human does—bringing ideas and hopes and fears and foibles—and I do not think the Kernewek ‘speaking’ computers will add even one human to our community.
Worse, if it does, there is evidence from Microsoft to suggest that the use of GenAI on language tasks, even once a week, impairs cognitive ability to learn, leading to decreased engagement with the topic, overreliance on the technology and hobbled skills in independent problem-solving.14 By using AI tools to ‘teach’ a learner Kernewek, we may in fact be impairing their ability to learn the language at all without this crutch. We will make regurgitators in place of speakers.
Perlin also emphasises the human element, saying that when we hold community central to our languages, as we do, the stochastic parrot can feel like a violation.15 At the moment, I can tell when someone is using AI ‘Kernewek’ to me. The idea that one day I will not know when an outsider—someone I would welcome if they took up a book or a class—is puppeting my ancestors’ jaws and speaking through them is ghoulish. It has the instant sting of colonialism, of appropriation when one could appreciate, of parroting when one could join our chorus.
Hawai’ian scholar Ha‘alilio Solomon agrees: “It is painful, because it reminds us of all the times that our culture and language has been appropriated. We have been fighting tooth and nail in an uphill climb for language revitalization.[…] People are going to think that this is an accurate representation of the Hawaiian language.”16
TRUST AND COMMUNITY FEELING
The anti-machine backlash has long been simmering but is now seemingly breaking to the surface.17
NBC NEWS
The explosion of insults for AI itself (clanker, tinskin, toaster), its output (slop, dross, brainrot) and its users (slopper, groksucker, botlicker, second-hand thinker)—as well as others more clearly based on real-world slurs than I am comfortable to include—tells a tale of the general attitude of distrust and disgust towards the technology and its use on anglophone and other majority language internet.18 While the attitude among tech bros and corporates remains bombastic, for the general public AI is “becoming interchangeable with things that sort of suck.”19
Further, it’s not just majority languages with this negative view of AI as taint. A quick sampling of social media comments and likes regarding AI and Scottish Gaelic by Professor Lamb showed a split of 54% negative, 33% positive and 13% neutral. (Lamb, 2024) The sentiment of the top-rated negative comment was that AI is harmful and the second-highest that AI should be kept away from heritage languages.
What are we telling our descendants? That our language and culture isn’t worth the personal effort? That’s how I might read it, if I were them.20
Kernewek survey respondent
Kernewek paints an even starker picture, especially among younger and more technologically-savvy learners and speakers. A survey on Cornish Discord and Whatsapp found that 65% felt AI would be bad (11.5%) or very bad (53%) for the language. When asked what the community response should be to AI, 46% said we should prevent it and 27% avoid it, with only over-60s thinking that we should work with it.20
31% of respondents said using AI in Kernewek would cause them to feel estranged from the language, while 54% said that they would feel strongly estranged and 23% a little estranged from any organisation, resource or teacher using AI.
The response from those who gave their knowledge of AI as either “expert” or “good” was particularly damning. Everyone in this group responded that AI would be harmful for the language, that the use of AI would estrange them from a source strongly and that we should prevent the use of AI for Kernewek.
IDENTITY, AUTHENTICITY AND DIVERSITY
Aristotelis Ioannis Paschalidis, writing for UNESCO, was not speaking specifically about minoritised languages when he asked this, but the question resonates even more strongly for us: “How much loss of identity is one willing to sacrifice for efficiency?”21
Identity is of paramount importance to Kernewek speakers. Ute Wimmer’s study Reversing Language Shift: the Case of Cornish identified the language’s “function as a symbol of national identity” as the second highest motive (66%**) among speakers and learners, beaten only by Cornish culture (80%).22 This would seem cause for celebration, but when AI is added to the mix, it becomes a risk. Vincent Koc of Hyperlink states that AI can “inadvertently contribute to the dilution of language and cultural identity.”23
He also identifies that automating language learning or generation “may diminish the richness and authenticity that comes from human speakers who carry cultural histories in their speech.” Indeed, four studies by the University of Southern California have shown that using LLMs to assist writing “is linked to notable declines in linguistic diversity and may interfere with the societal and psychological insights language provides.”24
This is in English, one of the richest and largest languages in the world. Imagine the possible impact on a smaller language like Kernewek—with less documentation, less data, a tiny speakerbase and basically no money—and on its many language varieties and orthographies. Particular to the Kernewek context, Late speakers are already struggling to be seen as valid under the dominance of Middle. Do we think AI knows the difference? Thoughtlessly, it will either mix everything together, confusing everyone, or it will use Middle to overwhelm Late.
Generative AI-driven content creation, by favoring standardized languages, risks the disappearance of regional dialects.25
Barcelona supercomputing Center ….
Not only are varieties at risk; AI threatens to drown Kernewek as a whole. Perlin agrees that the linguistic flattening that occurred over centuries in English could manifest overnight in a minoritised language with AI at the helm—as it would be, being able to effortlessly outstrip human Kernewek. He raises concerns of LLMs freezing a language in place and even defining what it means to know the language, especially with low numbers of native speakers.26
Garbage translations multiply online like fake news. Native speakers of the languages in question are bypassed as being “too hard to find,” compared with automated methods of vetting that are completely disconnected from real-life communication. While larger and more powerful language communities may be able to hold the bots to account and even make strategic use of them, it is all too easy to imagine [a minority language] being overwhelmed.26
Ross Perlin, Endangered Language Alliance
Uncontrolled and in the hands of tech giants, synthetic Kernewek will outnumber and outmanoeuvre human Kernewek.
DATA SOVEREIGNTY AND COLONIALISM
Indigenous data sovereignty is the right of [an indigenous nation] to govern the collection, ownership, and application of its own data.27
Native Nations Institute
There are, however, indigenous cultures that are working on a more equitable relationship with AI. Tech without the giant requires resources, but it allows communities to retain data sovereignty over the cultural asset that is their language. Te Mana Raraunga, the Māori Data Sovereignty Network, has created a list of principles for the creation, use and sharing of Māori data, prioritising the need to enhance control for current and future Māori.
They raise a key point that should be considered carefully by stewards of linguistic and cultural knowledge: “Data from us, and about us and our resources, are valuable assets. Once control of it is lost, it is difficult to regain.”28 Decisions must not be taken lightly or hastily; we can always say “yes” if we have previously said “no” to a particular dataset’s use, but can never say “no” if we have already said “yes”.
The AI field, like any other space, is occupied by people who are set in their ways and unintentionally have a very colonial perspective.29
Michael Running Wolf, First Languages AI Reality
This is vital in the context of the potential control of Kernewek data by powerful external corporations. Capitalist extractivism has long been a bane on societies in the imperial periphery and our Cornish society is no different, having faced centuries of its wealth and natural resources being stripped and sold by and large for the profit of those outside Cornwall.
The book Indigenous Data Sovereignty and Policy notes that current data relations can be seen as “a continuation of the processes and underlying belief systems of extraction, exploitation, accumulation and dispossession that have been visited on Indigenous populations through historical colonialism.”30 This extractive understanding of information is, they note, not disrupted but rather replicated by paying people for their data.
Ultimately, our language must not lie in outside hands governed by proprietary principles that do not allow us sufficient sovereignty over one of our most valuable natural resources: our language. We must have open data principles, not bow to corporate control. We must steer and steward the use of our data, rather than expose it to use against our interests and for the pockets of big tech.
Rather than approaching language preservation as a technical problem, I think indigenous communities need to be politically empowered, whether that be funding from governments or legal protections to use their languages.31
Dr. Fintan Mallory, Durham University
We must prioritise language-as-community and seek open, equitable and ethical use of our language, heritage and other cultural assets. We must avoid thinking of AI as the magic that it promises and invest in basic research, driven by our own community. Corporations will not save us and, indeed, may do us great harm.
NO CORNISH ON A DEAD PLANET
Global capitalism and governments […] are addicted to ‘free’ market ideology over the wellbeing of communities, people and the planet.32
Cymdeithas yr Iaith Maniffesto 2022
Honestly, most takedowns of AI would have hit this point already. It’s one of the main arguments against Generative AI, but in case you’re not familiar with it, we will briefly look over the main points.
Water used in cooling AI data centers must be drinkable water. AI guzzles this water. The University of California has reported that “global water demand from AI could reach 4.2-6.6 billion cubic meters by 2027. That exceeds 50 percent of the UK’s annual water use in 2023.”33 All this while the Global Commission on the Economics of Water has declared “a rapidly accelerating water crisis” to which Kernewek should not be contributing.34
We have become utterly dependent on private technologies manufactured and controlled by a handful of opaque companies [who] appear mostly indifferent to the social consequences of their activities and only invest minimally if obliged by government regulations to enhance their public image.35
Iker Erdocia, Dublin City University
AI requires vast quantities of hardware at the cost of mining rare earth minerals. These are difficult to extract and purify and come with heavy environmental and social costs. They are often extracted from mines in countries with poorer environmental and labour protections. Reset states that “communities living near these mines, often indigenous or minority groups, regularly face land degradation, water contamination and human rights abuses. Much of this can be directly linked to the AI hardware.”36 When the hardware inevitably cooks and is useless, it is then thrown out as e-waste into poor communities. The potential advancement of Kernewek must not come at the expense of our sister indigenous and minority communities.
Training an also AI requires huge amounts of energy, soon perhaps as much as a small country37 and has an enormous carbon footprint.38 What is clear is that—through water usage, extractive industry, energy consumption and carbon footprint—AI is bad news for the struggling environment of the planet we live on and there is no Cornish on a dead planet.
MAKING AI AN EX-PARROT
Rather than making minority languages more accessible, AI is now creating an ever expanding minefield for students and speakers of those languages to navigate.39
mit technology review
We have heard of the vast improbability of getting AI to be able to mimic Kernewek in light of the costs in data, work, time and technology. We have considered the likely choice of cold negligence or surveillance product and the importance of data sovereignty. We have read about the effects on the livelihoods of Cornish speakers, as well as the the catastrophic costs to the environment and indigenous peoples.
We have learned that linguistic flattening by AI impoverishes its subjects and how AI may decide for us how our language must operate. We have seen the inescapability of language as human and the risks of creating ‘learners’ who cannot learn and ‘speakers’ who cannot speak. We have seen the dangers to reputation and trust for any organisation who would shovel what is seen as ‘slop’.
We have heard why giving in to the juggernaut of AI would be a mistake for Kernewek and how our community does not support our laying down of the shield. Instead, we must fight. We must make Kernewek a space as free of slop as possible, we must educate botlickers into ethical and effective language learning and use, we must avoid second-hand thinking.
We must make our language a no AI zone, a network of reliable humans and their human creations, built on authenticity, community, effort and trust: a Kernewek for the people, of the people and by the people.
KERNEWEK YN-DANN GEVALAV
Mes yn sur y hyllyn ni gwellhe taklow dres termyn? Y fydh res meur a weres a gompanis SK, mes y talvia dhyn. Yn trist, Gabriel Nicholas, kesvroder hwithrans orth an Center for Democracy and Technology, re drovyas pan wrug kompani tek fondya gallosow selyek rag unn yeth, i a omgeslowenha yn ughel hag ena movya yn-rag.7
Kompanis tek bras yw yndella poran: kompanis. Ymons i ena rag gwaynya budh. Y’n gwettha prys, ny wra marghas rewlys gans yethow bras ri kentryn dhe gevarghewi yn gwellhe rag an re byghan.
Oll a’n deknegieth kows, chiow konnyk ha systemow ynterweythres lev usys hedhyw yw an askorrasow a hwithrans kenwerthel. Dhe vos sogh, yth yns i po rag dendyl arghans a’th kedhlow, po gwertha gwara ha gonisyow, po delenwel dha dybyansow. Nyns yw tra vyth a’n SK ma rag an les kemmyn. […] Mar nag eus argyans erbysek krev lowr, na wra gwaytya kompanis bras dhe fyski dhe askorra systemow kows Kembrek, Godhalek po Kernewek.8
Prof. Ian McLoughlin, pennskol kint
Ha mars ervirons bos SK Kernewek aventur a yll gwaynya budh, possybyl yw bos agan studh gweth ages dell via gans forsakyans. Dell lever Dr. Fintan Mallor, an fordh vrassa a waynya budh rag kompanis SK arghesys yn privedh yw kedreylya aga thoulys yn devisyow aspians.9 Drefen bos Kernewek onan a’n yethow boghes y’n RU nag yw aspiadow yn es y’n eur ma, hemm yw peryl kowrek rag gweythresieth Kernewek ha’gan strif a-barth omdhetermyans yn stat a vynn galweythegi dissent.
Ha ni ow tochya Kernewek ha’y savla yn-dann gevalav, gwren ni mires orth an kost denel. My a gellis ow soodh 13 bloodh yn yethow dhe SK kettooth ha dell veu eskorrans Sowsnek hewul lowr dhe askusya sevel orth tyli den. Y’n kas diwirhaval may kevyn SK hag a yll askorra Kernewek da, prag y hwrussa nebonan omankombra ow pe kowser? An tybyans a SK ow tenna oll an bewnans a’m taves ertach ha ni ow kwynnel dhe dreusvewa dell on yw skruthus.
Yn sempel, budh yw gorthenebel orth tus. Hedre vo SK an degen nowydh flamm a vudh, y fydh gorthenebel orth tus. Ha yeth yw hy thus.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Nyns eus styr dhe gesunyansow a lytherennow war skrin heb dewis ha heb mynnas.10
Ross Perlin, Endangered Language Alliance
Kyn nag yw yeth dibarow dhe dhensys, onan a’n rannow chif a vos denel yw. Ny yll bos lehes dhe gedhlow hepken, mes yth yw argerdh sosyel dres eghen.11 Ni oll a wor py mar synthesek y sen skoodhyans prener der SK po fatel yll SK fyllel orth konvedhes arliwyow. Dell gampol Dr. Mallory, “Yeth [yw] neppyth moy kepar hag enev a gemeneth. Ny yllir gwitha hemma yn jynn. Ny yllir assoylya kudyn denel kepar ha yethladhans gans gwel a yeth hag a remov an gerann denel.”12
Ny yll SK konvedhes Kernewek po taves vyth aral. Papynjay chonsus yw: y targan py ger yw gwirhaval wosa an huni kyns.13 Ny yll agan konvedhes. Ny yll mynnes tra vyth. Mar kwra derivas orthis y vos pes da dha weres, gow yw. My a vynn agan kemeneth dhe devi, mes ny wra kans jynn-amontya a yll ‘kewsel Kernewek’ keworra orti. Y hwra unn den—ow tri tybyansow ha govenegow hag ownow ha gwanderyow—ha ny dybav y hwra an jynnys-amontya kernwegorek keworra unn den hogen orth agan kemeneth.
Gwettha, mar kwra, yma dustuni a-dhyworth Microsoft hag a brof y hwra an devnydh a SKDin war oberennow yeth, unweyth an seythen hogen, aperya gallos godhvosel a dhyski, ow ledya orth omworrans lehes gans an desten, gorfydhyans y’n deknegieth ha sleyneth sprallys a assoylya kudynnow yn anserghek.14 Der usya toulys SK dhe ‘dhyski’ Kernewek, possybyl yw ni dhe shyndya gallos dyski an yeth vytholl heb an kroch ma. Ni a wra gul mimyoryon yn le Kernewegoryon.
Ynwedh Perlin a boslev an elven dhenel, ow leverel pan wren ni synsi kemeneth avel kres agan yethow, dell wren, an papynjay chonsus a yll bos klewys kepar ha defolyans.15 Y’n eur ma, my a aswon pan eus nebonan owth usya ‘Kernewek’ SK dhymm. An tybyans ny wrav vy unn jydh godhvos pan eus estren—nebonan a wrussen vy dynerghi mar pe lyver po klass ganses—ow popettya diwawen ow hengerens ha kewsel dresta yw bedhrosus. Yma dhe’n dra an wan dhistowgh a drevesigeth, a berghenegyans pan yllir gwerthveurhe, a bapynjaya pan yllir junya agan kesgan.
Unver yw skolheyk Hawai’i henwys Noah Ha‘alilio Solomon: “Ankensi yw, drefen ni dhe vos kofhes a’n prysyow oll re beu agan gonisogeth ha yeth perghenegys. Ni re beu owth omladh dre dhens hag ewines yn batel gales a-barth dasvewheans yeth.[…] Y hwra pobel krysi bos hemma representyans ewn a’n yeth a Hawai’i.”16
TREST HAG OMGLEWANS AN GEMENETH
Hir re beu an kil-lash gorthjynn ow kovryjyon mes lemmyn yma va ow terri an arenep dell hevel.17
NBC NEWS
Tardh an arvedhennow rag SK y honan (clanker, tinskin, toaster), y askorras (slop, dross, brainrot) ha’y usyoryon (slopper, groksucker, botlicker, second-hand thinker)—keffrys hag erel selys moy yn kler war geryow kas gwir dell ov attes gans aga heworra—a re hwedhel a stons ollgemmyn a wogrys ha divlases war-tu hag an deknegieth ha’y devnydh war an kesrosweyth Sowsnek ha yethow bras erel.18 Kynth yw an stons yn-mysk gwesyon dek ha korforeth hwath gwresek, rag an boblek gemmyn y hwra SK “dos ha bos keschanjyadow gans taklow tamm kawgh.”19
Pella, nyns yw marnas yethow moyhariv gans an gwel negedhek ma a SK avel podrek. Sampel uskis a gampollow media sosyel ha meusi ow tochya SK ha Godhalek Alban gans Professor Lamb a dhiskwedhas fals a 54% negedhek, 33% posedhek ha 13% heptu. (Lamb, 2024) Sentiment an kampol negedhek an moyha talvesys o bos SK dregynnus hag an nessa y talvia dhyn lettya SK rag kestav gans tavosow ertach.
Pyth eson ni ow leverel orth agan diyskynysi? Ny dal agan yeth ha gonisogeth an strivyans personel? Hemm yw martesen fatel wrussen vy y redya, a pen vy i.20
Gorthebydh sondyans Kernewek
Kernewek a baynt aven moy serth, yn arbennik gans dyskoryon ha kowsoryon yowynka ha moy skentel gans tek. Sondyans war Discord ha Whatsapp Kernewek a drovyas bos 65% a grysis y fia SK drog (11.5%) po pur dhrog (53%) rag an yeth. Pan veu govynnys pyth a dal bos gorthyp an gemeneth orth SK, 46% a leveris y kodh y hedhi ha 27% y woheles, gans an dus moy ha 60 bloodh hepken ow tybi y kodh oberi ganso.20
31% a worthebydhyon an sondyans a leveris y hwrussa an devnydh a SK yn Kernewek aga fellhe a’n yeth, hag ynwedh 54% a leveris y fiens i pellhes yn krev ha 23% pellhes tamm a by kowethas, asnodh po dyskador pynag ow tevnydhya SK.
An gorthyp a’n re a leveris bos aga godhvos a SK po “konnyk” po “da” o dampnus yn arbennik. Pubonan y’n bagas ma a worthebis y fia SK dregynnus rag Kernewek, y hwrussa an devnydh a SK gans pennfenten aga fellhe a’n bennfenten na yn krev hag y kodh dhyn hedhi an devnydh a SK rag Kernewek.
HONANIETH, LELDER HA DIVERSETH
Nyns esa Aristotelis Ioannis Paschalidis, ow skrifa a-barth UNESCO, ow kewsel yn komparek a-dro dhe yethow lyharivhes pan wrug ev y wovyn, mes an govyn a dhassen yn kreffa ragon: “Pygemmys koll a honanieth a vynnir sakrifia rag effeythuster?”21
Honanieth yw a’n moyha bri rag Kernewegoryon. Studhyans Ute Wimmer Reversing Language Shift: the Case of Cornish a henow “gweythres [an yeth] avel arwodh a honanieth kenedhlek” avel an nessa ughella skila (66%**) yn-mysk kowsoryon ha dyskoryon, fethys gans gonisogeth Kernow (80%) hepken.22 Yth havalsa hemma bos acheson solempnyans, mes pan vo SK keworrys, y teu ha bos peryl. Vincent Koc a Hyperlink a lever y hyll SK “kevri dre wall orth an gwannheans a yeth ha honanieth wonisogethel”.23
Ev a aswon ynwedh y hallsa awtomategi dyski po dinythi yeth “lehe an rychedh ha lelder hag a dheu a gowsoryon dhenel neb a dheg istoriow gonisogethel y’ga hows”. Yn hwir, peswar studhyans gwrys gans Pennskol Kaliforni Soth re dhiskwedhas bos devnydhya PYB dhe weres gans skrifa “kelmys orth dyfygyansow nosedhek yn diverseth yethel hag y hyll mellya gans an konvedhes brysoniethel ha kowethasel yw proviys gans yeth.”24
Ha hemm yw yn Sowsnek, onan a’n yethow an ryccha ha brassa y’n bys. Dismyk an effeyth war yeth byghanna kepar ha Kernewek—gans le a dhogvennans, le a gedhlow, sel kowsoryon munys hag ogas hag arghans mann—ha war y lies orgraf hag eghen yeth. Yn arbennik yn gettesten Kernewek, seulabrys yma kowsoryon Diwedhes ow strivya dhe vos gwelys avel vas gans gwartheyvans Kres. A dybyn y hwor SK an dyffrans? Heb preder, y hwra po kemyska puptra warbarth, ow sowdheni pubonan, po devnydhya Kres dhe fetha Diwedhes.
An gwruthyl a dhalgh herdhys gans SK Dinythus, dre favera yethow savonegys, a argyl an vansyans a rannyethow ranndiryel.25
Kresen woramontyorieth Barcelona
Nyns yw eghennow hepken yn peryl; SK a wodros beudhi Kernewek yn tien. Akordys yw Perlin y hallsa an platheans yethel a hwarva dres kansbledhynnyow yn Sowsnek hwarvos dres nos yn yeth lyharivhes gans SK orth an fronnow—dell via, ow pos gallosek a bassya Kernewek denel heb assay. Ev a venek prederow yn kever PYB ow rewi yeth yn hy le ha hogen ow settya pyth yw an styr a wodhvos an yeth, yn arbennik gans niverow munys a gowsoryon deythyek.26
Treylyansow leun a atal a liesha warlinen kepar ha nowodhow fug. Kowsoryon deythyek a’n yethow ma yw passyes avel bos “re gales dhe drovya”, komparys orth fordhow awtomategys a surheans kwalita hag yw disjunys yn tien a geskomunyans y’n bys gwir. Kynth yw possybyl rag kemenethow yeth brassa ha moy gallosek synsi an bottys ma dhe akont ha’ga devnydhya yn stratejek hogen, re es yw dismygi [yeth lyhariv] ow pos reverthys.26
Ross Perlin, Endangered Language Alliance
Heb kontrol hag yn diwla an gewri deknegieth, Kernewek synthesek a wra gornivera ha gorthrabellhe Kernewek denel.
SOVRANEDH KEDHLOW HA KOLONEGIETH
Sovranedh kedhlow teythyek yw an gwir gans [kenedhel teythyek] a woverna an kuntel, perghenogeth ha gweytha a’y hedhlow hy honan.27
Native Nations Institute
Byttegyns, yma gonisogethow teythyek hag usi owth oberi war geskowethyans moy ewnhynsek gans SK. Tek heb an kowr a res asnodhow, mes y as kemenethow gwitha sovranedh kedhlow war an gerthen wonisogethel hag yw aga yeth. Te Mana Raraunga, Rosweyth Sovranedh Kedhlow Māori, re wrug rol a bennrewlys rag an gwruthyl, devnydhya ha kevrenna a gedhlow Māori, ow ragwirhe an edhom a grefhe maystri rag Māori a-lemmyn hag a dheu.
I a venek poynt posek hag a dalvia bos konsidrys gans rach gans stywards a skians yethel ha gonisogethel: “Kedhlow ahanan, a-dro dhyn ha’gan asnodhow, yw kerthennow a bris. Pan vo maystri kellys, kales yw y dhaskemeres.”28 Ny dal gul erviransow yn skav po yn uskis; y hyllyn pupprys leverel “ea” mar kwrussyn leverel “na” kyns orth us sett kedhlow, mes ny yllyn nevra leverel “na” mar kwrussyn leverel “ea” seulabrys.
An desten SK, kepar ha pub le aral, yw leun a dus hag yw settys y’ga maneryow ha gans gwel pur drevesigel yn tidowl.29
Michael Running Wolf, First Languages AI Reality
Hemm yw pur bosek y’n gettesten a’n kontrol possybyl a gedhlow Kernewek gans korforethow gallosek a-ves. Estenegieth jatelydhek re beu molleth war gowethasow y’n amal emperourethek ha nyns yw kowethas Kernewek dyffrans, wosa enebi kansvledhynnyow a’y rychys hag asnodhow naturek ow pos destryppys ha gwerthys dre vras gans budh tus yn-mes a Gernow.
An lyver henwys Indigenous Data Sovereignty and Policy a verk lemmyn y hyllir gweles perthynyansow kedhlow avel “pesyans a’n argerdhow ha systemow-krysi isworwedhek a estennans, drogusyans, kuntellyans ha diberghenogeth re beu gwrys war boblansow Teythyek dres trevesigeth istorek.”30 An konvedhes estennek ma a gedhlow yw, dell verkons, hevelebys a-der goderrys gans tyli pobel rag aga hedhlow.
Wostiwedh, res yw ma na vo agan yeth gorrys yn diwla a-ves routys gans pennrewlys perghenogel na as dhyn sovranedh lowr a onan a’gan asnodhow naturel an moyha posek: agan yeth. Res yw dhyn kavos pennrewlys kedhlow ygor, a-der plegya orth kontrol korforethel. Res yw dhyn lewya ha gidya an devnydh a’gan kedhlow, a-der y usya erbynn agan lesow ha rag pocketys tek bras.
A-der drehedhes an arwithans a davosow avel kudyn teknegiethel, my a dyb bos res dhe gemenethow teythyek bos reythhes yn politek, po der arghasans a wovernansow po dre dhifresyansow laghel dhe dhevnydhya aga yethow.31
Dr. Fintan Mallory, Pennskol Durham
Res yw dhyn ragwirhe yeth-avel-kemeneth ha hwilas devnydh ygor, ewnhynsek hag ethegel a’gan kerthennow yeth, ertach ha gonisogethel. Res yw dhyn goheles tybi a SK avel an hus mayth ambos ha kevarghewi yn hwithrans selyek, lewys gans agan kemeneth. Ny wra korforethow agan selwel ha, hogen, i a yll agan shyndya.
NYNS EUS KERNEWEK WAR BLANET MAROW
Governansow ha kevalav ollvysel […] yw omres dhe ideologieth marghas ‘rydh’ moy es dell yns omres dhe sewena kemenethow, pobel ha’n planet.32
Cymdeithas yr Iaith Maniffesto 2022
An brassa rann a vreusyansow a SK a wrussa meneges hemma seulabrys. Onan a’n argyansow brassa yw erbynn SK Dinythus, mes rag own bos ankoth dhis, ni a wra mires orth an chif boyntys.
Res yw bos evadow an dowr goyeynhe pub kresen kedhlow SK. Y kollenk an dowr ma. Pennskol Kaliforni re dherivas “y hallsa demond dowr ollvysel SK hedhes 4.2-6.6 bilvil metrow kubek erbynn 2027. Henn yw moy es 50 kansran a us dowr bledhynnyek an RU yn 2023.”33 Y kettermyn, an Desedhek Ollvysel Erbysieth Dowr a dheklaryas “barras dowr ow tardha yn uskis” ma na dal Kernewek kevri dhodho.34
Ni re dheuth ha bos yn hwir omres dhe deknegiethow privedh gwrys ha kontrolys gans dornas a gompanis diskler [hag] a hevel bos mygyl dre vras orth an sewyansow sosyel a’ga gwriansow ha kevri yn ispoyntel marnas mars yns i konstrinys gans rewlys an wovernans dhe wellhe aga imach poblek.35
Iker Erdocia, Pennskol Sita Dulyn
Yma edhom dhe SK a vynsow kowrek a galesweyth orth kost palas monyow tanow. Kales yw estenna ha purhe an re ma hag yma kostow kerghynedhel ha sosyel poos. Estennys yns i yn fenowgh a hwelyow yn powyow gans difresyansow lakka rag lavur ha’n kerghynnedh. Reset a lever “yn fenowgh y hwra kemenethow yw trigys yn ogas dhe’n hwelyow, yn fenowgh bagasow lyhariv po teythyek, enebi gwethheans an tir, defolyans an dowr hag abusyans gwiryow denel. Meur a hemma a yll bos kelmys yn tidro orth an galesweyth SK.”36 Pan yw an galesweyth kegys yn sertan hag euver, ena tewlys yw avel e-wast yn kemenethow boghosek. Res yw nyns yw an avonsyans possybyl a Gernewek orth kost agan kemenethow hwor lyhariv ha teythyek.
Ynwedh res yw myns hujes a nerth rag trenya SK, yn skon martesen an keth myns ha pow byghan37 hag yma ol troos karbon kowrek.38 Kler yw—der usadow dowr, diwysyans estennek, konsumyans nerth hag ol troos karbon—bos SK yeyn nowodhow rag kerghynnedh ow strivya a’n planet mayth on ni trigys warnodho ha nyns eus Kernewek war blanet marow.
GUL DHE SK BOS EKS-PAPYNJAY
A-der gul dhe yethow lyhariv bos moy hedhadow, lemmyn yma SK ow kwruthyl tardhek pupprys owth omlesa rag studhyoryon ha kowsoryon a’n yethow ma dhe wolya.39
mit technology review
Ni re glewas a’n anwirhevelepter efan a wul dhe SK gallos mimya Kernewek yn golow an kostys yn kedhlow, ober, termyn ha teknegieth. Ni re gonsidras lycklod an dewisynter dispresyans yeyn po askorras-aspia ha’n posekter a sovranedh kedhlow. Ni re redyas a-dro dhe’n effeythyow war vewnansow Kernewegoryon, keffrys ha’n kostys katastrofek rag an kerghynnedh ha poblow teythyek.
Ni re dhyskas y hwra platheans yethel gans SK boghosekhe y destennow ha fatel yll SK martesen ervira a’gan parth fatel godh dh’agan yeth oberi. Ni re welas an anwoheladewder a yeth avel denel ha’n peryllyow a wul ‘dyskoryon’ na yll dyski ha ‘kowsoryon’ na yll kewsel. Ni re welas an peryllyow orth bri ha fydhyans rag kowethasow a wrussa palas an pyth hag yw gwelys avel ‘skomblans’.
Ni re glewas prag y fia omblegya orth an jagganat a SK error rag Kernewek ha dell na vynn agan kemeneth skoodhya gorra an skoos a-dhyworthyn. Yn y le, res yw dhyn batalyas. Res yw dhyn gul dhe Gernewek bos spas mar rydh a skomblans dell yll bos, res yw adhyski orth botlapyoryon yn dyski ha devnydh yeth yn ethegel hag yn effeythus, res yw goheles tybi wortaswerth.
Res yw dhyn gul dh’agan yeth bos parth heb SK, rosweyth a dus fydhyadow ha’ga gwriansow denel, drehevys war lelder, kemeneth, assay ha trest: Kernewek hag yw a-barth an bobel, a’n bobel ha gans an bobel.
Niwlen Ster
Notennow
* A prime example is the laughably-unaffordable restaurant RenMor, which The Headland Hotel thinks is a version of “Re’n Mor”, which they believe means “by the sea” as in “next to the sea” but actually means “by the sea!” like saying “by Zeus!”. This is both hilarious and enraging.
** A figure perhaps lower than it should be if you consider that many of the “emotional motives” which were not counted in this category, such as “I’m Cornish, what better reason do you need?”, do also refer to identity.
FENTENNOW
1. Judah, J. (2025) How AI and Wikipedia have sent vulnerable languages into a doom spiral, MIT Technology Review.
2. Ackermann, A. (2023) When AI doesn’t speak your language, Coda.
3. Crichton, D. (2024) AI and the Death of Human Languages, Lux.
4. Lamb, W. (2024). Could Artificial Intelligence save Scottish Gaelic?, The University of Edinburgh.
5. Dencik, L. (/2025) AI Inequalities: Minority Languages, TUC Cymru.
6. Joshi, P., Santy, S., Budhiraja, A., Bali, K., & Microsoft Research, India. (2020). The State and Fate of Linguistic Diversity and Inclusion in the NLP World. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
7. Ackermann, A. (op cit)
8. McLoughlin, I. (2018) How to teach AI to speak Welsh (and other minority languages), The Conversation.
9. Mallory, F. (2025) RISE UP Panel Discussion & Q&A: What AI Can and Cannot Do for Minoritised Languages, YouTube.
10. Perlin, R. (2024) AI Won’t Protect Endangered Languages, The Dial.
11. RISE UP (2025) #4 RISE UP Event Summary: What AI Can and Cannot Do For Minoritised Languages, RISE UP.
12. Mallory, F. (2024) European Day of Languages: Will lesser spoken languages soon only be kept alive by AI technology? Durham University.
13. Bender, E., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
14. Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025) The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Microsoft.
15. Perlin, R. (op cit)
16. Judah, J. (op cit)
17. Abbruzzese, J., & Wile, R. (2025) Is an AI backlash brewing? What ‘clanker’ says about growing frustrations with emerging tech, NBC News.
18. Webster, K. (2025) Why Using ChatGPT at Work Could Hurt Your Reputation, Inc. Magazine.
19. Herrman, J. (2024) Is That AI? Or Does It Just Suck?, Intelligencer.
20. Wilson, L. (2025) Skians Kreftus ha Kernewek/Artificial Intelligence and Cornish
21. Paschalidis, A. I. (2025) AI and the great linguistic flattening, UNESCO.
22. Wimmer, U. (2010). Reversing Language Shift: the Case of Cornish. Cornish Language Board, p. 113
23. Koc, V. (2025) Generative AI and Large Language Models in Language Preservation: Opportunities and Challenges, ResearchGate.
24. Sourati, Z., Karimi-Malekabadi, F., & Ozcan, M. (2025) The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models, ResearchGate.
25. Melero, M. (2024) The Future of Language (and Cultural) Diversity in the Age of AI, CLARIN.
26. Perlin, R. (op cit)
27. Russo Carroll, S., Rodriguez Lonebear, D., & Martinez, A. (2017). Data Governance for Native Nation Rebuilding, Native Nations Institute.
28. Te Mana Raraunga. (2018). Frequently Asked Questions, Te Mana Raraunga.
29. Ackermann, A. (op cit)
30. Walter, M., Kukutai, T., Carroll, S. R., & Rodriguez-Lonebear, D. (Eds.). (2020). Indigenous Data Sovereignty and Policy. Taylor & Francis, p. 24
31. Mallory, F. (op cit)
32. Cymdeithas yr Iaith (2022) Cymru Rydd, Cymru Werdd, Cymru Gymraeg., p. 27
33. O’Sullivan, L. (2025). How AI’s Failure on Linguistic Diversity is Deepening Global Inequality, RESET – Digital for Good.
34. Harvey, F. (2024). Global water crisis leaves half of world food production at risk in next 25 years, The Guardian.
35. Erdocia, I., Migge, B., & Schneider, B. (2024). Language is not a data set—Why overcoming ideologies of dataism is more important than ever in the age of AI. Journal of Sociolinguistics, 28(5), p. 23
36. O’Sullivan, L. (op cit)
37. Erdenesanaa, D. (2023) A.I. Could Soon Need as Much Electricity as an Entire Country, The New York Times
38. Heikkilä, M. (2022) We’re getting a better idea of AI’s true carbon footprint, MIT Technology Review.
39. Judah, J. (op cit)#4 #AI #ArtificialIntelligence #Breus #Cornish #Cornwall #data #generativeAI #history #jynn #kedhlow #Kernewek #Kernow #Kernowek #LLM #machine #PYB #SK #SKDinythus #SkiansKreftus #Sordya
-
Turning Saffron into Slop – Treylya Safran yn Skomblans
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
Note: Artificial Intelligence (AI) has come to be synonymous with Generative AI (GenAI) and with Large Language Models (LLMs), such as ChatGPT, in common parlance. Unless explicitly stated, I use the terms interchangeably.
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow*, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
LOW-RESOURCES AND LINGUISTIC TYPOLOGY
Simply adding a language to an AI model leads to a spike in poor-quality articles, drowning out quality writing by humans. AI has “industrialized the acts of destruction—which affect vulnerable languages most, since AI translations are typically far less reliable for them.”1 Wikipedia editors from varied languages evidence that machine translation tools have made it easier than ever before to create shoddy articles in minoritised languages, causing massive damage in minutes. AI leads to non-speakers producing much longer, truthier rubbish, Sámi computational linguistics expert Trond Trosterud notes: “the problem [is] that they are armed with Google Translate. Earlier they were armed only with dictionaries.”1
Kernewek, like all but 60 of the world’s roughly 7,000 languages, is designated “low-resource”, meaning it lacks sufficient data to train a machine.2 It is tempting, therefore, to assume that the solution is to provide more data. However, training an LLM requires petabytes of text, audio and video—manually categorised and in a machine-readable format—a vast trove that Kernewek simply does not have.3 Professor Will Lamb, Chair of Gaelic Ethnology and Linguistics at Edinburgh University, speaks of “millions of work hours devoted to just one aspect” of a working AI.4
Even if ChatGPT is trained on another language than English, the time and labour required may make it largely unviable. Current assessments of the performance of ChatGPT for different languages have shown that it performs worse in all tasks.5
Prof. Lina Dencik, Data Justice Lab
Furthermore, the amount of data and work is not the only barrier; at issue is the nature of the language itself. Microsoft has found that languages such as Breton—and thus Kernewek—cause a high rate of errors distinct from the size of their dataset, due to grammatical features, such as mutation, not present in well-sourced languages. As such, they remain poor without significant additional work.6 Essentially, simply adding more Kernewek may not help. Thus, engaging with AI is, for Kernewek, to tie ourselves to slop.
Noten: Skians Kreftus (SK) re dheuth ha bos kesstyr gans SK Dinythus (SKDin) ha gans Patronyow Yeth Bras (PYB), kepar ha ChatGPT, yn lavar kemmyn. Marnas bos menegys yn kler, my a us an termys yn keschanjyadow.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow*, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
ASNODHOW ISL HA TIPOLOGIETH YETHEL
Keworra yeth yn sempel orth patron SK a led orth spik yn erthyglow drog aga kwalita, ow peudhi skrif a gwalita gans tus. SK re wrug “diwysyansegi an aktys diswrians—hag a nas yethow goliadow an moyha, drefen bos treylyansow SK lieskweyth le lel yn tipek ragdha.”1 Golegydhyon Wikipedia a yethow divers a re dustuni re wrug medhelweyth-treylya y wul bos esya dell veu bythkweth kyns gwruthyl erthyglow pilyek yn yethow lyharivhes, ow kawsya damach kowrek yn mynysennow. SK a led orth digowsoryon owth askorra atal lieskweyth hirra ha gwirekka, konnyk yethonieth reknansek Sámi Trond Trosterud a not: “an kudyn [yw] aga bos ervys gans Google Translate. A-varra nyns ens ervys marnas gans gerlyvrow.”1
Kernewek, kepar hag oll marnas 60 a ogas lowr 7,000 yeth a’n bys, yw klassys avel “isel y asnodhow”, ow styrya nag eus dhodho kedhlow lowr dhe drenya jynn.2 Rakhenna, dynyek yw desevos bos an assoylyans profya moy a gedhlow. Byttegyns, res yw petavaytys a dekst, son ha gwydhyow—klassys dre leuv hag yn furvas redyadow gans jynn— dhe drenya PYB, tresorva efan nag eus dhe Gernewek yn sempel.3 Y kews Professor Will Lamb, Kaderyer Ethnologieth ha Yethonieth Wodhalek orth Pennskol Karedin, a “vilvilyow a ourys ober sakrys orth unn wedh hepken” a SK owth oberi.4
Hogen mars yw ChatGPT trenys war yeth a-der Sowsnek, an termyn hag ober yw res a styr y vos martesen anhewul dre vras. Arvreusyansow a-lemmyn a berformyans a ChatGPT rag yethow dyffrans re dhiskwedhas y perform gweth yn oberennow oll.5
Prof. Lina Dencik, Data Justice Lab
Pella, nyns yw an myns a gedhlow hag ober an unsel lett; a vern yw natur an yeth y honan. Microsoft re drovyas y kaws yethow kepar ha Bretonek—hag ytho Kernewek—kevradh ughel a wallow diblans a vraster aga sett kedhlow, drefen nasyow gramasek, kepar ha treylyansow, nag usi kevys yn yethow ughel aga asnodhow. Yndella, i a bes orth bos drog heb meur a ober keworransel.6 Yn essensek, possybyl yw ny wra keworra moy a Gernewek yn sempel gweres. Yndelma, oberi gans SK yw, rag Kernewek, omgelmi orth skomblans.
CORNISH UNDER CAPITALISM
But surely we can improve things over time? It will take a lot of help from AI companies, but it will be worth it. Sadly, Gabriel Nicholas, a research fellow at the Center for Democracy and Technology, has found that once a tech company has established basic capabilities for a language, they pat themselves on the back and move on.7
Big tech companies are just that: companies. They exist to make a profit. Unfortunately, a market dominated by big languages gives them no incentive to invest in improvements for small ones.
All of the speech technology, smart homes and voice interaction systems used today are the products of commercial research. To put it bluntly, they exist to either make money from your data, to sell you more goods and services, or to influence your thinking. None of this AI exists for the public good. […] Unless there is a strong enough economic argument, don’t expect big companies to rush into producing Welsh, Gaelic or Cornish speech systems.8
Prof. Ian McLoughlin, University of Kent
Should they decide that a Kernewek AI is a viable profit-making enterprise, our situation may even be worse than abandonment. As Dr. Fintan Mallory remarks, the dominant means of profit for privately-funded AI enterprises is to convert their tools into surveillance devices.9 As Kernewek is currently one of the UK’s only languages which is not currently easily surveillable, this poses a huge risk to Kernewek activism and the fight for self-determination in a state that seeks to criminalise dissent.
While we’re on the subject of Kernewek and its position under capitalism, let’s consider the human cost. I lost my 13-year career in language to AI as soon as English output became viable enough to excuse not paying a human. In the unlikely instance that we achieve an AI that can produce quality Kernewek, why would anyone bother paying speakers? The idea of AI sucking all the life out of my heritage language when we are struggling to survive as-is is appalling.
Simply put, profit is antithetical to people. While AI is the new favourite toy of profit, it will be antithetical to people. And a language is its people.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Combinations of characters on a screen mean nothing without agency and intention.10
Ross Perlin, Endangered Language Alliance
While language is not unique to humans, it is one of the chief parts of being human. It cannot be reduced to mere data, but is a highly social process.11 We all know how synthetic customer support via robot sounds or how AI fails to pick up nuance. As Dr. Mallory comments, “Language [is] something more like the soul of a community. You can’t store this in a machine. You can’t solve a human problem like linguicide with a view of language that removes the human component.”12
AI cannot comprehend Kernewek or any other language. It is a stochastic parrot: predicting what word is likely to follow the previous one.13 It cannot understand us. It cannot intend anything. If it tells you it feels delighted to help you, it is lying. I want our community to grow, but one hundred ‘Cornish-speaking’ computers do not add to it. One human does—bringing ideas and hopes and fears and foibles—and I do not think the Kernewek ‘speaking’ computers will add even one human to our community.
Worse, if it does, there is evidence from Microsoft to suggest that the use of GenAI on language tasks, even once a week, impairs cognitive ability to learn, leading to decreased engagement with the topic, overreliance on the technology and hobbled skills in independent problem-solving.14 By using AI tools to ‘teach’ a learner Kernewek, we may in fact be impairing their ability to learn the language at all without this crutch. We will make regurgitators in place of speakers.
Perlin also emphasises the human element, saying that when we hold community central to our languages, as we do, the stochastic parrot can feel like a violation.15 At the moment, I can tell when someone is using AI ‘Kernewek’ to me. The idea that one day I will not know when an outsider—someone I would welcome if they took up a book or a class—is puppeting my ancestors’ jaws and speaking through them is ghoulish. It has the instant sting of colonialism, of appropriation when one could appreciate, of parroting when one could join our chorus.
Hawai’ian scholar Ha‘alilio Solomon agrees: “It is painful, because it reminds us of all the times that our culture and language has been appropriated. We have been fighting tooth and nail in an uphill climb for language revitalization.[…] People are going to think that this is an accurate representation of the Hawaiian language.”16
TRUST AND COMMUNITY FEELING
The anti-machine backlash has long been simmering but is now seemingly breaking to the surface.17
NBC NEWS
The explosion of insults for AI itself (clanker, tinskin, toaster), its output (slop, dross, brainrot) and its users (slopper, groksucker, botlicker, second-hand thinker)—as well as others more clearly based on real-world slurs than I am comfortable to include—tells a tale of the general attitude of distrust and disgust towards the technology and its use on anglophone and other majority language internet.18 While the attitude among tech bros and corporates remains bombastic, for the general public AI is “becoming interchangeable with things that sort of suck.”19
Further, it’s not just majority languages with this negative view of AI as taint. A quick sampling of social media comments and likes regarding AI and Scottish Gaelic by Professor Lamb showed a split of 54% negative, 33% positive and 13% neutral. (Lamb, 2024) The sentiment of the top-rated negative comment was that AI is harmful and the second-highest that AI should be kept away from heritage languages.
What are we telling our descendants? That our language and culture isn’t worth the personal effort? That’s how I might read it, if I were them.20
Kernewek survey respondent
Kernewek paints an even starker picture, especially among younger and more technologically-savvy learners and speakers. A survey on Cornish Discord and Whatsapp found that 65% felt AI would be bad (11.5%) or very bad (53%) for the language. When asked what the community response should be to AI, 46% said we should prevent it and 27% avoid it, with only over-60s thinking that we should work with it.20
31% of respondents said using AI in Kernewek would cause them to feel estranged from the language, while 54% said that they would feel strongly estranged and 23% a little estranged from any organisation, resource or teacher using AI.
The response from those who gave their knowledge of AI as either “expert” or “good” was particularly damning. Everyone in this group responded that AI would be harmful for the language, that the use of AI would estrange them from a source strongly and that we should prevent the use of AI for Kernewek.
IDENTITY, AUTHENTICITY AND DIVERSITY
Aristotelis Ioannis Paschalidis, writing for UNESCO, was not speaking specifically about minoritised languages when he asked this, but the question resonates even more strongly for us: “How much loss of identity is one willing to sacrifice for efficiency?”21
Identity is of paramount importance to Kernewek speakers. Ute Wimmer’s study Reversing Language Shift: the Case of Cornish identified the language’s “function as a symbol of national identity” as the second highest motive (66%**) among speakers and learners, beaten only by Cornish culture (80%).22 This would seem cause for celebration, but when AI is added to the mix, it becomes a risk. Vincent Koc of Hyperlink states that AI can “inadvertently contribute to the dilution of language and cultural identity.”23
He also identifies that automating language learning or generation “may diminish the richness and authenticity that comes from human speakers who carry cultural histories in their speech.” Indeed, four studies by the University of Southern California have shown that using LLMs to assist writing “is linked to notable declines in linguistic diversity and may interfere with the societal and psychological insights language provides.”24
This is in English, one of the richest and largest languages in the world. Imagine the possible impact on a smaller language like Kernewek—with less documentation, less data, a tiny speakerbase and basically no money—and on its many language varieties and orthographies. Particular to the Kernewek context, Late speakers are already struggling to be seen as valid under the dominance of Middle. Do we think AI knows the difference? Thoughtlessly, it will either mix everything together, confusing everyone, or it will use Middle to overwhelm Late.
Generative AI-driven content creation, by favoring standardized languages, risks the disappearance of regional dialects.25
Barcelona supercomputing Center ….
Not only are varieties at risk; AI threatens to drown Kernewek as a whole. Perlin agrees that the linguistic flattening that occurred over centuries in English could manifest overnight in a minoritised language with AI at the helm—as it would be, being able to effortlessly outstrip human Kernewek. He raises concerns of LLMs freezing a language in place and even defining what it means to know the language, especially with low numbers of native speakers.26
Garbage translations multiply online like fake news. Native speakers of the languages in question are bypassed as being “too hard to find,” compared with automated methods of vetting that are completely disconnected from real-life communication. While larger and more powerful language communities may be able to hold the bots to account and even make strategic use of them, it is all too easy to imagine [a minority language] being overwhelmed.26
Ross Perlin, Endangered Language Alliance
Uncontrolled and in the hands of tech giants, synthetic Kernewek will outnumber and outmanoeuvre human Kernewek.
DATA SOVEREIGNTY AND COLONIALISM
Indigenous data sovereignty is the right of [an indigenous nation] to govern the collection, ownership, and application of its own data.27
Native Nations Institute
There are, however, indigenous cultures that are working on a more equitable relationship with AI. Tech without the giant requires resources, but it allows communities to retain data sovereignty over the cultural asset that is their language. Te Mana Raraunga, the Māori Data Sovereignty Network, has created a list of principles for the creation, use and sharing of Māori data, prioritising the need to enhance control for current and future Māori.
They raise a key point that should be considered carefully by stewards of linguistic and cultural knowledge: “Data from us, and about us and our resources, are valuable assets. Once control of it is lost, it is difficult to regain.”28 Decisions must not be taken lightly or hastily; we can always say “yes” if we have previously said “no” to a particular dataset’s use, but can never say “no” if we have already said “yes”.
The AI field, like any other space, is occupied by people who are set in their ways and unintentionally have a very colonial perspective.29
Michael Running Wolf, First Languages AI Reality
This is vital in the context of the potential control of Kernewek data by powerful external corporations. Capitalist extractivism has long been a bane on societies in the imperial periphery and our Cornish society is no different, having faced centuries of its wealth and natural resources being stripped and sold by and large for the profit of those outside Cornwall.
The book Indigenous Data Sovereignty and Policy notes that current data relations can be seen as “a continuation of the processes and underlying belief systems of extraction, exploitation, accumulation and dispossession that have been visited on Indigenous populations through historical colonialism.”30 This extractive understanding of information is, they note, not disrupted but rather replicated by paying people for their data.
Ultimately, our language must not lie in outside hands governed by proprietary principles that do not allow us sufficient sovereignty over one of our most valuable natural resources: our language. We must have open data principles, not bow to corporate control. We must steer and steward the use of our data, rather than expose it to use against our interests and for the pockets of big tech.
Rather than approaching language preservation as a technical problem, I think indigenous communities need to be politically empowered, whether that be funding from governments or legal protections to use their languages.31
Dr. Fintan Mallory, Durham University
We must prioritise language-as-community and seek open, equitable and ethical use of our language, heritage and other cultural assets. We must avoid thinking of AI as the magic that it promises and invest in basic research, driven by our own community. Corporations will not save us and, indeed, may do us great harm.
NO CORNISH ON A DEAD PLANET
Global capitalism and governments […] are addicted to ‘free’ market ideology over the wellbeing of communities, people and the planet.32
Cymdeithas yr Iaith Maniffesto 2022
Honestly, most takedowns of AI would have hit this point already. It’s one of the main arguments against Generative AI, but in case you’re not familiar with it, we will briefly look over the main points.
Water used in cooling AI data centers must be drinkable water. AI guzzles this water. The University of California has reported that “global water demand from AI could reach 4.2-6.6 billion cubic meters by 2027. That exceeds 50 percent of the UK’s annual water use in 2023.”33 All this while the Global Commission on the Economics of Water has declared “a rapidly accelerating water crisis” to which Kernewek should not be contributing.34
We have become utterly dependent on private technologies manufactured and controlled by a handful of opaque companies [who] appear mostly indifferent to the social consequences of their activities and only invest minimally if obliged by government regulations to enhance their public image.35
Iker Erdocia, Dublin City University
AI requires vast quantities of hardware at the cost of mining rare earth minerals. These are difficult to extract and purify and come with heavy environmental and social costs. They are often extracted from mines in countries with poorer environmental and labour protections. Reset states that “communities living near these mines, often indigenous or minority groups, regularly face land degradation, water contamination and human rights abuses. Much of this can be directly linked to the AI hardware.”36 When the hardware inevitably cooks and is useless, it is then thrown out as e-waste into poor communities. The potential advancement of Kernewek must not come at the expense of our sister indigenous and minority communities.
Training an also AI requires huge amounts of energy, soon perhaps as much as a small country37 and has an enormous carbon footprint.38 What is clear is that—through water usage, extractive industry, energy consumption and carbon footprint—AI is bad news for the struggling environment of the planet we live on and there is no Cornish on a dead planet.
MAKING AI AN EX-PARROT
Rather than making minority languages more accessible, AI is now creating an ever expanding minefield for students and speakers of those languages to navigate.39
mit technology review
We have heard of the vast improbability of getting AI to be able to mimic Kernewek in light of the costs in data, work, time and technology. We have considered the likely choice of cold negligence or surveillance product and the importance of data sovereignty. We have read about the effects on the livelihoods of Cornish speakers, as well as the the catastrophic costs to the environment and indigenous peoples.
We have learned that linguistic flattening by AI impoverishes its subjects and how AI may decide for us how our language must operate. We have seen the inescapability of language as human and the risks of creating ‘learners’ who cannot learn and ‘speakers’ who cannot speak. We have seen the dangers to reputation and trust for any organisation who would shovel what is seen as ‘slop’.
We have heard why giving in to the juggernaut of AI would be a mistake for Kernewek and how our community does not support our laying down of the shield. Instead, we must fight. We must make Kernewek a space as free of slop as possible, we must educate botlickers into ethical and effective language learning and use, we must avoid second-hand thinking.
We must make our language a no AI zone, a network of reliable humans and their human creations, built on authenticity, community, effort and trust: a Kernewek for the people, of the people and by the people.
KERNEWEK YN-DANN GEVALAV
Mes yn sur y hyllyn ni gwellhe taklow dres termyn? Y fydh res meur a weres a gompanis SK, mes y talvia dhyn. Yn trist, Gabriel Nicholas, kesvroder hwithrans orth an Center for Democracy and Technology, re drovyas pan wrug kompani tek fondya gallosow selyek rag unn yeth, i a omgeslowenha yn ughel hag ena movya yn-rag.7
Kompanis tek bras yw yndella poran: kompanis. Ymons i ena rag gwaynya budh. Y’n gwettha prys, ny wra marghas rewlys gans yethow bras ri kentryn dhe gevarghewi yn gwellhe rag an re byghan.
Oll a’n deknegieth kows, chiow konnyk ha systemow ynterweythres lev usys hedhyw yw an askorrasow a hwithrans kenwerthel. Dhe vos sogh, yth yns i po rag dendyl arghans a’th kedhlow, po gwertha gwara ha gonisyow, po delenwel dha dybyansow. Nyns yw tra vyth a’n SK ma rag an les kemmyn. […] Mar nag eus argyans erbysek krev lowr, na wra gwaytya kompanis bras dhe fyski dhe askorra systemow kows Kembrek, Godhalek po Kernewek.8
Prof. Ian McLoughlin, pennskol kint
Ha mars ervirons bos SK Kernewek aventur a yll gwaynya budh, possybyl yw bos agan studh gweth ages dell via gans forsakyans. Dell lever Dr. Fintan Mallor, an fordh vrassa a waynya budh rag kompanis SK arghesys yn privedh yw kedreylya aga thoulys yn devisyow aspians.9 Drefen bos Kernewek onan a’n yethow boghes y’n RU nag yw aspiadow yn es y’n eur ma, hemm yw peryl kowrek rag gweythresieth Kernewek ha’gan strif a-barth omdhetermyans yn stat a vynn galweythegi dissent.
Ha ni ow tochya Kernewek ha’y savla yn-dann gevalav, gwren ni mires orth an kost denel. My a gellis ow soodh 13 bloodh yn yethow dhe SK kettooth ha dell veu eskorrans Sowsnek hewul lowr dhe askusya sevel orth tyli den. Y’n kas diwirhaval may kevyn SK hag a yll askorra Kernewek da, prag y hwrussa nebonan omankombra ow pe kowser? An tybyans a SK ow tenna oll an bewnans a’m taves ertach ha ni ow kwynnel dhe dreusvewa dell on yw skruthus.
Yn sempel, budh yw gorthenebel orth tus. Hedre vo SK an degen nowydh flamm a vudh, y fydh gorthenebel orth tus. Ha yeth yw hy thus.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Nyns eus styr dhe gesunyansow a lytherennow war skrin heb dewis ha heb mynnas.10
Ross Perlin, Endangered Language Alliance
Kyn nag yw yeth dibarow dhe dhensys, onan a’n rannow chif a vos denel yw. Ny yll bos lehes dhe gedhlow hepken, mes yth yw argerdh sosyel dres eghen.11 Ni oll a wor py mar synthesek y sen skoodhyans prener der SK po fatel yll SK fyllel orth konvedhes arliwyow. Dell gampol Dr. Mallory, “Yeth [yw] neppyth moy kepar hag enev a gemeneth. Ny yllir gwitha hemma yn jynn. Ny yllir assoylya kudyn denel kepar ha yethladhans gans gwel a yeth hag a remov an gerann denel.”12
Ny yll SK konvedhes Kernewek po taves vyth aral. Papynjay chonsus yw: y targan py ger yw gwirhaval wosa an huni kyns.13 Ny yll agan konvedhes. Ny yll mynnes tra vyth. Mar kwra derivas orthis y vos pes da dha weres, gow yw. My a vynn agan kemeneth dhe devi, mes ny wra kans jynn-amontya a yll ‘kewsel Kernewek’ keworra orti. Y hwra unn den—ow tri tybyansow ha govenegow hag ownow ha gwanderyow—ha ny dybav y hwra an jynnys-amontya kernwegorek keworra unn den hogen orth agan kemeneth.
Gwettha, mar kwra, yma dustuni a-dhyworth Microsoft hag a brof y hwra an devnydh a SKDin war oberennow yeth, unweyth an seythen hogen, aperya gallos godhvosel a dhyski, ow ledya orth omworrans lehes gans an desten, gorfydhyans y’n deknegieth ha sleyneth sprallys a assoylya kudynnow yn anserghek.14 Der usya toulys SK dhe ‘dhyski’ Kernewek, possybyl yw ni dhe shyndya gallos dyski an yeth vytholl heb an kroch ma. Ni a wra gul mimyoryon yn le Kernewegoryon.
Ynwedh Perlin a boslev an elven dhenel, ow leverel pan wren ni synsi kemeneth avel kres agan yethow, dell wren, an papynjay chonsus a yll bos klewys kepar ha defolyans.15 Y’n eur ma, my a aswon pan eus nebonan owth usya ‘Kernewek’ SK dhymm. An tybyans ny wrav vy unn jydh godhvos pan eus estren—nebonan a wrussen vy dynerghi mar pe lyver po klass ganses—ow popettya diwawen ow hengerens ha kewsel dresta yw bedhrosus. Yma dhe’n dra an wan dhistowgh a drevesigeth, a berghenegyans pan yllir gwerthveurhe, a bapynjaya pan yllir junya agan kesgan.
Unver yw skolheyk Hawai’i henwys Noah Ha‘alilio Solomon: “Ankensi yw, drefen ni dhe vos kofhes a’n prysyow oll re beu agan gonisogeth ha yeth perghenegys. Ni re beu owth omladh dre dhens hag ewines yn batel gales a-barth dasvewheans yeth.[…] Y hwra pobel krysi bos hemma representyans ewn a’n yeth a Hawai’i.”16
TREST HAG OMGLEWANS AN GEMENETH
Hir re beu an kil-lash gorthjynn ow kovryjyon mes lemmyn yma va ow terri an arenep dell hevel.17
NBC NEWS
Tardh an arvedhennow rag SK y honan (clanker, tinskin, toaster), y askorras (slop, dross, brainrot) ha’y usyoryon (slopper, groksucker, botlicker, second-hand thinker)—keffrys hag erel selys moy yn kler war geryow kas gwir dell ov attes gans aga heworra—a re hwedhel a stons ollgemmyn a wogrys ha divlases war-tu hag an deknegieth ha’y devnydh war an kesrosweyth Sowsnek ha yethow bras erel.18 Kynth yw an stons yn-mysk gwesyon dek ha korforeth hwath gwresek, rag an boblek gemmyn y hwra SK “dos ha bos keschanjyadow gans taklow tamm kawgh.”19
Pella, nyns yw marnas yethow moyhariv gans an gwel negedhek ma a SK avel podrek. Sampel uskis a gampollow media sosyel ha meusi ow tochya SK ha Godhalek Alban gans Professor Lamb a dhiskwedhas fals a 54% negedhek, 33% posedhek ha 13% heptu. (Lamb, 2024) Sentiment an kampol negedhek an moyha talvesys o bos SK dregynnus hag an nessa y talvia dhyn lettya SK rag kestav gans tavosow ertach.
Pyth eson ni ow leverel orth agan diyskynysi? Ny dal agan yeth ha gonisogeth an strivyans personel? Hemm yw martesen fatel wrussen vy y redya, a pen vy i.20
Gorthebydh sondyans Kernewek
Kernewek a baynt aven moy serth, yn arbennik gans dyskoryon ha kowsoryon yowynka ha moy skentel gans tek. Sondyans war Discord ha Whatsapp Kernewek a drovyas bos 65% a grysis y fia SK drog (11.5%) po pur dhrog (53%) rag an yeth. Pan veu govynnys pyth a dal bos gorthyp an gemeneth orth SK, 46% a leveris y kodh y hedhi ha 27% y woheles, gans an dus moy ha 60 bloodh hepken ow tybi y kodh oberi ganso.20
31% a worthebydhyon an sondyans a leveris y hwrussa an devnydh a SK yn Kernewek aga fellhe a’n yeth, hag ynwedh 54% a leveris y fiens i pellhes yn krev ha 23% pellhes tamm a by kowethas, asnodh po dyskador pynag ow tevnydhya SK.
An gorthyp a’n re a leveris bos aga godhvos a SK po “konnyk” po “da” o dampnus yn arbennik. Pubonan y’n bagas ma a worthebis y fia SK dregynnus rag Kernewek, y hwrussa an devnydh a SK gans pennfenten aga fellhe a’n bennfenten na yn krev hag y kodh dhyn hedhi an devnydh a SK rag Kernewek.
HONANIETH, LELDER HA DIVERSETH
Nyns esa Aristotelis Ioannis Paschalidis, ow skrifa a-barth UNESCO, ow kewsel yn komparek a-dro dhe yethow lyharivhes pan wrug ev y wovyn, mes an govyn a dhassen yn kreffa ragon: “Pygemmys koll a honanieth a vynnir sakrifia rag effeythuster?”21
Honanieth yw a’n moyha bri rag Kernewegoryon. Studhyans Ute Wimmer Reversing Language Shift: the Case of Cornish a henow “gweythres [an yeth] avel arwodh a honanieth kenedhlek” avel an nessa ughella skila (66%**) yn-mysk kowsoryon ha dyskoryon, fethys gans gonisogeth Kernow (80%) hepken.22 Yth havalsa hemma bos acheson solempnyans, mes pan vo SK keworrys, y teu ha bos peryl. Vincent Koc a Hyperlink a lever y hyll SK “kevri dre wall orth an gwannheans a yeth ha honanieth wonisogethel”.23
Ev a aswon ynwedh y hallsa awtomategi dyski po dinythi yeth “lehe an rychedh ha lelder hag a dheu a gowsoryon dhenel neb a dheg istoriow gonisogethel y’ga hows”. Yn hwir, peswar studhyans gwrys gans Pennskol Kaliforni Soth re dhiskwedhas bos devnydhya PYB dhe weres gans skrifa “kelmys orth dyfygyansow nosedhek yn diverseth yethel hag y hyll mellya gans an konvedhes brysoniethel ha kowethasel yw proviys gans yeth.”24
Ha hemm yw yn Sowsnek, onan a’n yethow an ryccha ha brassa y’n bys. Dismyk an effeyth war yeth byghanna kepar ha Kernewek—gans le a dhogvennans, le a gedhlow, sel kowsoryon munys hag ogas hag arghans mann—ha war y lies orgraf hag eghen yeth. Yn arbennik yn gettesten Kernewek, seulabrys yma kowsoryon Diwedhes ow strivya dhe vos gwelys avel vas gans gwartheyvans Kres. A dybyn y hwor SK an dyffrans? Heb preder, y hwra po kemyska puptra warbarth, ow sowdheni pubonan, po devnydhya Kres dhe fetha Diwedhes.
An gwruthyl a dhalgh herdhys gans SK Dinythus, dre favera yethow savonegys, a argyl an vansyans a rannyethow ranndiryel.25
Kresen woramontyorieth Barcelona
Nyns yw eghennow hepken yn peryl; SK a wodros beudhi Kernewek yn tien. Akordys yw Perlin y hallsa an platheans yethel a hwarva dres kansbledhynnyow yn Sowsnek hwarvos dres nos yn yeth lyharivhes gans SK orth an fronnow—dell via, ow pos gallosek a bassya Kernewek denel heb assay. Ev a venek prederow yn kever PYB ow rewi yeth yn hy le ha hogen ow settya pyth yw an styr a wodhvos an yeth, yn arbennik gans niverow munys a gowsoryon deythyek.26
Treylyansow leun a atal a liesha warlinen kepar ha nowodhow fug. Kowsoryon deythyek a’n yethow ma yw passyes avel bos “re gales dhe drovya”, komparys orth fordhow awtomategys a surheans kwalita hag yw disjunys yn tien a geskomunyans y’n bys gwir. Kynth yw possybyl rag kemenethow yeth brassa ha moy gallosek synsi an bottys ma dhe akont ha’ga devnydhya yn stratejek hogen, re es yw dismygi [yeth lyhariv] ow pos reverthys.26
Ross Perlin, Endangered Language Alliance
Heb kontrol hag yn diwla an gewri deknegieth, Kernewek synthesek a wra gornivera ha gorthrabellhe Kernewek denel.
SOVRANEDH KEDHLOW HA KOLONEGIETH
Sovranedh kedhlow teythyek yw an gwir gans [kenedhel teythyek] a woverna an kuntel, perghenogeth ha gweytha a’y hedhlow hy honan.27
Native Nations Institute
Byttegyns, yma gonisogethow teythyek hag usi owth oberi war geskowethyans moy ewnhynsek gans SK. Tek heb an kowr a res asnodhow, mes y as kemenethow gwitha sovranedh kedhlow war an gerthen wonisogethel hag yw aga yeth. Te Mana Raraunga, Rosweyth Sovranedh Kedhlow Māori, re wrug rol a bennrewlys rag an gwruthyl, devnydhya ha kevrenna a gedhlow Māori, ow ragwirhe an edhom a grefhe maystri rag Māori a-lemmyn hag a dheu.
I a venek poynt posek hag a dalvia bos konsidrys gans rach gans stywards a skians yethel ha gonisogethel: “Kedhlow ahanan, a-dro dhyn ha’gan asnodhow, yw kerthennow a bris. Pan vo maystri kellys, kales yw y dhaskemeres.”28 Ny dal gul erviransow yn skav po yn uskis; y hyllyn pupprys leverel “ea” mar kwrussyn leverel “na” kyns orth us sett kedhlow, mes ny yllyn nevra leverel “na” mar kwrussyn leverel “ea” seulabrys.
An desten SK, kepar ha pub le aral, yw leun a dus hag yw settys y’ga maneryow ha gans gwel pur drevesigel yn tidowl.29
Michael Running Wolf, First Languages AI Reality
Hemm yw pur bosek y’n gettesten a’n kontrol possybyl a gedhlow Kernewek gans korforethow gallosek a-ves. Estenegieth jatelydhek re beu molleth war gowethasow y’n amal emperourethek ha nyns yw kowethas Kernewek dyffrans, wosa enebi kansvledhynnyow a’y rychys hag asnodhow naturek ow pos destryppys ha gwerthys dre vras gans budh tus yn-mes a Gernow.
An lyver henwys Indigenous Data Sovereignty and Policy a verk lemmyn y hyllir gweles perthynyansow kedhlow avel “pesyans a’n argerdhow ha systemow-krysi isworwedhek a estennans, drogusyans, kuntellyans ha diberghenogeth re beu gwrys war boblansow Teythyek dres trevesigeth istorek.”30 An konvedhes estennek ma a gedhlow yw, dell verkons, hevelebys a-der goderrys gans tyli pobel rag aga hedhlow.
Wostiwedh, res yw ma na vo agan yeth gorrys yn diwla a-ves routys gans pennrewlys perghenogel na as dhyn sovranedh lowr a onan a’gan asnodhow naturel an moyha posek: agan yeth. Res yw dhyn kavos pennrewlys kedhlow ygor, a-der plegya orth kontrol korforethel. Res yw dhyn lewya ha gidya an devnydh a’gan kedhlow, a-der y usya erbynn agan lesow ha rag pocketys tek bras.
A-der drehedhes an arwithans a davosow avel kudyn teknegiethel, my a dyb bos res dhe gemenethow teythyek bos reythhes yn politek, po der arghasans a wovernansow po dre dhifresyansow laghel dhe dhevnydhya aga yethow.31
Dr. Fintan Mallory, Pennskol Durham
Res yw dhyn ragwirhe yeth-avel-kemeneth ha hwilas devnydh ygor, ewnhynsek hag ethegel a’gan kerthennow yeth, ertach ha gonisogethel. Res yw dhyn goheles tybi a SK avel an hus mayth ambos ha kevarghewi yn hwithrans selyek, lewys gans agan kemeneth. Ny wra korforethow agan selwel ha, hogen, i a yll agan shyndya.
NYNS EUS KERNEWEK WAR BLANET MAROW
Governansow ha kevalav ollvysel […] yw omres dhe ideologieth marghas ‘rydh’ moy es dell yns omres dhe sewena kemenethow, pobel ha’n planet.32
Cymdeithas yr Iaith Maniffesto 2022
An brassa rann a vreusyansow a SK a wrussa meneges hemma seulabrys. Onan a’n argyansow brassa yw erbynn SK Dinythus, mes rag own bos ankoth dhis, ni a wra mires orth an chif boyntys.
Res yw bos evadow an dowr goyeynhe pub kresen kedhlow SK. Y kollenk an dowr ma. Pennskol Kaliforni re dherivas “y hallsa demond dowr ollvysel SK hedhes 4.2-6.6 bilvil metrow kubek erbynn 2027. Henn yw moy es 50 kansran a us dowr bledhynnyek an RU yn 2023.”33 Y kettermyn, an Desedhek Ollvysel Erbysieth Dowr a dheklaryas “barras dowr ow tardha yn uskis” ma na dal Kernewek kevri dhodho.34
Ni re dheuth ha bos yn hwir omres dhe deknegiethow privedh gwrys ha kontrolys gans dornas a gompanis diskler [hag] a hevel bos mygyl dre vras orth an sewyansow sosyel a’ga gwriansow ha kevri yn ispoyntel marnas mars yns i konstrinys gans rewlys an wovernans dhe wellhe aga imach poblek.35
Iker Erdocia, Pennskol Sita Dulyn
Yma edhom dhe SK a vynsow kowrek a galesweyth orth kost palas monyow tanow. Kales yw estenna ha purhe an re ma hag yma kostow kerghynedhel ha sosyel poos. Estennys yns i yn fenowgh a hwelyow yn powyow gans difresyansow lakka rag lavur ha’n kerghynnedh. Reset a lever “yn fenowgh y hwra kemenethow yw trigys yn ogas dhe’n hwelyow, yn fenowgh bagasow lyhariv po teythyek, enebi gwethheans an tir, defolyans an dowr hag abusyans gwiryow denel. Meur a hemma a yll bos kelmys yn tidro orth an galesweyth SK.”36 Pan yw an galesweyth kegys yn sertan hag euver, ena tewlys yw avel e-wast yn kemenethow boghosek. Res yw nyns yw an avonsyans possybyl a Gernewek orth kost agan kemenethow hwor lyhariv ha teythyek.
Ynwedh res yw myns hujes a nerth rag trenya SK, yn skon martesen an keth myns ha pow byghan37 hag yma ol troos karbon kowrek.38 Kler yw—der usadow dowr, diwysyans estennek, konsumyans nerth hag ol troos karbon—bos SK yeyn nowodhow rag kerghynnedh ow strivya a’n planet mayth on ni trigys warnodho ha nyns eus Kernewek war blanet marow.
GUL DHE SK BOS EKS-PAPYNJAY
A-der gul dhe yethow lyhariv bos moy hedhadow, lemmyn yma SK ow kwruthyl tardhek pupprys owth omlesa rag studhyoryon ha kowsoryon a’n yethow ma dhe wolya.39
mit technology review
Ni re glewas a’n anwirhevelepter efan a wul dhe SK gallos mimya Kernewek yn golow an kostys yn kedhlow, ober, termyn ha teknegieth. Ni re gonsidras lycklod an dewisynter dispresyans yeyn po askorras-aspia ha’n posekter a sovranedh kedhlow. Ni re redyas a-dro dhe’n effeythyow war vewnansow Kernewegoryon, keffrys ha’n kostys katastrofek rag an kerghynnedh ha poblow teythyek.
Ni re dhyskas y hwra platheans yethel gans SK boghosekhe y destennow ha fatel yll SK martesen ervira a’gan parth fatel godh dh’agan yeth oberi. Ni re welas an anwoheladewder a yeth avel denel ha’n peryllyow a wul ‘dyskoryon’ na yll dyski ha ‘kowsoryon’ na yll kewsel. Ni re welas an peryllyow orth bri ha fydhyans rag kowethasow a wrussa palas an pyth hag yw gwelys avel ‘skomblans’.
Ni re glewas prag y fia omblegya orth an jagganat a SK error rag Kernewek ha dell na vynn agan kemeneth skoodhya gorra an skoos a-dhyworthyn. Yn y le, res yw dhyn batalyas. Res yw dhyn gul dhe Gernewek bos spas mar rydh a skomblans dell yll bos, res yw adhyski orth botlapyoryon yn dyski ha devnydh yeth yn ethegel hag yn effeythus, res yw goheles tybi wortaswerth.
Res yw dhyn gul dh’agan yeth bos parth heb SK, rosweyth a dus fydhyadow ha’ga gwriansow denel, drehevys war lelder, kemeneth, assay ha trest: Kernewek hag yw a-barth an bobel, a’n bobel ha gans an bobel.
Niwlen Ster
Notennow
* A prime example is the laughably-unaffordable restaurant RenMor, which The Headland Hotel thinks is a version of “Re’n Mor”, which they believe means “by the sea” as in “next to the sea” but actually means “by the sea!” like saying “by Zeus!”. This is both hilarious and enraging.
** A figure perhaps lower than it should be if you consider that many of the “emotional motives” which were not counted in this category, such as “I’m Cornish, what better reason do you need?”, do also refer to identity.
FENTENNOW
1. Judah, J. (2025) How AI and Wikipedia have sent vulnerable languages into a doom spiral, MIT Technology Review.
2. Ackermann, A. (2023) When AI doesn’t speak your language, Coda.
3. Crichton, D. (2024) AI and the Death of Human Languages, Lux.
4. Lamb, W. (2024). Could Artificial Intelligence save Scottish Gaelic?, The University of Edinburgh.
5. Dencik, L. (/2025) AI Inequalities: Minority Languages, TUC Cymru.
6. Joshi, P., Santy, S., Budhiraja, A., Bali, K., & Microsoft Research, India. (2020). The State and Fate of Linguistic Diversity and Inclusion in the NLP World. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
7. Ackermann, A. (op cit)
8. McLoughlin, I. (2018) How to teach AI to speak Welsh (and other minority languages), The Conversation.
9. Mallory, F. (2025) RISE UP Panel Discussion & Q&A: What AI Can and Cannot Do for Minoritised Languages, YouTube.
10. Perlin, R. (2024) AI Won’t Protect Endangered Languages, The Dial.
11. RISE UP (2025) #4 RISE UP Event Summary: What AI Can and Cannot Do For Minoritised Languages, RISE UP.
12. Mallory, F. (2024) European Day of Languages: Will lesser spoken languages soon only be kept alive by AI technology? Durham University.
13. Bender, E., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
14. Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025) The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Microsoft.
15. Perlin, R. (op cit)
16. Judah, J. (op cit)
17. Abbruzzese, J., & Wile, R. (2025) Is an AI backlash brewing? What ‘clanker’ says about growing frustrations with emerging tech, NBC News.
18. Webster, K. (2025) Why Using ChatGPT at Work Could Hurt Your Reputation, Inc. Magazine.
19. Herrman, J. (2024) Is That AI? Or Does It Just Suck?, Intelligencer.
20. Wilson, L. (2025) Skians Kreftus ha Kernewek/Artificial Intelligence and Cornish
21. Paschalidis, A. I. (2025) AI and the great linguistic flattening, UNESCO.
22. Wimmer, U. (2010). Reversing Language Shift: the Case of Cornish. Cornish Language Board, p. 113
23. Koc, V. (2025) Generative AI and Large Language Models in Language Preservation: Opportunities and Challenges, ResearchGate.
24. Sourati, Z., Karimi-Malekabadi, F., & Ozcan, M. (2025) The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models, ResearchGate.
25. Melero, M. (2024) The Future of Language (and Cultural) Diversity in the Age of AI, CLARIN.
26. Perlin, R. (op cit)
27. Russo Carroll, S., Rodriguez Lonebear, D., & Martinez, A. (2017). Data Governance for Native Nation Rebuilding, Native Nations Institute.
28. Te Mana Raraunga. (2018). Frequently Asked Questions, Te Mana Raraunga.
29. Ackermann, A. (op cit)
30. Walter, M., Kukutai, T., Carroll, S. R., & Rodriguez-Lonebear, D. (Eds.). (2020). Indigenous Data Sovereignty and Policy. Taylor & Francis, p. 24
31. Mallory, F. (op cit)
32. Cymdeithas yr Iaith (2022) Cymru Rydd, Cymru Werdd, Cymru Gymraeg., p. 27
33. O’Sullivan, L. (2025). How AI’s Failure on Linguistic Diversity is Deepening Global Inequality, RESET – Digital for Good.
34. Harvey, F. (2024). Global water crisis leaves half of world food production at risk in next 25 years, The Guardian.
35. Erdocia, I., Migge, B., & Schneider, B. (2024). Language is not a data set—Why overcoming ideologies of dataism is more important than ever in the age of AI. Journal of Sociolinguistics, 28(5), p. 23
36. O’Sullivan, L. (op cit)
37. Erdenesanaa, D. (2023) A.I. Could Soon Need as Much Electricity as an Entire Country, The New York Times
38. Heikkilä, M. (2022) We’re getting a better idea of AI’s true carbon footprint, MIT Technology Review.
39. Judah, J. (op cit)#4 #AI #ArtificialIntelligence #Breus #Cornish #Cornwall #data #generativeAI #history #jynn #kedhlow #Kernewek #Kernow #Kernowek #LLM #machine #PYB #SK #SKDinythus #SkiansKreftus #Sordya
-
Turning Saffron into Slop – Treylya Safran yn Skomblans
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
Note: Artificial Intelligence (AI) has come to be synonymous with Generative AI (GenAI) and with Large Language Models (LLMs), such as ChatGPT, in common parlance. Unless explicitly stated, I use the terms interchangeably.
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow*, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
LOW-RESOURCES AND LINGUISTIC TYPOLOGY
Simply adding a language to an AI model leads to a spike in poor-quality articles, drowning out quality writing by humans. AI has “industrialized the acts of destruction—which affect vulnerable languages most, since AI translations are typically far less reliable for them.”1 Wikipedia editors from varied languages evidence that machine translation tools have made it easier than ever before to create shoddy articles in minoritised languages, causing massive damage in minutes. AI leads to non-speakers producing much longer, truthier rubbish, Sámi computational linguistics expert Trond Trosterud notes: “the problem [is] that they are armed with Google Translate. Earlier they were armed only with dictionaries.”1
Kernewek, like all but 60 of the world’s roughly 7,000 languages, is designated “low-resource”, meaning it lacks sufficient data to train a machine.2 It is tempting, therefore, to assume that the solution is to provide more data. However, training an LLM requires petabytes of text, audio and video—manually categorised and in a machine-readable format—a vast trove that Kernewek simply does not have.3 Professor Will Lamb, Chair of Gaelic Ethnology and Linguistics at Edinburgh University, speaks of “millions of work hours devoted to just one aspect” of a working AI.4
Even if ChatGPT is trained on another language than English, the time and labour required may make it largely unviable. Current assessments of the performance of ChatGPT for different languages have shown that it performs worse in all tasks.5
Prof. Lina Dencik, Data Justice Lab
Furthermore, the amount of data and work is not the only barrier; at issue is the nature of the language itself. Microsoft has found that languages such as Breton—and thus Kernewek—cause a high rate of errors distinct from the size of their dataset, due to grammatical features, such as mutation, not present in well-sourced languages. As such, they remain poor without significant additional work.6 Essentially, simply adding more Kernewek may not help. Thus, engaging with AI is, for Kernewek, to tie ourselves to slop.
Noten: Skians Kreftus (SK) re dheuth ha bos kesstyr gans SK Dinythus (SKDin) ha gans Patronyow Yeth Bras (PYB), kepar ha ChatGPT, yn lavar kemmyn. Marnas bos menegys yn kler, my a us an termys yn keschanjyadow.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow*, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
ASNODHOW ISL HA TIPOLOGIETH YETHEL
Keworra yeth yn sempel orth patron SK a led orth spik yn erthyglow drog aga kwalita, ow peudhi skrif a gwalita gans tus. SK re wrug “diwysyansegi an aktys diswrians—hag a nas yethow goliadow an moyha, drefen bos treylyansow SK lieskweyth le lel yn tipek ragdha.”1 Golegydhyon Wikipedia a yethow divers a re dustuni re wrug medhelweyth-treylya y wul bos esya dell veu bythkweth kyns gwruthyl erthyglow pilyek yn yethow lyharivhes, ow kawsya damach kowrek yn mynysennow. SK a led orth digowsoryon owth askorra atal lieskweyth hirra ha gwirekka, konnyk yethonieth reknansek Sámi Trond Trosterud a not: “an kudyn [yw] aga bos ervys gans Google Translate. A-varra nyns ens ervys marnas gans gerlyvrow.”1
Kernewek, kepar hag oll marnas 60 a ogas lowr 7,000 yeth a’n bys, yw klassys avel “isel y asnodhow”, ow styrya nag eus dhodho kedhlow lowr dhe drenya jynn.2 Rakhenna, dynyek yw desevos bos an assoylyans profya moy a gedhlow. Byttegyns, res yw petavaytys a dekst, son ha gwydhyow—klassys dre leuv hag yn furvas redyadow gans jynn— dhe drenya PYB, tresorva efan nag eus dhe Gernewek yn sempel.3 Y kews Professor Will Lamb, Kaderyer Ethnologieth ha Yethonieth Wodhalek orth Pennskol Karedin, a “vilvilyow a ourys ober sakrys orth unn wedh hepken” a SK owth oberi.4
Hogen mars yw ChatGPT trenys war yeth a-der Sowsnek, an termyn hag ober yw res a styr y vos martesen anhewul dre vras. Arvreusyansow a-lemmyn a berformyans a ChatGPT rag yethow dyffrans re dhiskwedhas y perform gweth yn oberennow oll.5
Prof. Lina Dencik, Data Justice Lab
Pella, nyns yw an myns a gedhlow hag ober an unsel lett; a vern yw natur an yeth y honan. Microsoft re drovyas y kaws yethow kepar ha Bretonek—hag ytho Kernewek—kevradh ughel a wallow diblans a vraster aga sett kedhlow, drefen nasyow gramasek, kepar ha treylyansow, nag usi kevys yn yethow ughel aga asnodhow. Yndella, i a bes orth bos drog heb meur a ober keworransel.6 Yn essensek, possybyl yw ny wra keworra moy a Gernewek yn sempel gweres. Yndelma, oberi gans SK yw, rag Kernewek, omgelmi orth skomblans.
CORNISH UNDER CAPITALISM
But surely we can improve things over time? It will take a lot of help from AI companies, but it will be worth it. Sadly, Gabriel Nicholas, a research fellow at the Center for Democracy and Technology, has found that once a tech company has established basic capabilities for a language, they pat themselves on the back and move on.7
Big tech companies are just that: companies. They exist to make a profit. Unfortunately, a market dominated by big languages gives them no incentive to invest in improvements for small ones.
All of the speech technology, smart homes and voice interaction systems used today are the products of commercial research. To put it bluntly, they exist to either make money from your data, to sell you more goods and services, or to influence your thinking. None of this AI exists for the public good. […] Unless there is a strong enough economic argument, don’t expect big companies to rush into producing Welsh, Gaelic or Cornish speech systems.8
Prof. Ian McLoughlin, University of Kent
Should they decide that a Kernewek AI is a viable profit-making enterprise, our situation may even be worse than abandonment. As Dr. Fintan Mallory remarks, the dominant means of profit for privately-funded AI enterprises is to convert their tools into surveillance devices.9 As Kernewek is currently one of the UK’s only languages which is not currently easily surveillable, this poses a huge risk to Kernewek activism and the fight for self-determination in a state that seeks to criminalise dissent.
While we’re on the subject of Kernewek and its position under capitalism, let’s consider the human cost. I lost my 13-year career in language to AI as soon as English output became viable enough to excuse not paying a human. In the unlikely instance that we achieve an AI that can produce quality Kernewek, why would anyone bother paying speakers? The idea of AI sucking all the life out of my heritage language when we are struggling to survive as-is is appalling.
Simply put, profit is antithetical to people. While AI is the new favourite toy of profit, it will be antithetical to people. And a language is its people.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Combinations of characters on a screen mean nothing without agency and intention.10
Ross Perlin, Endangered Language Alliance
While language is not unique to humans, it is one of the chief parts of being human. It cannot be reduced to mere data, but is a highly social process.11 We all know how synthetic customer support via robot sounds or how AI fails to pick up nuance. As Dr. Mallory comments, “Language [is] something more like the soul of a community. You can’t store this in a machine. You can’t solve a human problem like linguicide with a view of language that removes the human component.”12
AI cannot comprehend Kernewek or any other language. It is a stochastic parrot: predicting what word is likely to follow the previous one.13 It cannot understand us. It cannot intend anything. If it tells you it feels delighted to help you, it is lying. I want our community to grow, but one hundred ‘Cornish-speaking’ computers do not add to it. One human does—bringing ideas and hopes and fears and foibles—and I do not think the Kernewek ‘speaking’ computers will add even one human to our community.
Worse, if it does, there is evidence from Microsoft to suggest that the use of GenAI on language tasks, even once a week, impairs cognitive ability to learn, leading to decreased engagement with the topic, overreliance on the technology and hobbled skills in independent problem-solving.14 By using AI tools to ‘teach’ a learner Kernewek, we may in fact be impairing their ability to learn the language at all without this crutch. We will make regurgitators in place of speakers.
Perlin also emphasises the human element, saying that when we hold community central to our languages, as we do, the stochastic parrot can feel like a violation.15 At the moment, I can tell when someone is using AI ‘Kernewek’ to me. The idea that one day I will not know when an outsider—someone I would welcome if they took up a book or a class—is puppeting my ancestors’ jaws and speaking through them is ghoulish. It has the instant sting of colonialism, of appropriation when one could appreciate, of parroting when one could join our chorus.
Hawai’ian scholar Ha‘alilio Solomon agrees: “It is painful, because it reminds us of all the times that our culture and language has been appropriated. We have been fighting tooth and nail in an uphill climb for language revitalization.[…] People are going to think that this is an accurate representation of the Hawaiian language.”16
TRUST AND COMMUNITY FEELING
The anti-machine backlash has long been simmering but is now seemingly breaking to the surface.17
NBC NEWS
The explosion of insults for AI itself (clanker, tinskin, toaster), its output (slop, dross, brainrot) and its users (slopper, groksucker, botlicker, second-hand thinker)—as well as others more clearly based on real-world slurs than I am comfortable to include—tells a tale of the general attitude of distrust and disgust towards the technology and its use on anglophone and other majority language internet.18 While the attitude among tech bros and corporates remains bombastic, for the general public AI is “becoming interchangeable with things that sort of suck.”19
Further, it’s not just majority languages with this negative view of AI as taint. A quick sampling of social media comments and likes regarding AI and Scottish Gaelic by Professor Lamb showed a split of 54% negative, 33% positive and 13% neutral. (Lamb, 2024) The sentiment of the top-rated negative comment was that AI is harmful and the second-highest that AI should be kept away from heritage languages.
What are we telling our descendants? That our language and culture isn’t worth the personal effort? That’s how I might read it, if I were them.20
Kernewek survey respondent
Kernewek paints an even starker picture, especially among younger and more technologically-savvy learners and speakers. A survey on Cornish Discord and Whatsapp found that 65% felt AI would be bad (11.5%) or very bad (53%) for the language. When asked what the community response should be to AI, 46% said we should prevent it and 27% avoid it, with only over-60s thinking that we should work with it.20
31% of respondents said using AI in Kernewek would cause them to feel estranged from the language, while 54% said that they would feel strongly estranged and 23% a little estranged from any organisation, resource or teacher using AI.
The response from those who gave their knowledge of AI as either “expert” or “good” was particularly damning. Everyone in this group responded that AI would be harmful for the language, that the use of AI would estrange them from a source strongly and that we should prevent the use of AI for Kernewek.
IDENTITY, AUTHENTICITY AND DIVERSITY
Aristotelis Ioannis Paschalidis, writing for UNESCO, was not speaking specifically about minoritised languages when he asked this, but the question resonates even more strongly for us: “How much loss of identity is one willing to sacrifice for efficiency?”21
Identity is of paramount importance to Kernewek speakers. Ute Wimmer’s study Reversing Language Shift: the Case of Cornish identified the language’s “function as a symbol of national identity” as the second highest motive (66%**) among speakers and learners, beaten only by Cornish culture (80%).22 This would seem cause for celebration, but when AI is added to the mix, it becomes a risk. Vincent Koc of Hyperlink states that AI can “inadvertently contribute to the dilution of language and cultural identity.”23
He also identifies that automating language learning or generation “may diminish the richness and authenticity that comes from human speakers who carry cultural histories in their speech.” Indeed, four studies by the University of Southern California have shown that using LLMs to assist writing “is linked to notable declines in linguistic diversity and may interfere with the societal and psychological insights language provides.”24
This is in English, one of the richest and largest languages in the world. Imagine the possible impact on a smaller language like Kernewek—with less documentation, less data, a tiny speakerbase and basically no money—and on its many language varieties and orthographies. Particular to the Kernewek context, Late speakers are already struggling to be seen as valid under the dominance of Middle. Do we think AI knows the difference? Thoughtlessly, it will either mix everything together, confusing everyone, or it will use Middle to overwhelm Late.
Generative AI-driven content creation, by favoring standardized languages, risks the disappearance of regional dialects.25
Barcelona supercomputing Center ….
Not only are varieties at risk; AI threatens to drown Kernewek as a whole. Perlin agrees that the linguistic flattening that occurred over centuries in English could manifest overnight in a minoritised language with AI at the helm—as it would be, being able to effortlessly outstrip human Kernewek. He raises concerns of LLMs freezing a language in place and even defining what it means to know the language, especially with low numbers of native speakers.26
Garbage translations multiply online like fake news. Native speakers of the languages in question are bypassed as being “too hard to find,” compared with automated methods of vetting that are completely disconnected from real-life communication. While larger and more powerful language communities may be able to hold the bots to account and even make strategic use of them, it is all too easy to imagine [a minority language] being overwhelmed.26
Ross Perlin, Endangered Language Alliance
Uncontrolled and in the hands of tech giants, synthetic Kernewek will outnumber and outmanoeuvre human Kernewek.
DATA SOVEREIGNTY AND COLONIALISM
Indigenous data sovereignty is the right of [an indigenous nation] to govern the collection, ownership, and application of its own data.27
Native Nations Institute
There are, however, indigenous cultures that are working on a more equitable relationship with AI. Tech without the giant requires resources, but it allows communities to retain data sovereignty over the cultural asset that is their language. Te Mana Raraunga, the Māori Data Sovereignty Network, has created a list of principles for the creation, use and sharing of Māori data, prioritising the need to enhance control for current and future Māori.
They raise a key point that should be considered carefully by stewards of linguistic and cultural knowledge: “Data from us, and about us and our resources, are valuable assets. Once control of it is lost, it is difficult to regain.”28 Decisions must not be taken lightly or hastily; we can always say “yes” if we have previously said “no” to a particular dataset’s use, but can never say “no” if we have already said “yes”.
The AI field, like any other space, is occupied by people who are set in their ways and unintentionally have a very colonial perspective.29
Michael Running Wolf, First Languages AI Reality
This is vital in the context of the potential control of Kernewek data by powerful external corporations. Capitalist extractivism has long been a bane on societies in the imperial periphery and our Cornish society is no different, having faced centuries of its wealth and natural resources being stripped and sold by and large for the profit of those outside Cornwall.
The book Indigenous Data Sovereignty and Policy notes that current data relations can be seen as “a continuation of the processes and underlying belief systems of extraction, exploitation, accumulation and dispossession that have been visited on Indigenous populations through historical colonialism.”30 This extractive understanding of information is, they note, not disrupted but rather replicated by paying people for their data.
Ultimately, our language must not lie in outside hands governed by proprietary principles that do not allow us sufficient sovereignty over one of our most valuable natural resources: our language. We must have open data principles, not bow to corporate control. We must steer and steward the use of our data, rather than expose it to use against our interests and for the pockets of big tech.
Rather than approaching language preservation as a technical problem, I think indigenous communities need to be politically empowered, whether that be funding from governments or legal protections to use their languages.31
Dr. Fintan Mallory, Durham University
We must prioritise language-as-community and seek open, equitable and ethical use of our language, heritage and other cultural assets. We must avoid thinking of AI as the magic that it promises and invest in basic research, driven by our own community. Corporations will not save us and, indeed, may do us great harm.
NO CORNISH ON A DEAD PLANET
Global capitalism and governments […] are addicted to ‘free’ market ideology over the wellbeing of communities, people and the planet.32
Cymdeithas yr Iaith Maniffesto 2022
Honestly, most takedowns of AI would have hit this point already. It’s one of the main arguments against Generative AI, but in case you’re not familiar with it, we will briefly look over the main points.
Water used in cooling AI data centers must be drinkable water. AI guzzles this water. The University of California has reported that “global water demand from AI could reach 4.2-6.6 billion cubic meters by 2027. That exceeds 50 percent of the UK’s annual water use in 2023.”33 All this while the Global Commission on the Economics of Water has declared “a rapidly accelerating water crisis” to which Kernewek should not be contributing.34
We have become utterly dependent on private technologies manufactured and controlled by a handful of opaque companies [who] appear mostly indifferent to the social consequences of their activities and only invest minimally if obliged by government regulations to enhance their public image.35
Iker Erdocia, Dublin City University
AI requires vast quantities of hardware at the cost of mining rare earth minerals. These are difficult to extract and purify and come with heavy environmental and social costs. They are often extracted from mines in countries with poorer environmental and labour protections. Reset states that “communities living near these mines, often indigenous or minority groups, regularly face land degradation, water contamination and human rights abuses. Much of this can be directly linked to the AI hardware.”36 When the hardware inevitably cooks and is useless, it is then thrown out as e-waste into poor communities. The potential advancement of Kernewek must not come at the expense of our sister indigenous and minority communities.
Training an also AI requires huge amounts of energy, soon perhaps as much as a small country37 and has an enormous carbon footprint.38 What is clear is that—through water usage, extractive industry, energy consumption and carbon footprint—AI is bad news for the struggling environment of the planet we live on and there is no Cornish on a dead planet.
MAKING AI AN EX-PARROT
Rather than making minority languages more accessible, AI is now creating an ever expanding minefield for students and speakers of those languages to navigate.39
mit technology review
We have heard of the vast improbability of getting AI to be able to mimic Kernewek in light of the costs in data, work, time and technology. We have considered the likely choice of cold negligence or surveillance product and the importance of data sovereignty. We have read about the effects on the livelihoods of Cornish speakers, as well as the the catastrophic costs to the environment and indigenous peoples.
We have learned that linguistic flattening by AI impoverishes its subjects and how AI may decide for us how our language must operate. We have seen the inescapability of language as human and the risks of creating ‘learners’ who cannot learn and ‘speakers’ who cannot speak. We have seen the dangers to reputation and trust for any organisation who would shovel what is seen as ‘slop’.
We have heard why giving in to the juggernaut of AI would be a mistake for Kernewek and how our community does not support our laying down of the shield. Instead, we must fight. We must make Kernewek a space as free of slop as possible, we must educate botlickers into ethical and effective language learning and use, we must avoid second-hand thinking.
We must make our language a no AI zone, a network of reliable humans and their human creations, built on authenticity, community, effort and trust: a Kernewek for the people, of the people and by the people.
KERNEWEK YN-DANN GEVALAV
Mes yn sur y hyllyn ni gwellhe taklow dres termyn? Y fydh res meur a weres a gompanis SK, mes y talvia dhyn. Yn trist, Gabriel Nicholas, kesvroder hwithrans orth an Center for Democracy and Technology, re drovyas pan wrug kompani tek fondya gallosow selyek rag unn yeth, i a omgeslowenha yn ughel hag ena movya yn-rag.7
Kompanis tek bras yw yndella poran: kompanis. Ymons i ena rag gwaynya budh. Y’n gwettha prys, ny wra marghas rewlys gans yethow bras ri kentryn dhe gevarghewi yn gwellhe rag an re byghan.
Oll a’n deknegieth kows, chiow konnyk ha systemow ynterweythres lev usys hedhyw yw an askorrasow a hwithrans kenwerthel. Dhe vos sogh, yth yns i po rag dendyl arghans a’th kedhlow, po gwertha gwara ha gonisyow, po delenwel dha dybyansow. Nyns yw tra vyth a’n SK ma rag an les kemmyn. […] Mar nag eus argyans erbysek krev lowr, na wra gwaytya kompanis bras dhe fyski dhe askorra systemow kows Kembrek, Godhalek po Kernewek.8
Prof. Ian McLoughlin, pennskol kint
Ha mars ervirons bos SK Kernewek aventur a yll gwaynya budh, possybyl yw bos agan studh gweth ages dell via gans forsakyans. Dell lever Dr. Fintan Mallor, an fordh vrassa a waynya budh rag kompanis SK arghesys yn privedh yw kedreylya aga thoulys yn devisyow aspians.9 Drefen bos Kernewek onan a’n yethow boghes y’n RU nag yw aspiadow yn es y’n eur ma, hemm yw peryl kowrek rag gweythresieth Kernewek ha’gan strif a-barth omdhetermyans yn stat a vynn galweythegi dissent.
Ha ni ow tochya Kernewek ha’y savla yn-dann gevalav, gwren ni mires orth an kost denel. My a gellis ow soodh 13 bloodh yn yethow dhe SK kettooth ha dell veu eskorrans Sowsnek hewul lowr dhe askusya sevel orth tyli den. Y’n kas diwirhaval may kevyn SK hag a yll askorra Kernewek da, prag y hwrussa nebonan omankombra ow pe kowser? An tybyans a SK ow tenna oll an bewnans a’m taves ertach ha ni ow kwynnel dhe dreusvewa dell on yw skruthus.
Yn sempel, budh yw gorthenebel orth tus. Hedre vo SK an degen nowydh flamm a vudh, y fydh gorthenebel orth tus. Ha yeth yw hy thus.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Nyns eus styr dhe gesunyansow a lytherennow war skrin heb dewis ha heb mynnas.10
Ross Perlin, Endangered Language Alliance
Kyn nag yw yeth dibarow dhe dhensys, onan a’n rannow chif a vos denel yw. Ny yll bos lehes dhe gedhlow hepken, mes yth yw argerdh sosyel dres eghen.11 Ni oll a wor py mar synthesek y sen skoodhyans prener der SK po fatel yll SK fyllel orth konvedhes arliwyow. Dell gampol Dr. Mallory, “Yeth [yw] neppyth moy kepar hag enev a gemeneth. Ny yllir gwitha hemma yn jynn. Ny yllir assoylya kudyn denel kepar ha yethladhans gans gwel a yeth hag a remov an gerann denel.”12
Ny yll SK konvedhes Kernewek po taves vyth aral. Papynjay chonsus yw: y targan py ger yw gwirhaval wosa an huni kyns.13 Ny yll agan konvedhes. Ny yll mynnes tra vyth. Mar kwra derivas orthis y vos pes da dha weres, gow yw. My a vynn agan kemeneth dhe devi, mes ny wra kans jynn-amontya a yll ‘kewsel Kernewek’ keworra orti. Y hwra unn den—ow tri tybyansow ha govenegow hag ownow ha gwanderyow—ha ny dybav y hwra an jynnys-amontya kernwegorek keworra unn den hogen orth agan kemeneth.
Gwettha, mar kwra, yma dustuni a-dhyworth Microsoft hag a brof y hwra an devnydh a SKDin war oberennow yeth, unweyth an seythen hogen, aperya gallos godhvosel a dhyski, ow ledya orth omworrans lehes gans an desten, gorfydhyans y’n deknegieth ha sleyneth sprallys a assoylya kudynnow yn anserghek.14 Der usya toulys SK dhe ‘dhyski’ Kernewek, possybyl yw ni dhe shyndya gallos dyski an yeth vytholl heb an kroch ma. Ni a wra gul mimyoryon yn le Kernewegoryon.
Ynwedh Perlin a boslev an elven dhenel, ow leverel pan wren ni synsi kemeneth avel kres agan yethow, dell wren, an papynjay chonsus a yll bos klewys kepar ha defolyans.15 Y’n eur ma, my a aswon pan eus nebonan owth usya ‘Kernewek’ SK dhymm. An tybyans ny wrav vy unn jydh godhvos pan eus estren—nebonan a wrussen vy dynerghi mar pe lyver po klass ganses—ow popettya diwawen ow hengerens ha kewsel dresta yw bedhrosus. Yma dhe’n dra an wan dhistowgh a drevesigeth, a berghenegyans pan yllir gwerthveurhe, a bapynjaya pan yllir junya agan kesgan.
Unver yw skolheyk Hawai’i henwys Noah Ha‘alilio Solomon: “Ankensi yw, drefen ni dhe vos kofhes a’n prysyow oll re beu agan gonisogeth ha yeth perghenegys. Ni re beu owth omladh dre dhens hag ewines yn batel gales a-barth dasvewheans yeth.[…] Y hwra pobel krysi bos hemma representyans ewn a’n yeth a Hawai’i.”16
TREST HAG OMGLEWANS AN GEMENETH
Hir re beu an kil-lash gorthjynn ow kovryjyon mes lemmyn yma va ow terri an arenep dell hevel.17
NBC NEWS
Tardh an arvedhennow rag SK y honan (clanker, tinskin, toaster), y askorras (slop, dross, brainrot) ha’y usyoryon (slopper, groksucker, botlicker, second-hand thinker)—keffrys hag erel selys moy yn kler war geryow kas gwir dell ov attes gans aga heworra—a re hwedhel a stons ollgemmyn a wogrys ha divlases war-tu hag an deknegieth ha’y devnydh war an kesrosweyth Sowsnek ha yethow bras erel.18 Kynth yw an stons yn-mysk gwesyon dek ha korforeth hwath gwresek, rag an boblek gemmyn y hwra SK “dos ha bos keschanjyadow gans taklow tamm kawgh.”19
Pella, nyns yw marnas yethow moyhariv gans an gwel negedhek ma a SK avel podrek. Sampel uskis a gampollow media sosyel ha meusi ow tochya SK ha Godhalek Alban gans Professor Lamb a dhiskwedhas fals a 54% negedhek, 33% posedhek ha 13% heptu. (Lamb, 2024) Sentiment an kampol negedhek an moyha talvesys o bos SK dregynnus hag an nessa y talvia dhyn lettya SK rag kestav gans tavosow ertach.
Pyth eson ni ow leverel orth agan diyskynysi? Ny dal agan yeth ha gonisogeth an strivyans personel? Hemm yw martesen fatel wrussen vy y redya, a pen vy i.20
Gorthebydh sondyans Kernewek
Kernewek a baynt aven moy serth, yn arbennik gans dyskoryon ha kowsoryon yowynka ha moy skentel gans tek. Sondyans war Discord ha Whatsapp Kernewek a drovyas bos 65% a grysis y fia SK drog (11.5%) po pur dhrog (53%) rag an yeth. Pan veu govynnys pyth a dal bos gorthyp an gemeneth orth SK, 46% a leveris y kodh y hedhi ha 27% y woheles, gans an dus moy ha 60 bloodh hepken ow tybi y kodh oberi ganso.20
31% a worthebydhyon an sondyans a leveris y hwrussa an devnydh a SK yn Kernewek aga fellhe a’n yeth, hag ynwedh 54% a leveris y fiens i pellhes yn krev ha 23% pellhes tamm a by kowethas, asnodh po dyskador pynag ow tevnydhya SK.
An gorthyp a’n re a leveris bos aga godhvos a SK po “konnyk” po “da” o dampnus yn arbennik. Pubonan y’n bagas ma a worthebis y fia SK dregynnus rag Kernewek, y hwrussa an devnydh a SK gans pennfenten aga fellhe a’n bennfenten na yn krev hag y kodh dhyn hedhi an devnydh a SK rag Kernewek.
HONANIETH, LELDER HA DIVERSETH
Nyns esa Aristotelis Ioannis Paschalidis, ow skrifa a-barth UNESCO, ow kewsel yn komparek a-dro dhe yethow lyharivhes pan wrug ev y wovyn, mes an govyn a dhassen yn kreffa ragon: “Pygemmys koll a honanieth a vynnir sakrifia rag effeythuster?”21
Honanieth yw a’n moyha bri rag Kernewegoryon. Studhyans Ute Wimmer Reversing Language Shift: the Case of Cornish a henow “gweythres [an yeth] avel arwodh a honanieth kenedhlek” avel an nessa ughella skila (66%**) yn-mysk kowsoryon ha dyskoryon, fethys gans gonisogeth Kernow (80%) hepken.22 Yth havalsa hemma bos acheson solempnyans, mes pan vo SK keworrys, y teu ha bos peryl. Vincent Koc a Hyperlink a lever y hyll SK “kevri dre wall orth an gwannheans a yeth ha honanieth wonisogethel”.23
Ev a aswon ynwedh y hallsa awtomategi dyski po dinythi yeth “lehe an rychedh ha lelder hag a dheu a gowsoryon dhenel neb a dheg istoriow gonisogethel y’ga hows”. Yn hwir, peswar studhyans gwrys gans Pennskol Kaliforni Soth re dhiskwedhas bos devnydhya PYB dhe weres gans skrifa “kelmys orth dyfygyansow nosedhek yn diverseth yethel hag y hyll mellya gans an konvedhes brysoniethel ha kowethasel yw proviys gans yeth.”24
Ha hemm yw yn Sowsnek, onan a’n yethow an ryccha ha brassa y’n bys. Dismyk an effeyth war yeth byghanna kepar ha Kernewek—gans le a dhogvennans, le a gedhlow, sel kowsoryon munys hag ogas hag arghans mann—ha war y lies orgraf hag eghen yeth. Yn arbennik yn gettesten Kernewek, seulabrys yma kowsoryon Diwedhes ow strivya dhe vos gwelys avel vas gans gwartheyvans Kres. A dybyn y hwor SK an dyffrans? Heb preder, y hwra po kemyska puptra warbarth, ow sowdheni pubonan, po devnydhya Kres dhe fetha Diwedhes.
An gwruthyl a dhalgh herdhys gans SK Dinythus, dre favera yethow savonegys, a argyl an vansyans a rannyethow ranndiryel.25
Kresen woramontyorieth Barcelona
Nyns yw eghennow hepken yn peryl; SK a wodros beudhi Kernewek yn tien. Akordys yw Perlin y hallsa an platheans yethel a hwarva dres kansbledhynnyow yn Sowsnek hwarvos dres nos yn yeth lyharivhes gans SK orth an fronnow—dell via, ow pos gallosek a bassya Kernewek denel heb assay. Ev a venek prederow yn kever PYB ow rewi yeth yn hy le ha hogen ow settya pyth yw an styr a wodhvos an yeth, yn arbennik gans niverow munys a gowsoryon deythyek.26
Treylyansow leun a atal a liesha warlinen kepar ha nowodhow fug. Kowsoryon deythyek a’n yethow ma yw passyes avel bos “re gales dhe drovya”, komparys orth fordhow awtomategys a surheans kwalita hag yw disjunys yn tien a geskomunyans y’n bys gwir. Kynth yw possybyl rag kemenethow yeth brassa ha moy gallosek synsi an bottys ma dhe akont ha’ga devnydhya yn stratejek hogen, re es yw dismygi [yeth lyhariv] ow pos reverthys.26
Ross Perlin, Endangered Language Alliance
Heb kontrol hag yn diwla an gewri deknegieth, Kernewek synthesek a wra gornivera ha gorthrabellhe Kernewek denel.
SOVRANEDH KEDHLOW HA KOLONEGIETH
Sovranedh kedhlow teythyek yw an gwir gans [kenedhel teythyek] a woverna an kuntel, perghenogeth ha gweytha a’y hedhlow hy honan.27
Native Nations Institute
Byttegyns, yma gonisogethow teythyek hag usi owth oberi war geskowethyans moy ewnhynsek gans SK. Tek heb an kowr a res asnodhow, mes y as kemenethow gwitha sovranedh kedhlow war an gerthen wonisogethel hag yw aga yeth. Te Mana Raraunga, Rosweyth Sovranedh Kedhlow Māori, re wrug rol a bennrewlys rag an gwruthyl, devnydhya ha kevrenna a gedhlow Māori, ow ragwirhe an edhom a grefhe maystri rag Māori a-lemmyn hag a dheu.
I a venek poynt posek hag a dalvia bos konsidrys gans rach gans stywards a skians yethel ha gonisogethel: “Kedhlow ahanan, a-dro dhyn ha’gan asnodhow, yw kerthennow a bris. Pan vo maystri kellys, kales yw y dhaskemeres.”28 Ny dal gul erviransow yn skav po yn uskis; y hyllyn pupprys leverel “ea” mar kwrussyn leverel “na” kyns orth us sett kedhlow, mes ny yllyn nevra leverel “na” mar kwrussyn leverel “ea” seulabrys.
An desten SK, kepar ha pub le aral, yw leun a dus hag yw settys y’ga maneryow ha gans gwel pur drevesigel yn tidowl.29
Michael Running Wolf, First Languages AI Reality
Hemm yw pur bosek y’n gettesten a’n kontrol possybyl a gedhlow Kernewek gans korforethow gallosek a-ves. Estenegieth jatelydhek re beu molleth war gowethasow y’n amal emperourethek ha nyns yw kowethas Kernewek dyffrans, wosa enebi kansvledhynnyow a’y rychys hag asnodhow naturek ow pos destryppys ha gwerthys dre vras gans budh tus yn-mes a Gernow.
An lyver henwys Indigenous Data Sovereignty and Policy a verk lemmyn y hyllir gweles perthynyansow kedhlow avel “pesyans a’n argerdhow ha systemow-krysi isworwedhek a estennans, drogusyans, kuntellyans ha diberghenogeth re beu gwrys war boblansow Teythyek dres trevesigeth istorek.”30 An konvedhes estennek ma a gedhlow yw, dell verkons, hevelebys a-der goderrys gans tyli pobel rag aga hedhlow.
Wostiwedh, res yw ma na vo agan yeth gorrys yn diwla a-ves routys gans pennrewlys perghenogel na as dhyn sovranedh lowr a onan a’gan asnodhow naturel an moyha posek: agan yeth. Res yw dhyn kavos pennrewlys kedhlow ygor, a-der plegya orth kontrol korforethel. Res yw dhyn lewya ha gidya an devnydh a’gan kedhlow, a-der y usya erbynn agan lesow ha rag pocketys tek bras.
A-der drehedhes an arwithans a davosow avel kudyn teknegiethel, my a dyb bos res dhe gemenethow teythyek bos reythhes yn politek, po der arghasans a wovernansow po dre dhifresyansow laghel dhe dhevnydhya aga yethow.31
Dr. Fintan Mallory, Pennskol Durham
Res yw dhyn ragwirhe yeth-avel-kemeneth ha hwilas devnydh ygor, ewnhynsek hag ethegel a’gan kerthennow yeth, ertach ha gonisogethel. Res yw dhyn goheles tybi a SK avel an hus mayth ambos ha kevarghewi yn hwithrans selyek, lewys gans agan kemeneth. Ny wra korforethow agan selwel ha, hogen, i a yll agan shyndya.
NYNS EUS KERNEWEK WAR BLANET MAROW
Governansow ha kevalav ollvysel […] yw omres dhe ideologieth marghas ‘rydh’ moy es dell yns omres dhe sewena kemenethow, pobel ha’n planet.32
Cymdeithas yr Iaith Maniffesto 2022
An brassa rann a vreusyansow a SK a wrussa meneges hemma seulabrys. Onan a’n argyansow brassa yw erbynn SK Dinythus, mes rag own bos ankoth dhis, ni a wra mires orth an chif boyntys.
Res yw bos evadow an dowr goyeynhe pub kresen kedhlow SK. Y kollenk an dowr ma. Pennskol Kaliforni re dherivas “y hallsa demond dowr ollvysel SK hedhes 4.2-6.6 bilvil metrow kubek erbynn 2027. Henn yw moy es 50 kansran a us dowr bledhynnyek an RU yn 2023.”33 Y kettermyn, an Desedhek Ollvysel Erbysieth Dowr a dheklaryas “barras dowr ow tardha yn uskis” ma na dal Kernewek kevri dhodho.34
Ni re dheuth ha bos yn hwir omres dhe deknegiethow privedh gwrys ha kontrolys gans dornas a gompanis diskler [hag] a hevel bos mygyl dre vras orth an sewyansow sosyel a’ga gwriansow ha kevri yn ispoyntel marnas mars yns i konstrinys gans rewlys an wovernans dhe wellhe aga imach poblek.35
Iker Erdocia, Pennskol Sita Dulyn
Yma edhom dhe SK a vynsow kowrek a galesweyth orth kost palas monyow tanow. Kales yw estenna ha purhe an re ma hag yma kostow kerghynedhel ha sosyel poos. Estennys yns i yn fenowgh a hwelyow yn powyow gans difresyansow lakka rag lavur ha’n kerghynnedh. Reset a lever “yn fenowgh y hwra kemenethow yw trigys yn ogas dhe’n hwelyow, yn fenowgh bagasow lyhariv po teythyek, enebi gwethheans an tir, defolyans an dowr hag abusyans gwiryow denel. Meur a hemma a yll bos kelmys yn tidro orth an galesweyth SK.”36 Pan yw an galesweyth kegys yn sertan hag euver, ena tewlys yw avel e-wast yn kemenethow boghosek. Res yw nyns yw an avonsyans possybyl a Gernewek orth kost agan kemenethow hwor lyhariv ha teythyek.
Ynwedh res yw myns hujes a nerth rag trenya SK, yn skon martesen an keth myns ha pow byghan37 hag yma ol troos karbon kowrek.38 Kler yw—der usadow dowr, diwysyans estennek, konsumyans nerth hag ol troos karbon—bos SK yeyn nowodhow rag kerghynnedh ow strivya a’n planet mayth on ni trigys warnodho ha nyns eus Kernewek war blanet marow.
GUL DHE SK BOS EKS-PAPYNJAY
A-der gul dhe yethow lyhariv bos moy hedhadow, lemmyn yma SK ow kwruthyl tardhek pupprys owth omlesa rag studhyoryon ha kowsoryon a’n yethow ma dhe wolya.39
mit technology review
Ni re glewas a’n anwirhevelepter efan a wul dhe SK gallos mimya Kernewek yn golow an kostys yn kedhlow, ober, termyn ha teknegieth. Ni re gonsidras lycklod an dewisynter dispresyans yeyn po askorras-aspia ha’n posekter a sovranedh kedhlow. Ni re redyas a-dro dhe’n effeythyow war vewnansow Kernewegoryon, keffrys ha’n kostys katastrofek rag an kerghynnedh ha poblow teythyek.
Ni re dhyskas y hwra platheans yethel gans SK boghosekhe y destennow ha fatel yll SK martesen ervira a’gan parth fatel godh dh’agan yeth oberi. Ni re welas an anwoheladewder a yeth avel denel ha’n peryllyow a wul ‘dyskoryon’ na yll dyski ha ‘kowsoryon’ na yll kewsel. Ni re welas an peryllyow orth bri ha fydhyans rag kowethasow a wrussa palas an pyth hag yw gwelys avel ‘skomblans’.
Ni re glewas prag y fia omblegya orth an jagganat a SK error rag Kernewek ha dell na vynn agan kemeneth skoodhya gorra an skoos a-dhyworthyn. Yn y le, res yw dhyn batalyas. Res yw dhyn gul dhe Gernewek bos spas mar rydh a skomblans dell yll bos, res yw adhyski orth botlapyoryon yn dyski ha devnydh yeth yn ethegel hag yn effeythus, res yw goheles tybi wortaswerth.
Res yw dhyn gul dh’agan yeth bos parth heb SK, rosweyth a dus fydhyadow ha’ga gwriansow denel, drehevys war lelder, kemeneth, assay ha trest: Kernewek hag yw a-barth an bobel, a’n bobel ha gans an bobel.
Niwlen Ster
Notennow
* A prime example is the laughably-unaffordable restaurant RenMor, which The Headland Hotel thinks is a version of “Re’n Mor”, which they believe means “by the sea” as in “next to the sea” but actually means “by the sea!” like saying “by Zeus!”. This is both hilarious and enraging.
** A figure perhaps lower than it should be if you consider that many of the “emotional motives” which were not counted in this category, such as “I’m Cornish, what better reason do you need?”, do also refer to identity.
FENTENNOW
1. Judah, J. (2025) How AI and Wikipedia have sent vulnerable languages into a doom spiral, MIT Technology Review.
2. Ackermann, A. (2023) When AI doesn’t speak your language, Coda.
3. Crichton, D. (2024) AI and the Death of Human Languages, Lux.
4. Lamb, W. (2024). Could Artificial Intelligence save Scottish Gaelic?, The University of Edinburgh.
5. Dencik, L. (/2025) AI Inequalities: Minority Languages, TUC Cymru.
6. Joshi, P., Santy, S., Budhiraja, A., Bali, K., & Microsoft Research, India. (2020). The State and Fate of Linguistic Diversity and Inclusion in the NLP World. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
7. Ackermann, A. (op cit)
8. McLoughlin, I. (2018) How to teach AI to speak Welsh (and other minority languages), The Conversation.
9. Mallory, F. (2025) RISE UP Panel Discussion & Q&A: What AI Can and Cannot Do for Minoritised Languages, YouTube.
10. Perlin, R. (2024) AI Won’t Protect Endangered Languages, The Dial.
11. RISE UP (2025) #4 RISE UP Event Summary: What AI Can and Cannot Do For Minoritised Languages, RISE UP.
12. Mallory, F. (2024) European Day of Languages: Will lesser spoken languages soon only be kept alive by AI technology? Durham University.
13. Bender, E., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
14. Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025) The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Microsoft.
15. Perlin, R. (op cit)
16. Judah, J. (op cit)
17. Abbruzzese, J., & Wile, R. (2025) Is an AI backlash brewing? What ‘clanker’ says about growing frustrations with emerging tech, NBC News.
18. Webster, K. (2025) Why Using ChatGPT at Work Could Hurt Your Reputation, Inc. Magazine.
19. Herrman, J. (2024) Is That AI? Or Does It Just Suck?, Intelligencer.
20. Wilson, L. (2025) Skians Kreftus ha Kernewek/Artificial Intelligence and Cornish
21. Paschalidis, A. I. (2025) AI and the great linguistic flattening, UNESCO.
22. Wimmer, U. (2010). Reversing Language Shift: the Case of Cornish. Cornish Language Board, p. 113
23. Koc, V. (2025) Generative AI and Large Language Models in Language Preservation: Opportunities and Challenges, ResearchGate.
24. Sourati, Z., Karimi-Malekabadi, F., & Ozcan, M. (2025) The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models, ResearchGate.
25. Melero, M. (2024) The Future of Language (and Cultural) Diversity in the Age of AI, CLARIN.
26. Perlin, R. (op cit)
27. Russo Carroll, S., Rodriguez Lonebear, D., & Martinez, A. (2017). Data Governance for Native Nation Rebuilding, Native Nations Institute.
28. Te Mana Raraunga. (2018). Frequently Asked Questions, Te Mana Raraunga.
29. Ackermann, A. (op cit)
30. Walter, M., Kukutai, T., Carroll, S. R., & Rodriguez-Lonebear, D. (Eds.). (2020). Indigenous Data Sovereignty and Policy. Taylor & Francis, p. 24
31. Mallory, F. (op cit)
32. Cymdeithas yr Iaith (2022) Cymru Rydd, Cymru Werdd, Cymru Gymraeg., p. 27
33. O’Sullivan, L. (2025). How AI’s Failure on Linguistic Diversity is Deepening Global Inequality, RESET – Digital for Good.
34. Harvey, F. (2024). Global water crisis leaves half of world food production at risk in next 25 years, The Guardian.
35. Erdocia, I., Migge, B., & Schneider, B. (2024). Language is not a data set—Why overcoming ideologies of dataism is more important than ever in the age of AI. Journal of Sociolinguistics, 28(5), p. 23
36. O’Sullivan, L. (op cit)
37. Erdenesanaa, D. (2023) A.I. Could Soon Need as Much Electricity as an Entire Country, The New York Times
38. Heikkilä, M. (2022) We’re getting a better idea of AI’s true carbon footprint, MIT Technology Review.
39. Judah, J. (op cit)#4 #AI #ArtificialIntelligence #Breus #Cornish #Cornwall #data #generativeAI #history #jynn #kedhlow #Kernewek #Kernow #Kernowek #LLM #machine #PYB #SK #SKDinythus #SkiansKreftus #Sordya
-
Turning Saffron into Slop – Treylya Safran yn Skomblans
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
Note: Artificial Intelligence (AI) has come to be synonymous with Generative AI (GenAI) and with Large Language Models (LLMs), such as ChatGPT, in common parlance. Unless explicitly stated, I use the terms interchangeably.
Kernewek is under attack. The attacker? Machine-made rubbish. Fresh from companies dictionary-bashing to make terrible ‘translations’ for their black-and-gold-washing brandification of Kernow*, the shoddiness has spiralled.
Error-riddled AI ‘Kernewek textbooks’ have appeared on Amazon, by ‘authors’ who are at best well-meaning but harmful and at worst out to exploit us. Worse, a prominent crackpot is ‘translating’ conspiracy theories into ‘Cornish’ en masse. It’s not just nonsensical; it ties our language to fascism faster than we, making content by hand, can work to untie it.
There are those who believe that the best defence is to put down our shield and join the opposing forces: to ‘buy in’ to AI in the hope of coming out the other side with a useful tool for the language and a stronger community. Such hopes must be abandoned. What follows is a look why this approach is wrong-headed, as evidenced by universities, activists and indigenous groups.
LOW-RESOURCES AND LINGUISTIC TYPOLOGY
Simply adding a language to an AI model leads to a spike in poor-quality articles, drowning out quality writing by humans. AI has “industrialized the acts of destruction—which affect vulnerable languages most, since AI translations are typically far less reliable for them.”1 Wikipedia editors from varied languages evidence that machine translation tools have made it easier than ever before to create shoddy articles in minoritised languages, causing massive damage in minutes. AI leads to non-speakers producing much longer, truthier rubbish, Sámi computational linguistics expert Trond Trosterud notes: “the problem [is] that they are armed with Google Translate. Earlier they were armed only with dictionaries.”1
Kernewek, like all but 60 of the world’s roughly 7,000 languages, is designated “low-resource”, meaning it lacks sufficient data to train a machine.2 It is tempting, therefore, to assume that the solution is to provide more data. However, training an LLM requires petabytes of text, audio and video—manually categorised and in a machine-readable format—a vast trove that Kernewek simply does not have.3 Professor Will Lamb, Chair of Gaelic Ethnology and Linguistics at Edinburgh University, speaks of “millions of work hours devoted to just one aspect” of a working AI.4
Even if ChatGPT is trained on another language than English, the time and labour required may make it largely unviable. Current assessments of the performance of ChatGPT for different languages have shown that it performs worse in all tasks.5
Prof. Lina Dencik, Data Justice Lab
Furthermore, the amount of data and work is not the only barrier; at issue is the nature of the language itself. Microsoft has found that languages such as Breton—and thus Kernewek—cause a high rate of errors distinct from the size of their dataset, due to grammatical features, such as mutation, not present in well-sourced languages. As such, they remain poor without significant additional work.6 Essentially, simply adding more Kernewek may not help. Thus, engaging with AI is, for Kernewek, to tie ourselves to slop.
Noten: Skians Kreftus (SK) re dheuth ha bos kesstyr gans SK Dinythus (SKDin) ha gans Patronyow Yeth Bras (PYB), kepar ha ChatGPT, yn lavar kemmyn. Marnas bos menegys yn kler, my a us an termys yn keschanjyadow.
Kernewek yw yn-dann omsettyans. An omsettyer? Atal gwrys dre jynn. Nowydh devedhys a gompanis ow pylla gerlyvrow rag gul ‘treylyansow’ euthyk rag aga merkegyans yethwolghi a Gernow*, an pilyekter re wrug pesya.
‘Dysklyvrow’ ‘Kernewek’ gwallblagys re apperyas war Amazon, gans ‘awtours’ neb yw teg aga thowl dhe’n gwella ha drogusus aga hwans dhe’n gwettha. Lakka, yma koyntwas a vri ow ‘treylya’ tybiethow kesplottyans dhe ‘Gernewek’ yn routh. Nyns yw gocki hepken; y kelm agan yeth orth faskorieth uskissa es dell yllyn, dre wul dalgh dre leuv, oberi dh’y digelmi.
Yma nebes a grys bos agan gwella difres gorra an skoos dhyworthyn ha junya an ostys er agan pynn: dhe ‘unverhe’ gans SK gans govenek dos yn-mes gans toul dhe les rag an yeth ha kemeneth kreffa. Res yw hepkor govenegow a’n par na. An pyth hag a sew a vir orth prag yth yw an devedhyans ma penn-gam, dell yw dustunys gans pennskolyow, gweythresoryon ha bagasow teythyek.
ASNODHOW ISL HA TIPOLOGIETH YETHEL
Keworra yeth yn sempel orth patron SK a led orth spik yn erthyglow drog aga kwalita, ow peudhi skrif a gwalita gans tus. SK re wrug “diwysyansegi an aktys diswrians—hag a nas yethow goliadow an moyha, drefen bos treylyansow SK lieskweyth le lel yn tipek ragdha.”1 Golegydhyon Wikipedia a yethow divers a re dustuni re wrug medhelweyth-treylya y wul bos esya dell veu bythkweth kyns gwruthyl erthyglow pilyek yn yethow lyharivhes, ow kawsya damach kowrek yn mynysennow. SK a led orth digowsoryon owth askorra atal lieskweyth hirra ha gwirekka, konnyk yethonieth reknansek Sámi Trond Trosterud a not: “an kudyn [yw] aga bos ervys gans Google Translate. A-varra nyns ens ervys marnas gans gerlyvrow.”1
Kernewek, kepar hag oll marnas 60 a ogas lowr 7,000 yeth a’n bys, yw klassys avel “isel y asnodhow”, ow styrya nag eus dhodho kedhlow lowr dhe drenya jynn.2 Rakhenna, dynyek yw desevos bos an assoylyans profya moy a gedhlow. Byttegyns, res yw petavaytys a dekst, son ha gwydhyow—klassys dre leuv hag yn furvas redyadow gans jynn— dhe drenya PYB, tresorva efan nag eus dhe Gernewek yn sempel.3 Y kews Professor Will Lamb, Kaderyer Ethnologieth ha Yethonieth Wodhalek orth Pennskol Karedin, a “vilvilyow a ourys ober sakrys orth unn wedh hepken” a SK owth oberi.4
Hogen mars yw ChatGPT trenys war yeth a-der Sowsnek, an termyn hag ober yw res a styr y vos martesen anhewul dre vras. Arvreusyansow a-lemmyn a berformyans a ChatGPT rag yethow dyffrans re dhiskwedhas y perform gweth yn oberennow oll.5
Prof. Lina Dencik, Data Justice Lab
Pella, nyns yw an myns a gedhlow hag ober an unsel lett; a vern yw natur an yeth y honan. Microsoft re drovyas y kaws yethow kepar ha Bretonek—hag ytho Kernewek—kevradh ughel a wallow diblans a vraster aga sett kedhlow, drefen nasyow gramasek, kepar ha treylyansow, nag usi kevys yn yethow ughel aga asnodhow. Yndella, i a bes orth bos drog heb meur a ober keworransel.6 Yn essensek, possybyl yw ny wra keworra moy a Gernewek yn sempel gweres. Yndelma, oberi gans SK yw, rag Kernewek, omgelmi orth skomblans.
CORNISH UNDER CAPITALISM
But surely we can improve things over time? It will take a lot of help from AI companies, but it will be worth it. Sadly, Gabriel Nicholas, a research fellow at the Center for Democracy and Technology, has found that once a tech company has established basic capabilities for a language, they pat themselves on the back and move on.7
Big tech companies are just that: companies. They exist to make a profit. Unfortunately, a market dominated by big languages gives them no incentive to invest in improvements for small ones.
All of the speech technology, smart homes and voice interaction systems used today are the products of commercial research. To put it bluntly, they exist to either make money from your data, to sell you more goods and services, or to influence your thinking. None of this AI exists for the public good. […] Unless there is a strong enough economic argument, don’t expect big companies to rush into producing Welsh, Gaelic or Cornish speech systems.8
Prof. Ian McLoughlin, University of Kent
Should they decide that a Kernewek AI is a viable profit-making enterprise, our situation may even be worse than abandonment. As Dr. Fintan Mallory remarks, the dominant means of profit for privately-funded AI enterprises is to convert their tools into surveillance devices.9 As Kernewek is currently one of the UK’s only languages which is not currently easily surveillable, this poses a huge risk to Kernewek activism and the fight for self-determination in a state that seeks to criminalise dissent.
While we’re on the subject of Kernewek and its position under capitalism, let’s consider the human cost. I lost my 13-year career in language to AI as soon as English output became viable enough to excuse not paying a human. In the unlikely instance that we achieve an AI that can produce quality Kernewek, why would anyone bother paying speakers? The idea of AI sucking all the life out of my heritage language when we are struggling to survive as-is is appalling.
Simply put, profit is antithetical to people. While AI is the new favourite toy of profit, it will be antithetical to people. And a language is its people.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Combinations of characters on a screen mean nothing without agency and intention.10
Ross Perlin, Endangered Language Alliance
While language is not unique to humans, it is one of the chief parts of being human. It cannot be reduced to mere data, but is a highly social process.11 We all know how synthetic customer support via robot sounds or how AI fails to pick up nuance. As Dr. Mallory comments, “Language [is] something more like the soul of a community. You can’t store this in a machine. You can’t solve a human problem like linguicide with a view of language that removes the human component.”12
AI cannot comprehend Kernewek or any other language. It is a stochastic parrot: predicting what word is likely to follow the previous one.13 It cannot understand us. It cannot intend anything. If it tells you it feels delighted to help you, it is lying. I want our community to grow, but one hundred ‘Cornish-speaking’ computers do not add to it. One human does—bringing ideas and hopes and fears and foibles—and I do not think the Kernewek ‘speaking’ computers will add even one human to our community.
Worse, if it does, there is evidence from Microsoft to suggest that the use of GenAI on language tasks, even once a week, impairs cognitive ability to learn, leading to decreased engagement with the topic, overreliance on the technology and hobbled skills in independent problem-solving.14 By using AI tools to ‘teach’ a learner Kernewek, we may in fact be impairing their ability to learn the language at all without this crutch. We will make regurgitators in place of speakers.
Perlin also emphasises the human element, saying that when we hold community central to our languages, as we do, the stochastic parrot can feel like a violation.15 At the moment, I can tell when someone is using AI ‘Kernewek’ to me. The idea that one day I will not know when an outsider—someone I would welcome if they took up a book or a class—is puppeting my ancestors’ jaws and speaking through them is ghoulish. It has the instant sting of colonialism, of appropriation when one could appreciate, of parroting when one could join our chorus.
Hawai’ian scholar Ha‘alilio Solomon agrees: “It is painful, because it reminds us of all the times that our culture and language has been appropriated. We have been fighting tooth and nail in an uphill climb for language revitalization.[…] People are going to think that this is an accurate representation of the Hawaiian language.”16
TRUST AND COMMUNITY FEELING
The anti-machine backlash has long been simmering but is now seemingly breaking to the surface.17
NBC NEWS
The explosion of insults for AI itself (clanker, tinskin, toaster), its output (slop, dross, brainrot) and its users (slopper, groksucker, botlicker, second-hand thinker)—as well as others more clearly based on real-world slurs than I am comfortable to include—tells a tale of the general attitude of distrust and disgust towards the technology and its use on anglophone and other majority language internet.18 While the attitude among tech bros and corporates remains bombastic, for the general public AI is “becoming interchangeable with things that sort of suck.”19
Further, it’s not just majority languages with this negative view of AI as taint. A quick sampling of social media comments and likes regarding AI and Scottish Gaelic by Professor Lamb showed a split of 54% negative, 33% positive and 13% neutral. (Lamb, 2024) The sentiment of the top-rated negative comment was that AI is harmful and the second-highest that AI should be kept away from heritage languages.
What are we telling our descendants? That our language and culture isn’t worth the personal effort? That’s how I might read it, if I were them.20
Kernewek survey respondent
Kernewek paints an even starker picture, especially among younger and more technologically-savvy learners and speakers. A survey on Cornish Discord and Whatsapp found that 65% felt AI would be bad (11.5%) or very bad (53%) for the language. When asked what the community response should be to AI, 46% said we should prevent it and 27% avoid it, with only over-60s thinking that we should work with it.20
31% of respondents said using AI in Kernewek would cause them to feel estranged from the language, while 54% said that they would feel strongly estranged and 23% a little estranged from any organisation, resource or teacher using AI.
The response from those who gave their knowledge of AI as either “expert” or “good” was particularly damning. Everyone in this group responded that AI would be harmful for the language, that the use of AI would estrange them from a source strongly and that we should prevent the use of AI for Kernewek.
IDENTITY, AUTHENTICITY AND DIVERSITY
Aristotelis Ioannis Paschalidis, writing for UNESCO, was not speaking specifically about minoritised languages when he asked this, but the question resonates even more strongly for us: “How much loss of identity is one willing to sacrifice for efficiency?”21
Identity is of paramount importance to Kernewek speakers. Ute Wimmer’s study Reversing Language Shift: the Case of Cornish identified the language’s “function as a symbol of national identity” as the second highest motive (66%**) among speakers and learners, beaten only by Cornish culture (80%).22 This would seem cause for celebration, but when AI is added to the mix, it becomes a risk. Vincent Koc of Hyperlink states that AI can “inadvertently contribute to the dilution of language and cultural identity.”23
He also identifies that automating language learning or generation “may diminish the richness and authenticity that comes from human speakers who carry cultural histories in their speech.” Indeed, four studies by the University of Southern California have shown that using LLMs to assist writing “is linked to notable declines in linguistic diversity and may interfere with the societal and psychological insights language provides.”24
This is in English, one of the richest and largest languages in the world. Imagine the possible impact on a smaller language like Kernewek—with less documentation, less data, a tiny speakerbase and basically no money—and on its many language varieties and orthographies. Particular to the Kernewek context, Late speakers are already struggling to be seen as valid under the dominance of Middle. Do we think AI knows the difference? Thoughtlessly, it will either mix everything together, confusing everyone, or it will use Middle to overwhelm Late.
Generative AI-driven content creation, by favoring standardized languages, risks the disappearance of regional dialects.25
Barcelona supercomputing Center ….
Not only are varieties at risk; AI threatens to drown Kernewek as a whole. Perlin agrees that the linguistic flattening that occurred over centuries in English could manifest overnight in a minoritised language with AI at the helm—as it would be, being able to effortlessly outstrip human Kernewek. He raises concerns of LLMs freezing a language in place and even defining what it means to know the language, especially with low numbers of native speakers.26
Garbage translations multiply online like fake news. Native speakers of the languages in question are bypassed as being “too hard to find,” compared with automated methods of vetting that are completely disconnected from real-life communication. While larger and more powerful language communities may be able to hold the bots to account and even make strategic use of them, it is all too easy to imagine [a minority language] being overwhelmed.26
Ross Perlin, Endangered Language Alliance
Uncontrolled and in the hands of tech giants, synthetic Kernewek will outnumber and outmanoeuvre human Kernewek.
DATA SOVEREIGNTY AND COLONIALISM
Indigenous data sovereignty is the right of [an indigenous nation] to govern the collection, ownership, and application of its own data.27
Native Nations Institute
There are, however, indigenous cultures that are working on a more equitable relationship with AI. Tech without the giant requires resources, but it allows communities to retain data sovereignty over the cultural asset that is their language. Te Mana Raraunga, the Māori Data Sovereignty Network, has created a list of principles for the creation, use and sharing of Māori data, prioritising the need to enhance control for current and future Māori.
They raise a key point that should be considered carefully by stewards of linguistic and cultural knowledge: “Data from us, and about us and our resources, are valuable assets. Once control of it is lost, it is difficult to regain.”28 Decisions must not be taken lightly or hastily; we can always say “yes” if we have previously said “no” to a particular dataset’s use, but can never say “no” if we have already said “yes”.
The AI field, like any other space, is occupied by people who are set in their ways and unintentionally have a very colonial perspective.29
Michael Running Wolf, First Languages AI Reality
This is vital in the context of the potential control of Kernewek data by powerful external corporations. Capitalist extractivism has long been a bane on societies in the imperial periphery and our Cornish society is no different, having faced centuries of its wealth and natural resources being stripped and sold by and large for the profit of those outside Cornwall.
The book Indigenous Data Sovereignty and Policy notes that current data relations can be seen as “a continuation of the processes and underlying belief systems of extraction, exploitation, accumulation and dispossession that have been visited on Indigenous populations through historical colonialism.”30 This extractive understanding of information is, they note, not disrupted but rather replicated by paying people for their data.
Ultimately, our language must not lie in outside hands governed by proprietary principles that do not allow us sufficient sovereignty over one of our most valuable natural resources: our language. We must have open data principles, not bow to corporate control. We must steer and steward the use of our data, rather than expose it to use against our interests and for the pockets of big tech.
Rather than approaching language preservation as a technical problem, I think indigenous communities need to be politically empowered, whether that be funding from governments or legal protections to use their languages.31
Dr. Fintan Mallory, Durham University
We must prioritise language-as-community and seek open, equitable and ethical use of our language, heritage and other cultural assets. We must avoid thinking of AI as the magic that it promises and invest in basic research, driven by our own community. Corporations will not save us and, indeed, may do us great harm.
NO CORNISH ON A DEAD PLANET
Global capitalism and governments […] are addicted to ‘free’ market ideology over the wellbeing of communities, people and the planet.32
Cymdeithas yr Iaith Maniffesto 2022
Honestly, most takedowns of AI would have hit this point already. It’s one of the main arguments against Generative AI, but in case you’re not familiar with it, we will briefly look over the main points.
Water used in cooling AI data centers must be drinkable water. AI guzzles this water. The University of California has reported that “global water demand from AI could reach 4.2-6.6 billion cubic meters by 2027. That exceeds 50 percent of the UK’s annual water use in 2023.”33 All this while the Global Commission on the Economics of Water has declared “a rapidly accelerating water crisis” to which Kernewek should not be contributing.34
We have become utterly dependent on private technologies manufactured and controlled by a handful of opaque companies [who] appear mostly indifferent to the social consequences of their activities and only invest minimally if obliged by government regulations to enhance their public image.35
Iker Erdocia, Dublin City University
AI requires vast quantities of hardware at the cost of mining rare earth minerals. These are difficult to extract and purify and come with heavy environmental and social costs. They are often extracted from mines in countries with poorer environmental and labour protections. Reset states that “communities living near these mines, often indigenous or minority groups, regularly face land degradation, water contamination and human rights abuses. Much of this can be directly linked to the AI hardware.”36 When the hardware inevitably cooks and is useless, it is then thrown out as e-waste into poor communities. The potential advancement of Kernewek must not come at the expense of our sister indigenous and minority communities.
Training an also AI requires huge amounts of energy, soon perhaps as much as a small country37 and has an enormous carbon footprint.38 What is clear is that—through water usage, extractive industry, energy consumption and carbon footprint—AI is bad news for the struggling environment of the planet we live on and there is no Cornish on a dead planet.
MAKING AI AN EX-PARROT
Rather than making minority languages more accessible, AI is now creating an ever expanding minefield for students and speakers of those languages to navigate.39
mit technology review
We have heard of the vast improbability of getting AI to be able to mimic Kernewek in light of the costs in data, work, time and technology. We have considered the likely choice of cold negligence or surveillance product and the importance of data sovereignty. We have read about the effects on the livelihoods of Cornish speakers, as well as the the catastrophic costs to the environment and indigenous peoples.
We have learned that linguistic flattening by AI impoverishes its subjects and how AI may decide for us how our language must operate. We have seen the inescapability of language as human and the risks of creating ‘learners’ who cannot learn and ‘speakers’ who cannot speak. We have seen the dangers to reputation and trust for any organisation who would shovel what is seen as ‘slop’.
We have heard why giving in to the juggernaut of AI would be a mistake for Kernewek and how our community does not support our laying down of the shield. Instead, we must fight. We must make Kernewek a space as free of slop as possible, we must educate botlickers into ethical and effective language learning and use, we must avoid second-hand thinking.
We must make our language a no AI zone, a network of reliable humans and their human creations, built on authenticity, community, effort and trust: a Kernewek for the people, of the people and by the people.
KERNEWEK YN-DANN GEVALAV
Mes yn sur y hyllyn ni gwellhe taklow dres termyn? Y fydh res meur a weres a gompanis SK, mes y talvia dhyn. Yn trist, Gabriel Nicholas, kesvroder hwithrans orth an Center for Democracy and Technology, re drovyas pan wrug kompani tek fondya gallosow selyek rag unn yeth, i a omgeslowenha yn ughel hag ena movya yn-rag.7
Kompanis tek bras yw yndella poran: kompanis. Ymons i ena rag gwaynya budh. Y’n gwettha prys, ny wra marghas rewlys gans yethow bras ri kentryn dhe gevarghewi yn gwellhe rag an re byghan.
Oll a’n deknegieth kows, chiow konnyk ha systemow ynterweythres lev usys hedhyw yw an askorrasow a hwithrans kenwerthel. Dhe vos sogh, yth yns i po rag dendyl arghans a’th kedhlow, po gwertha gwara ha gonisyow, po delenwel dha dybyansow. Nyns yw tra vyth a’n SK ma rag an les kemmyn. […] Mar nag eus argyans erbysek krev lowr, na wra gwaytya kompanis bras dhe fyski dhe askorra systemow kows Kembrek, Godhalek po Kernewek.8
Prof. Ian McLoughlin, pennskol kint
Ha mars ervirons bos SK Kernewek aventur a yll gwaynya budh, possybyl yw bos agan studh gweth ages dell via gans forsakyans. Dell lever Dr. Fintan Mallor, an fordh vrassa a waynya budh rag kompanis SK arghesys yn privedh yw kedreylya aga thoulys yn devisyow aspians.9 Drefen bos Kernewek onan a’n yethow boghes y’n RU nag yw aspiadow yn es y’n eur ma, hemm yw peryl kowrek rag gweythresieth Kernewek ha’gan strif a-barth omdhetermyans yn stat a vynn galweythegi dissent.
Ha ni ow tochya Kernewek ha’y savla yn-dann gevalav, gwren ni mires orth an kost denel. My a gellis ow soodh 13 bloodh yn yethow dhe SK kettooth ha dell veu eskorrans Sowsnek hewul lowr dhe askusya sevel orth tyli den. Y’n kas diwirhaval may kevyn SK hag a yll askorra Kernewek da, prag y hwrussa nebonan omankombra ow pe kowser? An tybyans a SK ow tenna oll an bewnans a’m taves ertach ha ni ow kwynnel dhe dreusvewa dell on yw skruthus.
Yn sempel, budh yw gorthenebel orth tus. Hedre vo SK an degen nowydh flamm a vudh, y fydh gorthenebel orth tus. Ha yeth yw hy thus.
KENEDHEL HEB YETH, KENEDHEL HEB KOLON
Nyns eus styr dhe gesunyansow a lytherennow war skrin heb dewis ha heb mynnas.10
Ross Perlin, Endangered Language Alliance
Kyn nag yw yeth dibarow dhe dhensys, onan a’n rannow chif a vos denel yw. Ny yll bos lehes dhe gedhlow hepken, mes yth yw argerdh sosyel dres eghen.11 Ni oll a wor py mar synthesek y sen skoodhyans prener der SK po fatel yll SK fyllel orth konvedhes arliwyow. Dell gampol Dr. Mallory, “Yeth [yw] neppyth moy kepar hag enev a gemeneth. Ny yllir gwitha hemma yn jynn. Ny yllir assoylya kudyn denel kepar ha yethladhans gans gwel a yeth hag a remov an gerann denel.”12
Ny yll SK konvedhes Kernewek po taves vyth aral. Papynjay chonsus yw: y targan py ger yw gwirhaval wosa an huni kyns.13 Ny yll agan konvedhes. Ny yll mynnes tra vyth. Mar kwra derivas orthis y vos pes da dha weres, gow yw. My a vynn agan kemeneth dhe devi, mes ny wra kans jynn-amontya a yll ‘kewsel Kernewek’ keworra orti. Y hwra unn den—ow tri tybyansow ha govenegow hag ownow ha gwanderyow—ha ny dybav y hwra an jynnys-amontya kernwegorek keworra unn den hogen orth agan kemeneth.
Gwettha, mar kwra, yma dustuni a-dhyworth Microsoft hag a brof y hwra an devnydh a SKDin war oberennow yeth, unweyth an seythen hogen, aperya gallos godhvosel a dhyski, ow ledya orth omworrans lehes gans an desten, gorfydhyans y’n deknegieth ha sleyneth sprallys a assoylya kudynnow yn anserghek.14 Der usya toulys SK dhe ‘dhyski’ Kernewek, possybyl yw ni dhe shyndya gallos dyski an yeth vytholl heb an kroch ma. Ni a wra gul mimyoryon yn le Kernewegoryon.
Ynwedh Perlin a boslev an elven dhenel, ow leverel pan wren ni synsi kemeneth avel kres agan yethow, dell wren, an papynjay chonsus a yll bos klewys kepar ha defolyans.15 Y’n eur ma, my a aswon pan eus nebonan owth usya ‘Kernewek’ SK dhymm. An tybyans ny wrav vy unn jydh godhvos pan eus estren—nebonan a wrussen vy dynerghi mar pe lyver po klass ganses—ow popettya diwawen ow hengerens ha kewsel dresta yw bedhrosus. Yma dhe’n dra an wan dhistowgh a drevesigeth, a berghenegyans pan yllir gwerthveurhe, a bapynjaya pan yllir junya agan kesgan.
Unver yw skolheyk Hawai’i henwys Noah Ha‘alilio Solomon: “Ankensi yw, drefen ni dhe vos kofhes a’n prysyow oll re beu agan gonisogeth ha yeth perghenegys. Ni re beu owth omladh dre dhens hag ewines yn batel gales a-barth dasvewheans yeth.[…] Y hwra pobel krysi bos hemma representyans ewn a’n yeth a Hawai’i.”16
TREST HAG OMGLEWANS AN GEMENETH
Hir re beu an kil-lash gorthjynn ow kovryjyon mes lemmyn yma va ow terri an arenep dell hevel.17
NBC NEWS
Tardh an arvedhennow rag SK y honan (clanker, tinskin, toaster), y askorras (slop, dross, brainrot) ha’y usyoryon (slopper, groksucker, botlicker, second-hand thinker)—keffrys hag erel selys moy yn kler war geryow kas gwir dell ov attes gans aga heworra—a re hwedhel a stons ollgemmyn a wogrys ha divlases war-tu hag an deknegieth ha’y devnydh war an kesrosweyth Sowsnek ha yethow bras erel.18 Kynth yw an stons yn-mysk gwesyon dek ha korforeth hwath gwresek, rag an boblek gemmyn y hwra SK “dos ha bos keschanjyadow gans taklow tamm kawgh.”19
Pella, nyns yw marnas yethow moyhariv gans an gwel negedhek ma a SK avel podrek. Sampel uskis a gampollow media sosyel ha meusi ow tochya SK ha Godhalek Alban gans Professor Lamb a dhiskwedhas fals a 54% negedhek, 33% posedhek ha 13% heptu. (Lamb, 2024) Sentiment an kampol negedhek an moyha talvesys o bos SK dregynnus hag an nessa y talvia dhyn lettya SK rag kestav gans tavosow ertach.
Pyth eson ni ow leverel orth agan diyskynysi? Ny dal agan yeth ha gonisogeth an strivyans personel? Hemm yw martesen fatel wrussen vy y redya, a pen vy i.20
Gorthebydh sondyans Kernewek
Kernewek a baynt aven moy serth, yn arbennik gans dyskoryon ha kowsoryon yowynka ha moy skentel gans tek. Sondyans war Discord ha Whatsapp Kernewek a drovyas bos 65% a grysis y fia SK drog (11.5%) po pur dhrog (53%) rag an yeth. Pan veu govynnys pyth a dal bos gorthyp an gemeneth orth SK, 46% a leveris y kodh y hedhi ha 27% y woheles, gans an dus moy ha 60 bloodh hepken ow tybi y kodh oberi ganso.20
31% a worthebydhyon an sondyans a leveris y hwrussa an devnydh a SK yn Kernewek aga fellhe a’n yeth, hag ynwedh 54% a leveris y fiens i pellhes yn krev ha 23% pellhes tamm a by kowethas, asnodh po dyskador pynag ow tevnydhya SK.
An gorthyp a’n re a leveris bos aga godhvos a SK po “konnyk” po “da” o dampnus yn arbennik. Pubonan y’n bagas ma a worthebis y fia SK dregynnus rag Kernewek, y hwrussa an devnydh a SK gans pennfenten aga fellhe a’n bennfenten na yn krev hag y kodh dhyn hedhi an devnydh a SK rag Kernewek.
HONANIETH, LELDER HA DIVERSETH
Nyns esa Aristotelis Ioannis Paschalidis, ow skrifa a-barth UNESCO, ow kewsel yn komparek a-dro dhe yethow lyharivhes pan wrug ev y wovyn, mes an govyn a dhassen yn kreffa ragon: “Pygemmys koll a honanieth a vynnir sakrifia rag effeythuster?”21
Honanieth yw a’n moyha bri rag Kernewegoryon. Studhyans Ute Wimmer Reversing Language Shift: the Case of Cornish a henow “gweythres [an yeth] avel arwodh a honanieth kenedhlek” avel an nessa ughella skila (66%**) yn-mysk kowsoryon ha dyskoryon, fethys gans gonisogeth Kernow (80%) hepken.22 Yth havalsa hemma bos acheson solempnyans, mes pan vo SK keworrys, y teu ha bos peryl. Vincent Koc a Hyperlink a lever y hyll SK “kevri dre wall orth an gwannheans a yeth ha honanieth wonisogethel”.23
Ev a aswon ynwedh y hallsa awtomategi dyski po dinythi yeth “lehe an rychedh ha lelder hag a dheu a gowsoryon dhenel neb a dheg istoriow gonisogethel y’ga hows”. Yn hwir, peswar studhyans gwrys gans Pennskol Kaliforni Soth re dhiskwedhas bos devnydhya PYB dhe weres gans skrifa “kelmys orth dyfygyansow nosedhek yn diverseth yethel hag y hyll mellya gans an konvedhes brysoniethel ha kowethasel yw proviys gans yeth.”24
Ha hemm yw yn Sowsnek, onan a’n yethow an ryccha ha brassa y’n bys. Dismyk an effeyth war yeth byghanna kepar ha Kernewek—gans le a dhogvennans, le a gedhlow, sel kowsoryon munys hag ogas hag arghans mann—ha war y lies orgraf hag eghen yeth. Yn arbennik yn gettesten Kernewek, seulabrys yma kowsoryon Diwedhes ow strivya dhe vos gwelys avel vas gans gwartheyvans Kres. A dybyn y hwor SK an dyffrans? Heb preder, y hwra po kemyska puptra warbarth, ow sowdheni pubonan, po devnydhya Kres dhe fetha Diwedhes.
An gwruthyl a dhalgh herdhys gans SK Dinythus, dre favera yethow savonegys, a argyl an vansyans a rannyethow ranndiryel.25
Kresen woramontyorieth Barcelona
Nyns yw eghennow hepken yn peryl; SK a wodros beudhi Kernewek yn tien. Akordys yw Perlin y hallsa an platheans yethel a hwarva dres kansbledhynnyow yn Sowsnek hwarvos dres nos yn yeth lyharivhes gans SK orth an fronnow—dell via, ow pos gallosek a bassya Kernewek denel heb assay. Ev a venek prederow yn kever PYB ow rewi yeth yn hy le ha hogen ow settya pyth yw an styr a wodhvos an yeth, yn arbennik gans niverow munys a gowsoryon deythyek.26
Treylyansow leun a atal a liesha warlinen kepar ha nowodhow fug. Kowsoryon deythyek a’n yethow ma yw passyes avel bos “re gales dhe drovya”, komparys orth fordhow awtomategys a surheans kwalita hag yw disjunys yn tien a geskomunyans y’n bys gwir. Kynth yw possybyl rag kemenethow yeth brassa ha moy gallosek synsi an bottys ma dhe akont ha’ga devnydhya yn stratejek hogen, re es yw dismygi [yeth lyhariv] ow pos reverthys.26
Ross Perlin, Endangered Language Alliance
Heb kontrol hag yn diwla an gewri deknegieth, Kernewek synthesek a wra gornivera ha gorthrabellhe Kernewek denel.
SOVRANEDH KEDHLOW HA KOLONEGIETH
Sovranedh kedhlow teythyek yw an gwir gans [kenedhel teythyek] a woverna an kuntel, perghenogeth ha gweytha a’y hedhlow hy honan.27
Native Nations Institute
Byttegyns, yma gonisogethow teythyek hag usi owth oberi war geskowethyans moy ewnhynsek gans SK. Tek heb an kowr a res asnodhow, mes y as kemenethow gwitha sovranedh kedhlow war an gerthen wonisogethel hag yw aga yeth. Te Mana Raraunga, Rosweyth Sovranedh Kedhlow Māori, re wrug rol a bennrewlys rag an gwruthyl, devnydhya ha kevrenna a gedhlow Māori, ow ragwirhe an edhom a grefhe maystri rag Māori a-lemmyn hag a dheu.
I a venek poynt posek hag a dalvia bos konsidrys gans rach gans stywards a skians yethel ha gonisogethel: “Kedhlow ahanan, a-dro dhyn ha’gan asnodhow, yw kerthennow a bris. Pan vo maystri kellys, kales yw y dhaskemeres.”28 Ny dal gul erviransow yn skav po yn uskis; y hyllyn pupprys leverel “ea” mar kwrussyn leverel “na” kyns orth us sett kedhlow, mes ny yllyn nevra leverel “na” mar kwrussyn leverel “ea” seulabrys.
An desten SK, kepar ha pub le aral, yw leun a dus hag yw settys y’ga maneryow ha gans gwel pur drevesigel yn tidowl.29
Michael Running Wolf, First Languages AI Reality
Hemm yw pur bosek y’n gettesten a’n kontrol possybyl a gedhlow Kernewek gans korforethow gallosek a-ves. Estenegieth jatelydhek re beu molleth war gowethasow y’n amal emperourethek ha nyns yw kowethas Kernewek dyffrans, wosa enebi kansvledhynnyow a’y rychys hag asnodhow naturek ow pos destryppys ha gwerthys dre vras gans budh tus yn-mes a Gernow.
An lyver henwys Indigenous Data Sovereignty and Policy a verk lemmyn y hyllir gweles perthynyansow kedhlow avel “pesyans a’n argerdhow ha systemow-krysi isworwedhek a estennans, drogusyans, kuntellyans ha diberghenogeth re beu gwrys war boblansow Teythyek dres trevesigeth istorek.”30 An konvedhes estennek ma a gedhlow yw, dell verkons, hevelebys a-der goderrys gans tyli pobel rag aga hedhlow.
Wostiwedh, res yw ma na vo agan yeth gorrys yn diwla a-ves routys gans pennrewlys perghenogel na as dhyn sovranedh lowr a onan a’gan asnodhow naturel an moyha posek: agan yeth. Res yw dhyn kavos pennrewlys kedhlow ygor, a-der plegya orth kontrol korforethel. Res yw dhyn lewya ha gidya an devnydh a’gan kedhlow, a-der y usya erbynn agan lesow ha rag pocketys tek bras.
A-der drehedhes an arwithans a davosow avel kudyn teknegiethel, my a dyb bos res dhe gemenethow teythyek bos reythhes yn politek, po der arghasans a wovernansow po dre dhifresyansow laghel dhe dhevnydhya aga yethow.31
Dr. Fintan Mallory, Pennskol Durham
Res yw dhyn ragwirhe yeth-avel-kemeneth ha hwilas devnydh ygor, ewnhynsek hag ethegel a’gan kerthennow yeth, ertach ha gonisogethel. Res yw dhyn goheles tybi a SK avel an hus mayth ambos ha kevarghewi yn hwithrans selyek, lewys gans agan kemeneth. Ny wra korforethow agan selwel ha, hogen, i a yll agan shyndya.
NYNS EUS KERNEWEK WAR BLANET MAROW
Governansow ha kevalav ollvysel […] yw omres dhe ideologieth marghas ‘rydh’ moy es dell yns omres dhe sewena kemenethow, pobel ha’n planet.32
Cymdeithas yr Iaith Maniffesto 2022
An brassa rann a vreusyansow a SK a wrussa meneges hemma seulabrys. Onan a’n argyansow brassa yw erbynn SK Dinythus, mes rag own bos ankoth dhis, ni a wra mires orth an chif boyntys.
Res yw bos evadow an dowr goyeynhe pub kresen kedhlow SK. Y kollenk an dowr ma. Pennskol Kaliforni re dherivas “y hallsa demond dowr ollvysel SK hedhes 4.2-6.6 bilvil metrow kubek erbynn 2027. Henn yw moy es 50 kansran a us dowr bledhynnyek an RU yn 2023.”33 Y kettermyn, an Desedhek Ollvysel Erbysieth Dowr a dheklaryas “barras dowr ow tardha yn uskis” ma na dal Kernewek kevri dhodho.34
Ni re dheuth ha bos yn hwir omres dhe deknegiethow privedh gwrys ha kontrolys gans dornas a gompanis diskler [hag] a hevel bos mygyl dre vras orth an sewyansow sosyel a’ga gwriansow ha kevri yn ispoyntel marnas mars yns i konstrinys gans rewlys an wovernans dhe wellhe aga imach poblek.35
Iker Erdocia, Pennskol Sita Dulyn
Yma edhom dhe SK a vynsow kowrek a galesweyth orth kost palas monyow tanow. Kales yw estenna ha purhe an re ma hag yma kostow kerghynedhel ha sosyel poos. Estennys yns i yn fenowgh a hwelyow yn powyow gans difresyansow lakka rag lavur ha’n kerghynnedh. Reset a lever “yn fenowgh y hwra kemenethow yw trigys yn ogas dhe’n hwelyow, yn fenowgh bagasow lyhariv po teythyek, enebi gwethheans an tir, defolyans an dowr hag abusyans gwiryow denel. Meur a hemma a yll bos kelmys yn tidro orth an galesweyth SK.”36 Pan yw an galesweyth kegys yn sertan hag euver, ena tewlys yw avel e-wast yn kemenethow boghosek. Res yw nyns yw an avonsyans possybyl a Gernewek orth kost agan kemenethow hwor lyhariv ha teythyek.
Ynwedh res yw myns hujes a nerth rag trenya SK, yn skon martesen an keth myns ha pow byghan37 hag yma ol troos karbon kowrek.38 Kler yw—der usadow dowr, diwysyans estennek, konsumyans nerth hag ol troos karbon—bos SK yeyn nowodhow rag kerghynnedh ow strivya a’n planet mayth on ni trigys warnodho ha nyns eus Kernewek war blanet marow.
GUL DHE SK BOS EKS-PAPYNJAY
A-der gul dhe yethow lyhariv bos moy hedhadow, lemmyn yma SK ow kwruthyl tardhek pupprys owth omlesa rag studhyoryon ha kowsoryon a’n yethow ma dhe wolya.39
mit technology review
Ni re glewas a’n anwirhevelepter efan a wul dhe SK gallos mimya Kernewek yn golow an kostys yn kedhlow, ober, termyn ha teknegieth. Ni re gonsidras lycklod an dewisynter dispresyans yeyn po askorras-aspia ha’n posekter a sovranedh kedhlow. Ni re redyas a-dro dhe’n effeythyow war vewnansow Kernewegoryon, keffrys ha’n kostys katastrofek rag an kerghynnedh ha poblow teythyek.
Ni re dhyskas y hwra platheans yethel gans SK boghosekhe y destennow ha fatel yll SK martesen ervira a’gan parth fatel godh dh’agan yeth oberi. Ni re welas an anwoheladewder a yeth avel denel ha’n peryllyow a wul ‘dyskoryon’ na yll dyski ha ‘kowsoryon’ na yll kewsel. Ni re welas an peryllyow orth bri ha fydhyans rag kowethasow a wrussa palas an pyth hag yw gwelys avel ‘skomblans’.
Ni re glewas prag y fia omblegya orth an jagganat a SK error rag Kernewek ha dell na vynn agan kemeneth skoodhya gorra an skoos a-dhyworthyn. Yn y le, res yw dhyn batalyas. Res yw dhyn gul dhe Gernewek bos spas mar rydh a skomblans dell yll bos, res yw adhyski orth botlapyoryon yn dyski ha devnydh yeth yn ethegel hag yn effeythus, res yw goheles tybi wortaswerth.
Res yw dhyn gul dh’agan yeth bos parth heb SK, rosweyth a dus fydhyadow ha’ga gwriansow denel, drehevys war lelder, kemeneth, assay ha trest: Kernewek hag yw a-barth an bobel, a’n bobel ha gans an bobel.
Niwlen Ster
Notennow
* A prime example is the laughably-unaffordable restaurant RenMor, which The Headland Hotel thinks is a version of “Re’n Mor”, which they believe means “by the sea” as in “next to the sea” but actually means “by the sea!” like saying “by Zeus!”. This is both hilarious and enraging.
** A figure perhaps lower than it should be if you consider that many of the “emotional motives” which were not counted in this category, such as “I’m Cornish, what better reason do you need?”, do also refer to identity.
FENTENNOW
1. Judah, J. (2025) How AI and Wikipedia have sent vulnerable languages into a doom spiral, MIT Technology Review.
2. Ackermann, A. (2023) When AI doesn’t speak your language, Coda.
3. Crichton, D. (2024) AI and the Death of Human Languages, Lux.
4. Lamb, W. (2024). Could Artificial Intelligence save Scottish Gaelic?, The University of Edinburgh.
5. Dencik, L. (/2025) AI Inequalities: Minority Languages, TUC Cymru.
6. Joshi, P., Santy, S., Budhiraja, A., Bali, K., & Microsoft Research, India. (2020). The State and Fate of Linguistic Diversity and Inclusion in the NLP World. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
7. Ackermann, A. (op cit)
8. McLoughlin, I. (2018) How to teach AI to speak Welsh (and other minority languages), The Conversation.
9. Mallory, F. (2025) RISE UP Panel Discussion & Q&A: What AI Can and Cannot Do for Minoritised Languages, YouTube.
10. Perlin, R. (2024) AI Won’t Protect Endangered Languages, The Dial.
11. RISE UP (2025) #4 RISE UP Event Summary: What AI Can and Cannot Do For Minoritised Languages, RISE UP.
12. Mallory, F. (2024) European Day of Languages: Will lesser spoken languages soon only be kept alive by AI technology? Durham University.
13. Bender, E., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
14. Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025) The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Microsoft.
15. Perlin, R. (op cit)
16. Judah, J. (op cit)
17. Abbruzzese, J., & Wile, R. (2025) Is an AI backlash brewing? What ‘clanker’ says about growing frustrations with emerging tech, NBC News.
18. Webster, K. (2025) Why Using ChatGPT at Work Could Hurt Your Reputation, Inc. Magazine.
19. Herrman, J. (2024) Is That AI? Or Does It Just Suck?, Intelligencer.
20. Wilson, L. (2025) Skians Kreftus ha Kernewek/Artificial Intelligence and Cornish
21. Paschalidis, A. I. (2025) AI and the great linguistic flattening, UNESCO.
22. Wimmer, U. (2010). Reversing Language Shift: the Case of Cornish. Cornish Language Board, p. 113
23. Koc, V. (2025) Generative AI and Large Language Models in Language Preservation: Opportunities and Challenges, ResearchGate.
24. Sourati, Z., Karimi-Malekabadi, F., & Ozcan, M. (2025) The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models, ResearchGate.
25. Melero, M. (2024) The Future of Language (and Cultural) Diversity in the Age of AI, CLARIN.
26. Perlin, R. (op cit)
27. Russo Carroll, S., Rodriguez Lonebear, D., & Martinez, A. (2017). Data Governance for Native Nation Rebuilding, Native Nations Institute.
28. Te Mana Raraunga. (2018). Frequently Asked Questions, Te Mana Raraunga.
29. Ackermann, A. (op cit)
30. Walter, M., Kukutai, T., Carroll, S. R., & Rodriguez-Lonebear, D. (Eds.). (2020). Indigenous Data Sovereignty and Policy. Taylor & Francis, p. 24
31. Mallory, F. (op cit)
32. Cymdeithas yr Iaith (2022) Cymru Rydd, Cymru Werdd, Cymru Gymraeg., p. 27
33. O’Sullivan, L. (2025). How AI’s Failure on Linguistic Diversity is Deepening Global Inequality, RESET – Digital for Good.
34. Harvey, F. (2024). Global water crisis leaves half of world food production at risk in next 25 years, The Guardian.
35. Erdocia, I., Migge, B., & Schneider, B. (2024). Language is not a data set—Why overcoming ideologies of dataism is more important than ever in the age of AI. Journal of Sociolinguistics, 28(5), p. 23
36. O’Sullivan, L. (op cit)
37. Erdenesanaa, D. (2023) A.I. Could Soon Need as Much Electricity as an Entire Country, The New York Times
38. Heikkilä, M. (2022) We’re getting a better idea of AI’s true carbon footprint, MIT Technology Review.
39. Judah, J. (op cit)#4 #AI #ArtificialIntelligence #Breus #Cornish #Cornwall #data #generativeAI #history #jynn #kedhlow #Kernewek #Kernow #Kernowek #LLM #machine #PYB #SK #SKDinythus #SkiansKreftus #Sordya
-
When GitHub released Copilot, its AI programming assistant, skeptics dismissed it as just another autocomplete tool. Two years later, a Harvard Business School study reveals something far more profound: AI isn't just changing how developers write code – it's fundamentally reshaping how software teams work.
The study, which tracked 187,489 developers over two years, offers the first comprehensive look at how AI transforms knowledge work. The findings are striking: developers with Copilot write more code, work more independently, and experiment more freely with new technologies. But these changes come with unexpected tradeoffs that challenge traditional software development practices.
The Numbers Tell a Story
When developers gain access to Copilot, their work patterns shift dramatically:
- Coding activities increase by 12.37%
- Project management tasks decrease by 24.93%
- Collaboration with other developers drops by 79.3%
- Experimentation with new programming languages rises by 21.79%
"These aren't just productivity gains," explains Dr. Frank Nagle, the study's lead author. "We're seeing a fundamental shift in how developers approach their work. They're coding more, managing less, and working more independently than ever before."
The Collaboration Paradox
Perhaps the most surprising finding is how AI affects team dynamics. Traditional software development emphasizes collaboration – code reviews, pair programming, and constant communication. But Copilot users increasingly work alone, solving problems independently rather than reaching out to teammates.
This shift brings both benefits and risks. While development speed increases, teams report concerns about knowledge sharing and code quality. "We're moving faster, but we had to deliberately create new ways to keep everyone aligned," says Emma Rodriguez, an Engineering Manager at Atlassian. "The natural collaboration that happened through code reviews isn't happening organically anymore."
The Economic Impact
The financial implications are substantial. By analyzing the programming languages developers learn with Copilot's assistance, researchers estimate an average potential salary increase of $1,683 per developer annually. Across all current Copilot users, this represents nearly half a billion dollars in economic impact.
But the real value may be in democratizing expertise. Less experienced developers show the largest gains in productivity and autonomy, suggesting AI could help flatten traditional engineering hierarchies.
AI and the Linchpin Problem: Can AI Help Save Open Source Software?
Open source software is the invisible engine powering much of the digital world, from the servers running websites to the apps on our phones. And yet, the people who build and maintain this crucial infrastructure are often overworked and underappreciated, leading to burnout and even the abandonment of critical projects. This is the linchpin problem in a nutshell: a small group of highly skilled individuals shoulder a disproportionate burden, putting the entire ecosystem at risk. But could AI be the answer?
Less Management, More Coding
Think about the typical open source software maintainer. They are often bombarded with requests to fix bugs, add features, and answer questions from users. These tasks, while essential, can be incredibly time-consuming and take them away from the actual coding they love. The researchers found that when developers were given free access to Copilot, they shifted their work patterns dramatically. Coding activity increased, while project management activities decreased. This suggests that Copilot is not just making developers more productive, but also freeing them from the managerial burden, allowing them to focus on what they do best.
Going Solo with AI
Another interesting finding was that Copilot led developers to work more autonomously and less collaboratively. This might seem counterintuitive, as open source software is all about collaboration, right? But think about it: if AI can help you solve problems and write code faster, you might not need to rely on others as much. This could reduce the friction and overhead often associated with collaboration, especially in the decentralised world of open source.
AI Encourages Experimentation
The study also provided evidence that AI could encourage developers to explore new technologies and programming languages. This is particularly interesting because it suggests that AI could help boost innovation in the open source world. Developers may feel more comfortable trying new things if they have an AI assistant to guide them.
Leveling the Playing Field
Perhaps the most encouraging finding was that the benefits of Copilot were strongest for developers with lower levels of experience and skill. This suggests that AI could help level the playing field in open source development, making it easier for newcomers to contribute and reducing the reliance on a small group of experts.
A Future for Open Source?
While this is just one study, the findings are certainly intriguing. It seems that AI has the potential to change the nature of open source software development in profound ways, addressing the linchpin problem and potentially creating a more sustainable and inclusive ecosystem.
Reshaping the Future of Work
The study's implications extend far beyond software development. As similar AI assistants emerge for other knowledge work domains – from legal research to financial analysis – we might expect similar patterns:
- Increased individual productivity
- Reduced administrative overhead
- More autonomous work patterns
- Greater emphasis on exploration and creativity
The key insight isn't that AI makes workers more productive (though it does), but that it fundamentally changes behavior in ways that ripple through entire organizations.
Looking Ahead
Organizations adopting AI assistants like Copilot need to actively address these changes. Some key considerations:
- Create structured opportunities for knowledge sharing
- Develop new quality control mechanisms that work with autonomous developers
- Invest in tools and practices that preserve institutional knowledge
- Build cultures that balance AI efficiency with human collaboration
The future of software development – and knowledge work more broadly – won't be about humans versus AI, but about mastering this new dance between human creativity and AI capability. The developers who thrive will be those who learn to leverage AI's strengths while preserving the human elements that make great software possible.
As one senior developer put it: "Copilot isn't replacing me – it's letting me focus on the parts of programming I actually enjoy. The trick is learning when to let the AI drive and when to take the wheel yourself."
However, the key insight isn't that AI makes developers more productive (though it does), but that it changes their behavior in ways that ripple through entire organizations. As similar AI assistants emerge for other knowledge work domains – from legal research to financial analysis – we might expect to see similar patterns of increased autonomy, more experimentation, and shifting collaboration models.
Unlock the Future of Business with AI
Dive into our immersive workshops and equip your team with the tools and knowledge to lead in the AI era.
Get in touch with us (FEHLENDE ÜBERSETZUNG) -
When GitHub released Copilot, its AI programming assistant, skeptics dismissed it as just another autocomplete tool. Two years later, a Harvard Business School study reveals something far more profound: AI isn't just changing how developers write code – it's fundamentally reshaping how software teams work.
The study, which tracked 187,489 developers over two years, offers the first comprehensive look at how AI transforms knowledge work. The findings are striking: developers with Copilot write more code, work more independently, and experiment more freely with new technologies. But these changes come with unexpected tradeoffs that challenge traditional software development practices.
The Numbers Tell a Story
When developers gain access to Copilot, their work patterns shift dramatically:
- Coding activities increase by 12.37%
- Project management tasks decrease by 24.93%
- Collaboration with other developers drops by 79.3%
- Experimentation with new programming languages rises by 21.79%
"These aren't just productivity gains," explains Dr. Frank Nagle, the study's lead author. "We're seeing a fundamental shift in how developers approach their work. They're coding more, managing less, and working more independently than ever before."
The Collaboration Paradox
Perhaps the most surprising finding is how AI affects team dynamics. Traditional software development emphasizes collaboration – code reviews, pair programming, and constant communication. But Copilot users increasingly work alone, solving problems independently rather than reaching out to teammates.
This shift brings both benefits and risks. While development speed increases, teams report concerns about knowledge sharing and code quality. "We're moving faster, but we had to deliberately create new ways to keep everyone aligned," says Emma Rodriguez, an Engineering Manager at Atlassian. "The natural collaboration that happened through code reviews isn't happening organically anymore."
The Economic Impact
The financial implications are substantial. By analyzing the programming languages developers learn with Copilot's assistance, researchers estimate an average potential salary increase of $1,683 per developer annually. Across all current Copilot users, this represents nearly half a billion dollars in economic impact.
But the real value may be in democratizing expertise. Less experienced developers show the largest gains in productivity and autonomy, suggesting AI could help flatten traditional engineering hierarchies.
AI and the Linchpin Problem: Can AI Help Save Open Source Software?
Open source software is the invisible engine powering much of the digital world, from the servers running websites to the apps on our phones. And yet, the people who build and maintain this crucial infrastructure are often overworked and underappreciated, leading to burnout and even the abandonment of critical projects. This is the linchpin problem in a nutshell: a small group of highly skilled individuals shoulder a disproportionate burden, putting the entire ecosystem at risk. But could AI be the answer?
Less Management, More Coding
Think about the typical open source software maintainer. They are often bombarded with requests to fix bugs, add features, and answer questions from users. These tasks, while essential, can be incredibly time-consuming and take them away from the actual coding they love. The researchers found that when developers were given free access to Copilot, they shifted their work patterns dramatically. Coding activity increased, while project management activities decreased. This suggests that Copilot is not just making developers more productive, but also freeing them from the managerial burden, allowing them to focus on what they do best.
Going Solo with AI
Another interesting finding was that Copilot led developers to work more autonomously and less collaboratively. This might seem counterintuitive, as open source software is all about collaboration, right? But think about it: if AI can help you solve problems and write code faster, you might not need to rely on others as much. This could reduce the friction and overhead often associated with collaboration, especially in the decentralised world of open source.
AI Encourages Experimentation
The study also provided evidence that AI could encourage developers to explore new technologies and programming languages. This is particularly interesting because it suggests that AI could help boost innovation in the open source world. Developers may feel more comfortable trying new things if they have an AI assistant to guide them.
Leveling the Playing Field
Perhaps the most encouraging finding was that the benefits of Copilot were strongest for developers with lower levels of experience and skill. This suggests that AI could help level the playing field in open source development, making it easier for newcomers to contribute and reducing the reliance on a small group of experts.
A Future for Open Source?
While this is just one study, the findings are certainly intriguing. It seems that AI has the potential to change the nature of open source software development in profound ways, addressing the linchpin problem and potentially creating a more sustainable and inclusive ecosystem.
Reshaping the Future of Work
The study's implications extend far beyond software development. As similar AI assistants emerge for other knowledge work domains – from legal research to financial analysis – we might expect similar patterns:
- Increased individual productivity
- Reduced administrative overhead
- More autonomous work patterns
- Greater emphasis on exploration and creativity
The key insight isn't that AI makes workers more productive (though it does), but that it fundamentally changes behavior in ways that ripple through entire organizations.
Looking Ahead
Organizations adopting AI assistants like Copilot need to actively address these changes. Some key considerations:
- Create structured opportunities for knowledge sharing
- Develop new quality control mechanisms that work with autonomous developers
- Invest in tools and practices that preserve institutional knowledge
- Build cultures that balance AI efficiency with human collaboration
The future of software development – and knowledge work more broadly – won't be about humans versus AI, but about mastering this new dance between human creativity and AI capability. The developers who thrive will be those who learn to leverage AI's strengths while preserving the human elements that make great software possible.
As one senior developer put it: "Copilot isn't replacing me – it's letting me focus on the parts of programming I actually enjoy. The trick is learning when to let the AI drive and when to take the wheel yourself."
However, the key insight isn't that AI makes developers more productive (though it does), but that it changes their behavior in ways that ripple through entire organizations. As similar AI assistants emerge for other knowledge work domains – from legal research to financial analysis – we might expect to see similar patterns of increased autonomy, more experimentation, and shifting collaboration models.
Unlock the Future of Business with AI
Dive into our immersive workshops and equip your team with the tools and knowledge to lead in the AI era.
Get in touch with us -
Richard Mayer’s research on multimedia for learning actually proves text works better
Educational technology professionals cite Richard Mayer’s 2008 study more than any other research on multimedia instruction.
They are citing the wrong conclusion.
Mayer did not prove multimedia enhances learning.
He proved multimedia creates cognitive problems requiring ten different workarounds – and accidentally built the case for text-based instruction.
What Richard Mayer actually found
Through hundreds of controlled experiments, Richard Mayer identified ten principles for multimedia design.
The pattern is striking: most principles involve removing elements from presentations.
Five principles focus on reducing “extraneous processing” – cognitive waste that multimedia creates.
- Remove irrelevant material.
- Highlight essential information buried among distractions.
- Eliminate simultaneous animation, narration, and text because learners perform better with only two elements.
- Place corresponding words and pictures close together.
- Present them simultaneously, not sequentially.
Three principles manage “essential processing” when content is complex.
- Break presentations into learner-controlled segments.
- Use spoken rather than printed text with graphics.
- Provide pre-training before complex multimedia instruction.
Two principles foster deeper learning.
- Combine words and pictures rather than words alone.
- Use conversational rather than formal language.
The hidden message: multimedia instruction is so cognitively demanding that it requires ten specialized principles to avoid harming learning.
Richard Mayer’s split attention revelation
Mayer’s modality principle seems to endorse multimedia: learners perform better with graphics plus spoken text than graphics plus printed text.
Educational technologists celebrate this as proof that multimedia works.
They miss the real insight.
Graphics with printed text create split attention – learners cannot simultaneously look at pictures while reading words.
They must constantly switch between visual elements, wasting cognitive resources on coordination rather than learning.
Richard Mayer’s solution uses different channels: visual graphics with auditory narration.
But this still requires complex mental coordination between multiple input streams while maintaining focus on learning objectives.
Text-based instruction eliminates split attention entirely.
(There are deeply-rooted cultural and historical reasons for the distrust of text.)
Learners process information through one coherent channel that naturally supports sequential, analytical thinking.
The damage control principles in Richard Mayer’s principles
Step back from individual findings and Mayer’s principles reveal themselves as damage control.
The coherence principle removes distractions that multimedia introduces.
The redundancy principle eliminates conflicts between competing inputs.
The segmenting principle provides control that multimedia complexity demands.
The pre-training principle prepares learners for cognitive challenges that simpler instruction avoids.
Each principle represents additional design constraints requiring specialized expertise and extensive testing.
They exist because multimedia instruction is fundamentally problematic.
Text extends Richard Mayer’s logic
At The Geneva Learning Foundation, we work with 70,000 health practitioners using text-based peer learning.
Nigerian practitioners write about extreme heat forcing people to sleep outdoors, increasing malaria exposure.
Colleagues in Brazil, Chad, Ghana, and India read these accounts, analyze climate-health connections, and provide structured feedback through expert-designed rubrics.
No graphics.
No audio coordination.
No split attention problems.
Read our article: Against chocolate-covered broccoli: text-based alternatives to expensive multimedia content
Direct engagement with content that supports rather than complicates learning.
This approach achieves Richard Mayer’s goals through elimination rather than optimization.
Ultimate coherence by presenting only essential information.
Zero redundancy through single-channel processing.
Natural segmenting through text’s inherent reader control.
No pre-training needed because text presents information in logical, sequential structures.
The multimedia principle reconsidered
Mayer’s most famous finding – people learn better from words and pictures than words alone – deserves scrutiny.
This emerged from comparing passive multimedia consumption to passive text reading.
It equates learning with recall.
Neither condition included structured peer interaction, collaborative analysis, or iterative revision that characterize more complex learning.
When learners create knowledge through text-based peer learning, they achieve outcomes that passive consumption of any media cannot match.
The effect size for active text-based learning exceeds Mayer’s multimedia findings while avoiding cognitive coordination problems.
The economic evidence
Mayer’s ten principles exist because multimedia design is expensive and complex.
Each principle represents additional constraints demanding specialized expertise.
Typical multimedia modules are expensive.
Text-based peer learning costs a fraction of this amount while producing superior outcomes.
Resources should flow toward learning infrastructure such as expert rubrics and facilitated dialogue – elements that actually drive learning rather than manage cognitive problems.
The real choice
Educational technology leaders face a fundamental decision: invest in managing multimedia’s problems or adopt approaches that avoid those problems entirely.
Mayer’s research illuminates multimedia’s cognitive costs.
His ten principles represent sophisticated damage control, not learning enhancement.
They minimize harm rather than maximize potential.
Text-based instruction honors Mayer’s deeper insights while rejecting surface implications.
It achieves the cognitive efficiency his principles attempt to restore to multimedia environments.
References
- Berrocal, Y., Regan, J., Fisher, J., Darr, A., Hammersmith, L., Aiyer, M., 2021. Implementing Rubric-Based Peer Review for Video Microlecture Design in Health Professions Education. Med.Sci.Educ. 31, 1761–1765. https://doi.org/10.1007/s40670-021-01437-1
- Clark, R.C., Mayer, R.E. (Eds.), 2016. e‐Learning and the Science of Instruction: Proven Guidelines for Consumers and Designers of Multimedia Learning, 1st ed. Wiley. https://doi.org/10.1002/9781119239086
- Feenberg, A. The written world: On the theory and practice of computer conferencing. Mindweave: Communication, computers, and distance education 22–39 (1989).
- Mayer, R.E., 2008. Applying the science of learning: Evidence-based principles for the design of multimedia instruction. American Psychologist 63, 760–769. https://doi.org/10.1037/0003-066X.63.8.760
- Mayer, R.E., 2005. Cognitive Theory of Multimedia Learning, in: Mayer, R. (Ed.), The Cambridge Handbook of Multimedia Learning. Cambridge University Press, pp. 31–48. https://doi.org/10.1017/CBO9780511816819.004
- Mayer, R.E., Heiser, J., Lonn, S., 2001. Cognitive constraints on multimedia learning: When presenting more material results in less understanding. Journal of Educational Psychology 93, 187–198. https://doi.org/10.1037/0022-0663.93.1.187
- Mayer, R.E., Moreno, R., 2003. Nine Ways to Reduce Cognitive Load in Multimedia Learning. Educational Psychologist 38, 43–52. https://doi.org/10.1207/S15326985EP3801_6
- Mayer, R.E., Moreno, R., 2002. Animation as an Aid to Multimedia Learning. Educational Psychology Review 14, 87–99. https://doi.org/10.1023/A:1013184611077
- Plass, J.L., Chun, D.M., Mayer, R.E., Leutner, D., 2003. Cognitive load in reading a foreign language text with multimedia aids and the influence of verbal and spatial abilities. Computers in Human Behavior 19, 221–243. https://doi.org/10.1016/S0747-5632(02)00015-8
- Sweller, J., 2005. Implications of Cognitive Load Theory for Multimedia Learning, in: Mayer, R. (Ed.), The Cambridge Handbook of Multimedia Learning. Cambridge University Press, pp. 19–30. https://doi.org/10.1017/CBO9780511816819.003
Image: The Geneva Learning Foundation Collection © 2025
#cognitiveLoad #CognitiveLoadTheory #eLearning #instruction #learning #multimedia #multimediaLearning #RichardMayer #text -
Richard Mayer’s research on multimedia for learning actually proves text works better
Educational technology professionals cite Richard Mayer’s 2008 study more than any other research on multimedia instruction.
They are citing the wrong conclusion.
Mayer did not prove multimedia enhances learning.
He proved multimedia creates cognitive problems requiring ten different workarounds – and accidentally built the case for text-based instruction.
What Richard Mayer actually found
Through hundreds of controlled experiments, Richard Mayer identified ten principles for multimedia design.
The pattern is striking: most principles involve removing elements from presentations.
Five principles focus on reducing “extraneous processing” – cognitive waste that multimedia creates.
- Remove irrelevant material.
- Highlight essential information buried among distractions.
- Eliminate simultaneous animation, narration, and text because learners perform better with only two elements.
- Place corresponding words and pictures close together.
- Present them simultaneously, not sequentially.
Three principles manage “essential processing” when content is complex.
- Break presentations into learner-controlled segments.
- Use spoken rather than printed text with graphics.
- Provide pre-training before complex multimedia instruction.
Two principles foster deeper learning.
- Combine words and pictures rather than words alone.
- Use conversational rather than formal language.
The hidden message: multimedia instruction is so cognitively demanding that it requires ten specialized principles to avoid harming learning.
Richard Mayer’s split attention revelation
Mayer’s modality principle seems to endorse multimedia: learners perform better with graphics plus spoken text than graphics plus printed text.
Educational technologists celebrate this as proof that multimedia works.
They miss the real insight.
Graphics with printed text create split attention – learners cannot simultaneously look at pictures while reading words.
They must constantly switch between visual elements, wasting cognitive resources on coordination rather than learning.
Richard Mayer’s solution uses different channels: visual graphics with auditory narration.
But this still requires complex mental coordination between multiple input streams while maintaining focus on learning objectives.
Text-based instruction eliminates split attention entirely.
(There are deeply-rooted cultural and historical reasons for the distrust of text.)
Learners process information through one coherent channel that naturally supports sequential, analytical thinking.
The damage control principles in Richard Mayer’s principles
Step back from individual findings and Mayer’s principles reveal themselves as damage control.
The coherence principle removes distractions that multimedia introduces.
The redundancy principle eliminates conflicts between competing inputs.
The segmenting principle provides control that multimedia complexity demands.
The pre-training principle prepares learners for cognitive challenges that simpler instruction avoids.
Each principle represents additional design constraints requiring specialized expertise and extensive testing.
They exist because multimedia instruction is fundamentally problematic.
Text extends Richard Mayer’s logic
At The Geneva Learning Foundation, we work with 70,000 health practitioners using text-based peer learning.
Nigerian practitioners write about extreme heat forcing people to sleep outdoors, increasing malaria exposure.
Colleagues in Brazil, Chad, Ghana, and India read these accounts, analyze climate-health connections, and provide structured feedback through expert-designed rubrics.
No graphics.
No audio coordination.
No split attention problems.
Read our article: Against chocolate-covered broccoli: text-based alternatives to expensive multimedia content
Direct engagement with content that supports rather than complicates learning.
This approach achieves Richard Mayer’s goals through elimination rather than optimization.
Ultimate coherence by presenting only essential information.
Zero redundancy through single-channel processing.
Natural segmenting through text’s inherent reader control.
No pre-training needed because text presents information in logical, sequential structures.
The multimedia principle reconsidered
Mayer’s most famous finding – people learn better from words and pictures than words alone – deserves scrutiny.
This emerged from comparing passive multimedia consumption to passive text reading.
It equates learning with recall.
Neither condition included structured peer interaction, collaborative analysis, or iterative revision that characterize more complex learning.
When learners create knowledge through text-based peer learning, they achieve outcomes that passive consumption of any media cannot match.
The effect size for active text-based learning exceeds Mayer’s multimedia findings while avoiding cognitive coordination problems.
The economic evidence
Mayer’s ten principles exist because multimedia design is expensive and complex.
Each principle represents additional constraints demanding specialized expertise.
Typical multimedia modules are expensive.
Text-based peer learning costs a fraction of this amount while producing superior outcomes.
Resources should flow toward learning infrastructure such as expert rubrics and facilitated dialogue – elements that actually drive learning rather than manage cognitive problems.
The real choice
Educational technology leaders face a fundamental decision: invest in managing multimedia’s problems or adopt approaches that avoid those problems entirely.
Mayer’s research illuminates multimedia’s cognitive costs.
His ten principles represent sophisticated damage control, not learning enhancement.
They minimize harm rather than maximize potential.
Text-based instruction honors Mayer’s deeper insights while rejecting surface implications.
It achieves the cognitive efficiency his principles attempt to restore to multimedia environments.
References
- Berrocal, Y., Regan, J., Fisher, J., Darr, A., Hammersmith, L., Aiyer, M., 2021. Implementing Rubric-Based Peer Review for Video Microlecture Design in Health Professions Education. Med.Sci.Educ. 31, 1761–1765. https://doi.org/10.1007/s40670-021-01437-1
- Clark, R.C., Mayer, R.E. (Eds.), 2016. e‐Learning and the Science of Instruction: Proven Guidelines for Consumers and Designers of Multimedia Learning, 1st ed. Wiley. https://doi.org/10.1002/9781119239086
- Feenberg, A. The written world: On the theory and practice of computer conferencing. Mindweave: Communication, computers, and distance education 22–39 (1989).
- Mayer, R.E., 2008. Applying the science of learning: Evidence-based principles for the design of multimedia instruction. American Psychologist 63, 760–769. https://doi.org/10.1037/0003-066X.63.8.760
- Mayer, R.E., 2005. Cognitive Theory of Multimedia Learning, in: Mayer, R. (Ed.), The Cambridge Handbook of Multimedia Learning. Cambridge University Press, pp. 31–48. https://doi.org/10.1017/CBO9780511816819.004
- Mayer, R.E., Heiser, J., Lonn, S., 2001. Cognitive constraints on multimedia learning: When presenting more material results in less understanding. Journal of Educational Psychology 93, 187–198. https://doi.org/10.1037/0022-0663.93.1.187
- Mayer, R.E., Moreno, R., 2003. Nine Ways to Reduce Cognitive Load in Multimedia Learning. Educational Psychologist 38, 43–52. https://doi.org/10.1207/S15326985EP3801_6
- Mayer, R.E., Moreno, R., 2002. Animation as an Aid to Multimedia Learning. Educational Psychology Review 14, 87–99. https://doi.org/10.1023/A:1013184611077
- Plass, J.L., Chun, D.M., Mayer, R.E., Leutner, D., 2003. Cognitive load in reading a foreign language text with multimedia aids and the influence of verbal and spatial abilities. Computers in Human Behavior 19, 221–243. https://doi.org/10.1016/S0747-5632(02)00015-8
- Sweller, J., 2005. Implications of Cognitive Load Theory for Multimedia Learning, in: Mayer, R. (Ed.), The Cambridge Handbook of Multimedia Learning. Cambridge University Press, pp. 19–30. https://doi.org/10.1017/CBO9780511816819.003
Image: The Geneva Learning Foundation Collection © 2025
#cognitiveLoad #CognitiveLoadTheory #eLearning #instruction #learning #multimedia #multimediaLearning #RichardMayer #text -
Critical evidence gaps in the Lancet Countdown on health and climate change
The 2024 report of the Lancet Countdown on health and climate change “reveals the health threats of climate change have reached record-breaking levels” and provides “the most up-to-date assessment of the links between health and climate change”.
Yet its treatment of experiential knowledge – particularly the direct observations and understanding developed by frontline health workers and communities – reveals both progress and persistent gaps in how major global health assessments value different forms of knowing.
The fundamental tension appears right at the start.
The report notes a significant challenge: “A global scarcity of internationally standardised data hinders the capacity to optimally monitor the observed health impacts of climate change and evaluate the health-protective effect of implemented interventions.”
This framing privileges standardized, quantifiable data over other forms of knowledge.
Yet paradoxically, the report recognizes that “health workers are already intimate witnesses to the impacts of climate change on the health of the communities they serve, possessing valuable knowledge that should inform both science and policy.”
This recognition of frontline experience as a valid source of knowledge is significant, even if not fully integrated into the report’s methodology.
Health workers’ experiences are not merely anecdotal but represent a crucial form of evidence gathering and early warning that conventional research methods cannot match.
When a nurse in Bangladesh notices changing patterns of heat-related illness in specific neighborhoods, or when a community health worker in Kenya observes shifts in disease transmission seasons, they are detecting signals that might take epidemiological studies decades to formally document.
Can we afford to wait?
As the report acknowledges that we face “record-breaking threats to their wellbeing, health, and survival from the rapidly changing climate,” why wait for traditional longitudinal studies to validate what health workers are already seeing?
Explore the value of health workers’ experiential knowledge: Jones, I., Mbuh, C., Sadki, R., Eller, K., Rhoda, D., 2023. On the frontline of climate change and health: A health worker eyewitness report. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.10204660
Their observations, if their significance and value were fully recognized, could provide vital early insights into emerging health threats and guide rapid, life-saving adaptations.
This is especially critical given the report’s call to alarm that climate change impacts are “increasingly claiming lives and livelihoods worldwide” and that “delays in climate change mitigation and adaptation have intensified these impacts.”
The humanitarian imperative to act quickly makes health workers’ experiential knowledge not just valuable but essential – they are the canaries in the coal mine of our climate crisis, and their insights could help bridge critical evidence gaps while more traditional research catches up.
The report’s most thoughtful engagement with alternative forms of knowledge comes in its treatment of Indigenous knowledge systems.
A panel titled “Indigenous knowledge for a healthy future” explicitly acknowledges that “Indigenous peoples maintain deep connections with the natural environment that are important for the social, livelihood, cultural, and spiritual practices that underpin their health and wellbeing.”
More importantly, it recognizes that “Indigenous knowledge has been shown to be the key to protect Indigenous health in times of health emergencies when official health systems and governments are unable to provide assistance to Indigenous communities.”
However, the report also acknowledges that “Indigenous medicine and worldviews are rarely considered within health care or health risk preparedness and response.”
This gap between recognizing the value of Indigenous knowledge and actually incorporating it into health systems and policies reflects a broader challenge.
A crucial observation comes in the report’s data discussion: available data are “rarely disaggregated by relevant groups (eg, gender, age, indigeneity, ethnicity, and socioeconomic level)” and “Indigenous knowledge is often overlooked, and Indigenous populations are seldom taken into consideration in the production and reporting of evidence and data.”
This gap in representation means that crucial experiential knowledge is systematically excluded from our understanding of climate change’s health impacts.
Perhaps most tellingly, while the report calls for “improved data” to evaluate progress on international commitments, it focuses primarily on standardized quantitative metrics rather than developing new frameworks that could better integrate experiential knowledge.
This reveals an underlying epistemological bias – while experiential knowledge is acknowledged as valuable, the report’s methodology remains firmly grounded in traditional scientific approaches.
Looking forward, truly leveraging experiential knowledge in understanding climate change’s health impacts will require more than just acknowledgment.
It will require developing new methodological frameworks that can systematically incorporate and validate different forms of knowing, while ensuring that frontline voices – whether from health workers, Indigenous communities, or other groups with direct experience – are centered rather than marginalized in our understanding of this global crisis.
For the Lancet Countdown to fully live up to its mission of tracking progress on health and climate change, future reports will need to more fundamentally rethink how they recognize, validate, and incorporate experiential knowledge.
The seeds of this transformation are present in the 2024 report.
Doing so is both necessary to improve science and consistent with The Lancet Countdown’s commitment to “operate an open and iterative process of indicator improvement, welcoming proposals for new indicators… from the world’s most vulnerable countries”.
References
- Romanello, M., et al., 2024. The 2024 report of the Lancet Countdown on health and climate change: facing record-breaking threats from delayed action. The Lancet 404, 1847–1896. https://doi.org/10.1016/S0140-6736(24)01822-1
- Jones, I., Mbuh, C., Sadki, R., Eller, K., Rhoda, D., 2023. On the frontline of climate change and health: A health worker eyewitness report. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.10204660
- Jones, I., Mbuh, C., Sadki, R., Steed, I., 2024. Climate change and health: Health workers on climate, community, and the urgent need for action. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.11194918
- Sadki, R., 2025. WHO Global Conference on Climate and Health: New pathways to overcome structural barriers blocking effective climate and health action. https://doi.org/10.59350/redasadki.21322
- Sadki, R., 2025. Climate change and health: a new peer learning programme by and for health workers from the most climate-vulnerable countries. https://doi.org/10.59350/redasadki.21339
- Sadki, R., 2024. Knowing-in-action: Bridging the theory-practice divide in global health. https://doi.org/10.59350/4evj5-vm802
- Sadki, R., 2024. Strengthening primary health care in a changing climate. https://doi.org/10.59350/5s2zf-s6879
- Sadki, R., 2024. World Health Summit: to rebuild trust in global health, invest in health workers as community leaders. https://doi.org/10.59350/343na-80712
- Sadki, R., 2024. The cost of inaction: Quantifying the impact of climate change on health. https://doi.org/10.59350/gn95w-jpt34
- Sanchez, J.J. et al. (2025) ‘The climate crisis and human health: identifying grand challenges through participatory research’, The Lancet Global Health, p. S2214109X25000038. Available at: https://doi.org/10.1016/S2214-109X(25)00003-8.
Image: The Geneva Learning Foundation Collection © 2024
#climateAndHealth #COP29 #CriticalEvidenceGapsInTheLancetCountdownOnHealthAndClimateChange #epistemology #experientialKnowledge #IndigenousKnowledge #localKnowledge #quantitativeData
-
Critical evidence gaps in the Lancet Countdown on health and climate change
The 2024 report of the Lancet Countdown on health and climate change “reveals the health threats of climate change have reached record-breaking levels” and provides “the most up-to-date assessment of the links between health and climate change”.
Yet its treatment of experiential knowledge – particularly the direct observations and understanding developed by frontline health workers and communities – reveals both progress and persistent gaps in how major global health assessments value different forms of knowing.
The fundamental tension appears right at the start.
The report notes a significant challenge: “A global scarcity of internationally standardised data hinders the capacity to optimally monitor the observed health impacts of climate change and evaluate the health-protective effect of implemented interventions.”
This framing privileges standardized, quantifiable data over other forms of knowledge.
Yet paradoxically, the report recognizes that “health workers are already intimate witnesses to the impacts of climate change on the health of the communities they serve, possessing valuable knowledge that should inform both science and policy.”
This recognition of frontline experience as a valid source of knowledge is significant, even if not fully integrated into the report’s methodology.
Health workers’ experiences are not merely anecdotal but represent a crucial form of evidence gathering and early warning that conventional research methods cannot match.
When a nurse in Bangladesh notices changing patterns of heat-related illness in specific neighborhoods, or when a community health worker in Kenya observes shifts in disease transmission seasons, they are detecting signals that might take epidemiological studies decades to formally document.
Can we afford to wait?
As the report acknowledges that we face “record-breaking threats to their wellbeing, health, and survival from the rapidly changing climate,” why wait for traditional longitudinal studies to validate what health workers are already seeing?
Explore the value of health workers’ experiential knowledge: Jones, I., Mbuh, C., Sadki, R., Eller, K., Rhoda, D., 2023. On the frontline of climate change and health: A health worker eyewitness report. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.10204660
Their observations, if their significance and value were fully recognized, could provide vital early insights into emerging health threats and guide rapid, life-saving adaptations.
This is especially critical given the report’s call to alarm that climate change impacts are “increasingly claiming lives and livelihoods worldwide” and that “delays in climate change mitigation and adaptation have intensified these impacts.”
The humanitarian imperative to act quickly makes health workers’ experiential knowledge not just valuable but essential – they are the canaries in the coal mine of our climate crisis, and their insights could help bridge critical evidence gaps while more traditional research catches up.
The report’s most thoughtful engagement with alternative forms of knowledge comes in its treatment of Indigenous knowledge systems.
A panel titled “Indigenous knowledge for a healthy future” explicitly acknowledges that “Indigenous peoples maintain deep connections with the natural environment that are important for the social, livelihood, cultural, and spiritual practices that underpin their health and wellbeing.”
More importantly, it recognizes that “Indigenous knowledge has been shown to be the key to protect Indigenous health in times of health emergencies when official health systems and governments are unable to provide assistance to Indigenous communities.”
However, the report also acknowledges that “Indigenous medicine and worldviews are rarely considered within health care or health risk preparedness and response.”
This gap between recognizing the value of Indigenous knowledge and actually incorporating it into health systems and policies reflects a broader challenge.
A crucial observation comes in the report’s data discussion: available data are “rarely disaggregated by relevant groups (eg, gender, age, indigeneity, ethnicity, and socioeconomic level)” and “Indigenous knowledge is often overlooked, and Indigenous populations are seldom taken into consideration in the production and reporting of evidence and data.”
This gap in representation means that crucial experiential knowledge is systematically excluded from our understanding of climate change’s health impacts.
Perhaps most tellingly, while the report calls for “improved data” to evaluate progress on international commitments, it focuses primarily on standardized quantitative metrics rather than developing new frameworks that could better integrate experiential knowledge.
This reveals an underlying epistemological bias – while experiential knowledge is acknowledged as valuable, the report’s methodology remains firmly grounded in traditional scientific approaches.
Looking forward, truly leveraging experiential knowledge in understanding climate change’s health impacts will require more than just acknowledgment.
It will require developing new methodological frameworks that can systematically incorporate and validate different forms of knowing, while ensuring that frontline voices – whether from health workers, Indigenous communities, or other groups with direct experience – are centered rather than marginalized in our understanding of this global crisis.
For the Lancet Countdown to fully live up to its mission of tracking progress on health and climate change, future reports will need to more fundamentally rethink how they recognize, validate, and incorporate experiential knowledge.
The seeds of this transformation are present in the 2024 report.
Doing so is both necessary to improve science and consistent with The Lancet Countdown’s commitment to “operate an open and iterative process of indicator improvement, welcoming proposals for new indicators… from the world’s most vulnerable countries”.
References
- Romanello, M., et al., 2024. The 2024 report of the Lancet Countdown on health and climate change: facing record-breaking threats from delayed action. The Lancet 404, 1847–1896. https://doi.org/10.1016/S0140-6736(24)01822-1
- Jones, I., Mbuh, C., Sadki, R., Eller, K., Rhoda, D., 2023. On the frontline of climate change and health: A health worker eyewitness report. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.10204660
- Jones, I., Mbuh, C., Sadki, R., Steed, I., 2024. Climate change and health: Health workers on climate, community, and the urgent need for action. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.11194918
- Sadki, R., 2025. WHO Global Conference on Climate and Health: New pathways to overcome structural barriers blocking effective climate and health action. https://doi.org/10.59350/redasadki.21322
- Sadki, R., 2025. Climate change and health: a new peer learning programme by and for health workers from the most climate-vulnerable countries. https://doi.org/10.59350/redasadki.21339
- Sadki, R., 2024. Knowing-in-action: Bridging the theory-practice divide in global health. https://doi.org/10.59350/4evj5-vm802
- Sadki, R., 2024. Strengthening primary health care in a changing climate. https://doi.org/10.59350/5s2zf-s6879
- Sadki, R., 2024. World Health Summit: to rebuild trust in global health, invest in health workers as community leaders. https://doi.org/10.59350/343na-80712
- Sadki, R., 2024. The cost of inaction: Quantifying the impact of climate change on health. https://doi.org/10.59350/gn95w-jpt34
- Sanchez, J.J. et al. (2025) ‘The climate crisis and human health: identifying grand challenges through participatory research’, The Lancet Global Health, p. S2214109X25000038. Available at: https://doi.org/10.1016/S2214-109X(25)00003-8.
Image: The Geneva Learning Foundation Collection © 2024
#climateAndHealth #COP29 #CriticalEvidenceGapsInTheLancetCountdownOnHealthAndClimateChange #epistemology #experientialKnowledge #IndigenousKnowledge #localKnowledge #quantitativeData
-
Climate change and health: perspectives from developing countries
Today, the Geneva Learning Foundation’s Charlotte Mbuh delivered a scientific presentation titled “On the frontline of climate change and health: A health worker eyewitness report” at the University of Hamburg’s Online Expert Seminar on Climate Change and Health: Perspectives from Developing Countries.
https://www.youtube.com/watch?v=7cR-mFCj2xk
Mbuh shared insights from a report based on observations from frontline health workers on the impact of climate change on health in their communities.
The Geneva Learning Foundation, a Swiss non-profit, facilitated a special event “From community to planet: Health professionals on the frontlines of climate change” on 28 July 2023, engaging 4,700 health practitioners from 68 countries who shared 1,260 observations.
“93% of respondents believed that there was a link between climate change and health, and they reported a direct local experience of a wide range of climatic and environmental impacts,” Mbuh stated.
The most commonly reported impacts were on farming and farmland, the distribution of disease-carrying insects, and urban areas becoming hotter.
Health impacts linked to these climatic and environmental changes included increased malnutrition and/or undernutrition, increased waterborne diseases, and changes to the incidence and distribution of vector-borne diseases.
Mbuh emphasized that these impacts were particularly prevalent in smaller communities or mid-sized towns.
Mbuh highlighted the unique role of frontline health workers as trusted advisors to their communities: “Frontline health workers are trusted advisors of the communities that they serve, and they have unique insights to local realities and are strategically positioned to bring about change,” she said.
The Geneva Learning Foundation aims to leverage its digitally-enabled peer learning network of health workers to drive change across different levels of the health system and geographical boundaries.
Mbuh concluded : “These experiences demonstrate the importance of community engagement, sustainable practices, and support from relevant stakeholders in addressing the climate health nexus and building resilience in the face of a changing climate.”
The presentation underscored the urgent need to invest in frontline health workers at the local level to build resilience against the impacts of climate change on health, particularly in vulnerable communities in developing countries.
The event was organized by the International Expert Centre of Climate Change and Health (IECCCH) at the Research and Transfer Centre Sustainable Development and Climate Change Management, Hamburg University of Applied Sciences, in collaboration with the European School of Sustainability Science and Research (ESSSR), the UK Consortium on Sustainability Research (UK-CSR), and the Inter-University Sustainable Development Research Programme (IUSDRP).
Photo: The Geneva Learning Foundation Collection © 2024
Share this:
#CharlotteMbuh #climateChange #developingCountries #ExpertCentreOfClimateChangeAndHealth #globalHealth #HamburgUniversityOfAppliedSciences #health
-
Climate change and health: perspectives from developing countries
Today, the Geneva Learning Foundation’s Charlotte Mbuh delivered a scientific presentation titled “On the frontline of climate change and health: A health worker eyewitness report” at the University of Hamburg’s Online Expert Seminar on Climate Change and Health: Perspectives from Developing Countries.
https://www.youtube.com/watch?v=7cR-mFCj2xk
Mbuh shared insights from a report based on observations from frontline health workers on the impact of climate change on health in their communities.
The Geneva Learning Foundation, a Swiss non-profit, facilitated a special event “From community to planet: Health professionals on the frontlines of climate change” on 28 July 2023, engaging 4,700 health practitioners from 68 countries who shared 1,260 observations.
“93% of respondents believed that there was a link between climate change and health, and they reported a direct local experience of a wide range of climatic and environmental impacts,” Mbuh stated.
The most commonly reported impacts were on farming and farmland, the distribution of disease-carrying insects, and urban areas becoming hotter.
Health impacts linked to these climatic and environmental changes included increased malnutrition and/or undernutrition, increased waterborne diseases, and changes to the incidence and distribution of vector-borne diseases.
Mbuh emphasized that these impacts were particularly prevalent in smaller communities or mid-sized towns.
Mbuh highlighted the unique role of frontline health workers as trusted advisors to their communities: “Frontline health workers are trusted advisors of the communities that they serve, and they have unique insights to local realities and are strategically positioned to bring about change,” she said.
The Geneva Learning Foundation aims to leverage its digitally-enabled peer learning network of health workers to drive change across different levels of the health system and geographical boundaries.
Mbuh concluded : “These experiences demonstrate the importance of community engagement, sustainable practices, and support from relevant stakeholders in addressing the climate health nexus and building resilience in the face of a changing climate.”
The presentation underscored the urgent need to invest in frontline health workers at the local level to build resilience against the impacts of climate change on health, particularly in vulnerable communities in developing countries.
The event was organized by the International Expert Centre of Climate Change and Health (IECCCH) at the Research and Transfer Centre Sustainable Development and Climate Change Management, Hamburg University of Applied Sciences, in collaboration with the European School of Sustainability Science and Research (ESSSR), the UK Consortium on Sustainability Research (UK-CSR), and the Inter-University Sustainable Development Research Programme (IUSDRP).
Photo: The Geneva Learning Foundation Collection © 2024
Share this:
#CharlotteMbuh #climateChange #developingCountries #ExpertCentreOfClimateChangeAndHealth #globalHealth #HamburgUniversityOfAppliedSciences #health
-
Development is adaptation: Bill Gates’s shift is actually about linking climate change and health
Bill Gates’ latest public memo marks a significant shift in how the world’s most influential philanthropist frames the challenge of climate change. He sees a future in which responding to climate threats and promoting well-being become two sides of the same mission, declaring, “development is adaptation.”
Gates argues that the principal metric for climate action should not be global temperature or near-term emission reductions alone, but measured improvement in the lives of the world’s most vulnerable populations.
He argues that the focus of climate action should be on the “greatest possible impact for the most vulnerable people.”
The suffering of poor communities must take priority, since, in his view, “climate change, disease, and poverty are all major problems. We should deal with them in proportion to the suffering they cause.”
Climate change is about the health of the most vulnerable
This position resonates with a core message that has emerged across global health over the past several years: climate change is about health.
New data from the 2025 Lancet Countdown draw a stark picture:
- Heat-related mortality has risen 63 percent since the 1990s.
- Deaths from wildfire smoke and air pollution caused by fossil fuels continue to climb.
- Food insecurity, driven by erratic weather, is destabilizing health and economies at once.
- Thirteen out of twenty key health indicators linked to climate impacts now signal urgent action is needed.
Health professionals, policy coalitions, scientists, and patient advocates have succeeded in bringing this nexus between climate and health squarely to the global agenda, culminating in recent summits where health finally shared the main stage with energy and economics.
Yet just as the science and advocacy align, political attention risks fragmenting.
Despite sweeping reports, evidence, and high-level declarations, momentum can ebb.
There is now a risk that the transformative potential embedded in the climate-health linkage may not be fully realized.
Here, Gates’s pivot could actually be the inflection point that the field needs.
The case for health workforce-centered adaptation
For nearly a decade, The Geneva Learning Foundation (TGLF) has been advocating and demonstrating that meeting complex humanitarian, health, and development challenges requires strengthening not just technical capacities or disease programs, but the underlying connective tissue of the health system: its workforce.
TGLF’s digital peer-learning platform now connects over 70,000 health workers across more than 130 countries.
These practitioners – mostly in government service, often in low-resource or crisis-affected settings – are the first to observe, and often the first to respond to, the local impacts of climate change on health.
Their reports show that health impacts are immediate and multi-faceted: rising malnutrition from crop failures, increases in waterborne diseases following floods, new burdens from air pollution and heat, and psychological distress from repeated disasters.
What sets this approach apart is its systemic focus.
Climate change is not a threat that can be “verticalized”.
It demands responses that are adaptive, distributed, and coordinated across all levels of the health system.
TGLF’s innovation lies in harnessing a distributed network to surface and scale locally-grounded solutions:
- Peer learning accelerates uptake of proven local adaptations – health workers share and adapt practices in near real time.
- Digital networks enable early warning and rapid, cost-effective dissemination of context-appropriate strategies, even in remote settings.
- The workforce is not merely the channel for external solutions but is positioned as a community of problem-solvers and trusted communicators, leaders in building and sustaining resilience.
Data from these initiatives demonstrate that such networked learning delivers results at scale, often with return on investment superior to parallel vertical programs, and increases system resilience and flexibility.
Development is adaptation: the need for human capital investment
The urgency and logic of these approaches are reinforced by ongoing policy developments ahead of COP30:
- WHO’s Global Action Plan on Climate Change and Health, adopted at the World Health Assembly in May 2025, recognizes that without context-sensitive system strengthening, existing approaches are insufficient, and positions knowledge and workforce mobilization as strategic imperatives.
- The COP30 Belem Health Action Plan establishes adaptation of the health sector to climate change as an international priority, calling for holistic, cross-sectoral strategies, and “community engagement and participation as foundational to implementation.”
Without empowered and connected health workers, no global action plan will reach those most at risk or maintain public trust.
A strategic investment imperative: why the next breakthrough must be human-centered
The philanthropic search for cost-effective, scalable, and measurable impact has built immense legacies in reducing child mortality and combating infectious disease.
Gates’ own approach of pioneering “vertical” innovations, optimizing delivery through metrics, and prioritizing technical solutions has been transformative, especially at the intersection of science and delivery.
However, emerging science show the limits of technical “magic bullets” absent robust, interconnected local systems.
Trust, legitimacy, and action flow from the relationships health workers build in – and with – their communities.
If development is adaptation, what does this mean for the next phase in climate-health philanthropy?
If the measure of climate action’s value is the scale and speed at which lives are improved and disasters averted, investing in the human infrastructure of the health system is the most evidence-based, cost-effective, and legacy-ensuring play available.
- Investing in the health workforce is itself a breakthrough technology: It increases the absorptive capacity of low-resource health systems, making innovations stick and catalyzing uptake well beyond single-disease silos or narrow infrastructure projects.
- Long-term, system-wide resilience is built by equipping health workers – not simply with technology or training from above, but with platforms for peer learning, rapid response, and locally-driven adaptation coordinated through agile networks.
- The network effect is real: A million motivated and networked health practitioners is likely to surface, refine, and implement interventions at a scale and pace that outstrips most top-down models. Digitally-enabled peer learning, tested by TGLF, could link to AI systems to provide distributed AI-human intelligence that supports effective action.
Without these bridges, even the best technology or policies will fail to gain a durable footprint at community level, especially as climate impacts deepen.
Health is where climate change action matters most
The world is waking to the reality that technical solutions alone cannot future-proof health against climate risks.
We need to focus on the highest-value levers.
This starts with a distributed, networked workforce at the coalface of the crisis, empowered to adapt, share, and lead.
In a world of accelerating climate shocks and retreating political will, the boldest, most rational bet for sustained global impact is to go “horizontal” – to invest in the people and the systems that connect them.
By helping build adaptive, digitally connected networks of health professionals, philanthropy can reinforce the foundation upon which all high-impact innovation rests.
This is not a departure from the pursuit of technology-driven change, but rather the necessary evolution to ensure every breakthrough finds its mark – and that trust in science and public health stays strong under pressure.
If ever there was a time for rigorous, data-driven engagement that bridges technology, health, and community resilience, this is it.
Every indicator – scientific, economic, social – suggests that communities will confront more climate disruptions in the coming years.
Investing in the people who translate science into health, who stand with their communities in crisis, is the most robust, scalable, and sustainable bet that any philanthropist or society can make.
By focusing on these vital human connections, the world can ensure that innovation works where it matters most – and that the next chapter of climate action measures true success by the health, security, and opportunity it delivers for all.
History will honor those whose support creates not only tools and policies, but the living networks of trust and craft upon which community resilience depends.
That is the climate breakthrough waiting to happen.
References
- COP30 Belém Action Plan. (2025). The Belém Health Action Plan for the Adaptation of the Health Sector to Climate Change. https://www.who.int/teams/environment-climate-change-and-health/climate-change-and-health/advocacy-partnerships/talks/health-at-cop30
- Ebi, K.L., et al. (2025). The attribution of human health outcomes to climate change: transdisciplinary practical guidance. Climatic Change, 178, 143. https://doi.org/10.1007/s10584-025-03976-7
- Ebi, K.L., Haines, A. (2019). The imperative for climate action to protect health. The New England Journal of Medicine, 380, 263–273. https://doi.org/10.1056/NEJMra1807873
- Jacobson, J., Brooks, A., Mbuh, C., Sadki, R. (2023). Learning from frontline health workers in the climate change era. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.7316466
- Jones, I., Mbuh, C., Sadki, R., Steed, I. (2024). Climate change and health: Health workers on climate, community, and the urgent need for action (Version 1.0). The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.11194918
- Romanello, M., et al. (2025). The 2025 report of the Lancet Countdown on health and climate change. The Lancet S0140673625019191. https://doi.org/10.1016/S0140-6736(25)01919-1
- Sadki, R. (2024). Health at COP29: Workforce crisis meets climate crisis. The Geneva Learning Foundation. https://doi.org/10.59350/sdmgt-ptt98
- Sadki, R. (2024). Strengthening primary health care in a changing climate. The Geneva Learning Foundation. https://doi.org/10.59350/5s2zf-s6879
- Sadki, R. (2024). The cost of inaction: Quantifying the impact of climate change on health. The Geneva Learning Foundation. https://doi.org/10.59350/gn95w-jpt34
- Sanchez, J.J., et al. (2025). The climate crisis and human health: identifying grand challenges through participatory research. The Lancet Global Health. https://doi.org/10.1016/S2214-109X(25)00003-8
- Storeng, K. T. (2014). The GAVI Alliance and the Gates approach to health system strengthening. Global Public Health, 9(8), 865–879. https://doi.org/10.1080/17441692.2014.940362
- World Health Organization. (2025). Draft Global Action Plan on Climate Change and Health. Seventy-eighth World Health Assembly. https://apps.who.int/gb/ebwha/pdf_files/WHA78/A78_4Add2-en.pdf
#billGates #climateAndHealth #climateChangeAndHealth #development #humanCapitalInvestment #lancetCountdown #peerLearning #theGenevaLearningFoundation #workforce
-
Development is adaptation: Bill Gates’s shift is actually about linking climate change and health
Bill Gates’ latest public memo marks a significant shift in how the world’s most influential philanthropist frames the challenge of climate change. He sees a future in which responding to climate threats and promoting well-being become two sides of the same mission, declaring, “development is adaptation.”
Gates argues that the principal metric for climate action should not be global temperature or near-term emission reductions alone, but measured improvement in the lives of the world’s most vulnerable populations.
He argues that the focus of climate action should be on the “greatest possible impact for the most vulnerable people.”
The suffering of poor communities must take priority, since, in his view, “climate change, disease, and poverty are all major problems. We should deal with them in proportion to the suffering they cause.”
Climate change is about the health of the most vulnerable
This position resonates with a core message that has emerged across global health over the past several years: climate change is about health.
New data from the 2025 Lancet Countdown draw a stark picture:
- Heat-related mortality has risen 63 percent since the 1990s.
- Deaths from wildfire smoke and air pollution caused by fossil fuels continue to climb.
- Food insecurity, driven by erratic weather, is destabilizing health and economies at once.
- Thirteen out of twenty key health indicators linked to climate impacts now signal urgent action is needed.
Health professionals, policy coalitions, scientists, and patient advocates have succeeded in bringing this nexus between climate and health squarely to the global agenda, culminating in recent summits where health finally shared the main stage with energy and economics.
Yet just as the science and advocacy align, political attention risks fragmenting.
Despite sweeping reports, evidence, and high-level declarations, momentum can ebb.
There is now a risk that the transformative potential embedded in the climate-health linkage may not be fully realized.
Here, Gates’s pivot could actually be the inflection point that the field needs.
The case for health workforce-centered adaptation
For nearly a decade, The Geneva Learning Foundation (TGLF) has been advocating and demonstrating that meeting complex humanitarian, health, and development challenges requires strengthening not just technical capacities or disease programs, but the underlying connective tissue of the health system: its workforce.
TGLF’s digital peer-learning platform now connects over 70,000 health workers across more than 130 countries.
These practitioners – mostly in government service, often in low-resource or crisis-affected settings – are the first to observe, and often the first to respond to, the local impacts of climate change on health.
Their reports show that health impacts are immediate and multi-faceted: rising malnutrition from crop failures, increases in waterborne diseases following floods, new burdens from air pollution and heat, and psychological distress from repeated disasters.
What sets this approach apart is its systemic focus.
Climate change is not a threat that can be “verticalized”.
It demands responses that are adaptive, distributed, and coordinated across all levels of the health system.
TGLF’s innovation lies in harnessing a distributed network to surface and scale locally-grounded solutions:
- Peer learning accelerates uptake of proven local adaptations – health workers share and adapt practices in near real time.
- Digital networks enable early warning and rapid, cost-effective dissemination of context-appropriate strategies, even in remote settings.
- The workforce is not merely the channel for external solutions but is positioned as a community of problem-solvers and trusted communicators, leaders in building and sustaining resilience.
Data from these initiatives demonstrate that such networked learning delivers results at scale, often with return on investment superior to parallel vertical programs, and increases system resilience and flexibility.
Development is adaptation: the need for human capital investment
The urgency and logic of these approaches are reinforced by ongoing policy developments ahead of COP30:
- WHO’s Global Action Plan on Climate Change and Health, adopted at the World Health Assembly in May 2025, recognizes that without context-sensitive system strengthening, existing approaches are insufficient, and positions knowledge and workforce mobilization as strategic imperatives.
- The COP30 Belem Health Action Plan establishes adaptation of the health sector to climate change as an international priority, calling for holistic, cross-sectoral strategies, and “community engagement and participation as foundational to implementation.”
Without empowered and connected health workers, no global action plan will reach those most at risk or maintain public trust.
A strategic investment imperative: why the next breakthrough must be human-centered
The philanthropic search for cost-effective, scalable, and measurable impact has built immense legacies in reducing child mortality and combating infectious disease.
Gates’ own approach of pioneering “vertical” innovations, optimizing delivery through metrics, and prioritizing technical solutions has been transformative, especially at the intersection of science and delivery.
However, emerging science show the limits of technical “magic bullets” absent robust, interconnected local systems.
Trust, legitimacy, and action flow from the relationships health workers build in – and with – their communities.
If development is adaptation, what does this mean for the next phase in climate-health philanthropy?
If the measure of climate action’s value is the scale and speed at which lives are improved and disasters averted, investing in the human infrastructure of the health system is the most evidence-based, cost-effective, and legacy-ensuring play available.
- Investing in the health workforce is itself a breakthrough technology: It increases the absorptive capacity of low-resource health systems, making innovations stick and catalyzing uptake well beyond single-disease silos or narrow infrastructure projects.
- Long-term, system-wide resilience is built by equipping health workers – not simply with technology or training from above, but with platforms for peer learning, rapid response, and locally-driven adaptation coordinated through agile networks.
- The network effect is real: A million motivated and networked health practitioners is likely to surface, refine, and implement interventions at a scale and pace that outstrips most top-down models. Digitally-enabled peer learning, tested by TGLF, could link to AI systems to provide distributed AI-human intelligence that supports effective action.
Without these bridges, even the best technology or policies will fail to gain a durable footprint at community level, especially as climate impacts deepen.
Health is where climate change action matters most
The world is waking to the reality that technical solutions alone cannot future-proof health against climate risks.
We need to focus on the highest-value levers.
This starts with a distributed, networked workforce at the coalface of the crisis, empowered to adapt, share, and lead.
In a world of accelerating climate shocks and retreating political will, the boldest, most rational bet for sustained global impact is to go “horizontal” – to invest in the people and the systems that connect them.
By helping build adaptive, digitally connected networks of health professionals, philanthropy can reinforce the foundation upon which all high-impact innovation rests.
This is not a departure from the pursuit of technology-driven change, but rather the necessary evolution to ensure every breakthrough finds its mark – and that trust in science and public health stays strong under pressure.
If ever there was a time for rigorous, data-driven engagement that bridges technology, health, and community resilience, this is it.
Every indicator – scientific, economic, social – suggests that communities will confront more climate disruptions in the coming years.
Investing in the people who translate science into health, who stand with their communities in crisis, is the most robust, scalable, and sustainable bet that any philanthropist or society can make.
By focusing on these vital human connections, the world can ensure that innovation works where it matters most – and that the next chapter of climate action measures true success by the health, security, and opportunity it delivers for all.
History will honor those whose support creates not only tools and policies, but the living networks of trust and craft upon which community resilience depends.
That is the climate breakthrough waiting to happen.
References
- COP30 Belém Action Plan. (2025). The Belém Health Action Plan for the Adaptation of the Health Sector to Climate Change. https://www.who.int/teams/environment-climate-change-and-health/climate-change-and-health/advocacy-partnerships/talks/health-at-cop30
- Ebi, K.L., et al. (2025). The attribution of human health outcomes to climate change: transdisciplinary practical guidance. Climatic Change, 178, 143. https://doi.org/10.1007/s10584-025-03976-7
- Ebi, K.L., Haines, A. (2019). The imperative for climate action to protect health. The New England Journal of Medicine, 380, 263–273. https://doi.org/10.1056/NEJMra1807873
- Jacobson, J., Brooks, A., Mbuh, C., Sadki, R. (2023). Learning from frontline health workers in the climate change era. The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.7316466
- Jones, I., Mbuh, C., Sadki, R., Steed, I. (2024). Climate change and health: Health workers on climate, community, and the urgent need for action (Version 1.0). The Geneva Learning Foundation. https://doi.org/10.5281/zenodo.11194918
- Romanello, M., et al. (2025). The 2025 report of the Lancet Countdown on health and climate change. The Lancet S0140673625019191. https://doi.org/10.1016/S0140-6736(25)01919-1
- Sadki, R. (2024). Health at COP29: Workforce crisis meets climate crisis. The Geneva Learning Foundation. https://doi.org/10.59350/sdmgt-ptt98
- Sadki, R. (2024). Strengthening primary health care in a changing climate. The Geneva Learning Foundation. https://doi.org/10.59350/5s2zf-s6879
- Sadki, R. (2024). The cost of inaction: Quantifying the impact of climate change on health. The Geneva Learning Foundation. https://doi.org/10.59350/gn95w-jpt34
- Sanchez, J.J., et al. (2025). The climate crisis and human health: identifying grand challenges through participatory research. The Lancet Global Health. https://doi.org/10.1016/S2214-109X(25)00003-8
- Storeng, K. T. (2014). The GAVI Alliance and the Gates approach to health system strengthening. Global Public Health, 9(8), 865–879. https://doi.org/10.1080/17441692.2014.940362
- World Health Organization. (2025). Draft Global Action Plan on Climate Change and Health. Seventy-eighth World Health Assembly. https://apps.who.int/gb/ebwha/pdf_files/WHA78/A78_4Add2-en.pdf
#billGates #climateAndHealth #climateChangeAndHealth #development #humanCapitalInvestment #lancetCountdown #peerLearning #theGenevaLearningFoundation #workforce
-
Integrating community-based monitoring (CBM) into a comprehensive learning-to-action model
According to Gavi, “community-based monitoring” or “CBM” is a process where service users collect data on various aspects of health service provision to monitor program implementation, identify gaps, and collaboratively develop solutions with providers.
- Community-based monitoring (CBM) has emerged as a promising strategy for enhancing immunization program performance and equity.
- CBM interventions have been implemented across different settings and populations, including remote rural areas, urban poor, fragile/conflict-affected regions, and marginalized groups such as indigenous populations and people living with HIV.
By engaging service users, CBM aims to foster greater accountability and responsiveness to local needs.
- However, realizing CBM’s potential in practice has proven challenging.
- Without a coherent approach, CBM risks becoming just another disconnected tool.
The Geneva Learning Foundation’s innovative learning-to-action model offers a compelling framework within which CBM could be applied to immunization challenges.
The model’s comprehensive design creates an enabling environment for effectively integrating diverse monitoring data sources – and this could include community perspectives.
Health workers as trusted community advisers… and members of the community
A distinctive feature of TGLF’s model is its emphasis on health workers’ role as trusted advisors to the communities they serve.
The model recognizes that local health staff are not merely service providers, but often deeply embedded community members with intimate knowledge of local realities.
For example, in TGLF’s immunization learning initiatives, participating health workers frequently share insights into the social, cultural, and economic factors shaping vaccine hesitancy and uptake in their communities.
- They discuss the everyday barriers families face, from misinformation to transportation challenges, and strategize context-specific outreach approaches.
- This grounding in community realities positions health workers as vital bridges for facilitating community engagement in monitoring.
When local staff are empowered as active agents of learning and change, they can more effectively champion community participation, translating insights into tangible improvements.
Could CBM fit into a more comprehensive system from local monitoring to action?
TGLF’s model supports health workers in this bridging role by providing a comprehensive framework for local monitoring and action.
Through peer learning networks and problem-solving cycles, the model equips health staff to collect, interpret, and act on unconventional monitoring data from their communities.
For instance, in TGLF’s 2022 “Full Learning Cycle” initiative, 6,185 local health workers from 99 countries examined key immunization indicators to inform their analyses of root causes and then map out corrective actions.
- Participants began monitoring their own local health indicators, such as vaccination coverage rates.
- For many, this was the first time they had been prompted to use this data for problem-solving a real-world challenge they face, rather than just reporting up the next level of the health system.
They discussed many factors critical for tailoring immunization strategies.
This transition – from being passive data collectors to active data users – has proven transformative.
It positions health workers not as cogs in a reporting machine, but as empowered analysts and strategists.
By discussing real metrics with peers, participants make data actionable and contextually meaningful.
Guided by expert-designed rubrics and facilitated discussions, health workers translated this localized monitoring data into practical improvement plans.
For an epidemiologist, this represents a significant shift from traditional top-down monitoring paradigms.
By valuing and actioning local knowledge, TGLF’s model demonstrates how community insights can be systematically integrated into immunization decision-making.
However, until now, its actors have been health workers, many of them members of the communities they serve, not service users themselves.
CBM’s focus on monitoring is important – but leaves out key issues around community participation, decision-making autonomy, and strategy.
How could we integrate CBM into a transformative approach?
TGLF’s experiences suggest that CBM could be embedded within comprehensive learning-to-action systems focused on locally-led change.
TGLF’s model is more than a monitoring intervention.
- It combines structured learning, rapid solution sharing, root cause analysis, action planning, and peer accountability to drive measurable improvements.
- These mutually reinforcing components create an enabling environment for health workers to translate insights into impact.
In this framing, community monitoring becomes one critical input within a continuous, collaborative process of problem-solving and adaptation.
Several features of TGLF’s model illustrate how this integration could work in practice:
- Peer accountability structures, where health workers regularly convene to review progress, share challenges, and iterate solutions, create natural entry points for discussing and actioning community feedback.
- Rapid dissemination channels, like TGLF’s “Ideas Engine” for spreading promising practices across contexts, enable local innovations in response to community-identified gaps to be efficiently scaled.
- Emphasis on root cause analysis and systemic thinking equips health workers to interpret community insights within a broader ecosystem lens, connecting localized issues to upstream determinants.
- Cultivation of connected leadership empowers local actors to champion community priorities and navigate complex change processes.
TGLF’s extensive digital network connects health workers across system levels and contexts, enabling them to learn from each other’s experiences with no upper limit to the number of participants.
By contrast, CBM seems to assume that a community is limited to a physical area, which fails to recognize that problem-solving complex challenges requires expanding the range of inputs used.
Within a networked approach that connects both community members and health workers across boundaries of geography, health system level, and roles, CBM could become an integral component of a transformative approach to health system improvement – one that recognizes communities and local health workers as capable architects of context-responsive solutions.
Fundamentally, the TGLF model invites a shift in mindset about whose expertise counts in monitoring and driving system change.
CBM could provide the ‘connective tissue’ for health workers to revise how they listen and learn with the communities they serve.
For immunization programs grappling with persistent inequities, this shift from passive compliance to proactive local problem-solving is critical.
As the COVID-19 crisis has underscored, rapidly evolving public health challenges demand localized action that harnesses the full range of community expertise.
TGLF’s model offers a tested framework for actualizing this vision at scale.
By investing in local health workers’ capacity to learn, adapt, and lead change in partnership with the communities they serve, the model illuminates a promising pathway for integrating CBM into immunization monitoring and beyond.
For epidemiologists and global health practitioners, TGLF’s approach invites a reframing of how we conceptualize and operationalize community engagement in health system monitoring.
It challenges us to move beyond tokenistic participation towards genuine co-design and co-ownership of monitoring processes with local actors.
Realizing this vision will require significant shifts in mindsets, power dynamics, and resource flows.
But as TGLF’s initiatives demonstrate, when we invest in the leadership of those closest to the challenges we seek to solve, transformative possibilities emerge.
Further rigorous research comparing the impacts of different CBM integration models could help accelerate this paradigm shift, surfacing critical lessons for the immunization field and global health more broadly.
TGLF’s model not only offers compelling lessons for reimagining monitoring and improvement in immunization programs, it also provides a pathway for integrating CBM into a system that supports actual change.
CBM practitioners are likely to struggle with how to incorporate it into existing practices.
By investing in frontline health workers as change agents, and surrounding them with an empowering learning ecosystem, the model offers a path to then bring in community monitoring.
Without such leadership from health workers, it is unlikely that communities are able to participate.
The journey to authentic community engagement in health system monitoring is undoubtedly complex.
But if we are to deliver on the promise of equitable immunization for all, it is a journey we must undertake.
TGLF’s model lights one promising path forward – one that positions communities and local health workers as the beating heart of a learning health system.
While Gavi’s evidence brief affirms the promise of CBM for immunization, TGLF’s experience with its own model suggests the full potential of CBM may be realized by embedding it within more comprehensive, digitally-enabled learning systems that activate health workers as agents of change – and do so with both physical and digital communities implementing new forms of peer and community accountability that complement conventional kinds (supervision, administration, donor, etc.).
Share this:
#communityBasedMonitoring #continuousLearning #globalHealth #healthWorkers #HRH #HumanResourcesForHealth #immunization #ImmunizationAgenda2030 #learningStrategy #TheGenevaLearningFoundation #zeroDoseChildren #ZeroDoseLearningHubZDLH_
-
A Chinese national pretended to be U.S. engineers and researchers for almost five years, from 2017 to 2021, and walked away with sensitive aerospace and weapons development software from NASA, the Air Force, the Navy, and the Army. There was no hacking or breaking through firewalls. People simply emailed him what he asked for, because they believed he was someone they knew.
This worries me more than any zero-day vulnerability. The NASA OIG reported that Song Wu asked for the same software several times without explaining why he needed it. Most people miss this kind of red flag because no one teaches them to spot it. We invest millions in technology controls but spend very little on training people to pause and think like a threat actor before sending information.
Export controls are not only about legal compliance. They are also about human behavior. Your employees make export control decisions every day, often without realizing it.
When was the last time your organization ran a spear-phishing simulation aimed at your researchers, not just your finance team?
If your security awareness program doesn't cover identity deception and unusual software requests, it is not thorough enough.
https://thehackernews.com/2026/04/nasa-employees-duped-in-chinese.html
#Cybersecurity #NationalSecurity #Espionage #SecurityAwareness #InfoSec #security #privacy #cloud #infosec -
A Chinese national pretended to be U.S. engineers and researchers for almost five years, from 2017 to 2021, and walked away with sensitive aerospace and weapons development software from NASA, the Air Force, the Navy, and the Army. There was no hacking or breaking through firewalls. People simply emailed him what he asked for, because they believed he was someone they knew.
This worries me more than any zero-day vulnerability. The NASA OIG reported that Song Wu asked for the same software several times without explaining why he needed it. Most people miss this kind of red flag because no one teaches them to spot it. We invest millions in technology controls but spend very little on training people to pause and think like a threat actor before sending information.
Export controls are not only about legal compliance. They are also about human behavior. Your employees make export control decisions every day, often without realizing it.
When was the last time your organization ran a spear-phishing simulation aimed at your researchers, not just your finance team?
If your security awareness program doesn't cover identity deception and unusual software requests, it is not thorough enough.
https://thehackernews.com/2026/04/nasa-employees-duped-in-chinese.html
#Cybersecurity #NationalSecurity #Espionage #SecurityAwareness #InfoSec #security #privacy #cloud #infosec -
A Chinese national pretended to be U.S. engineers and researchers for almost five years, from 2017 to 2021, and walked away with sensitive aerospace and weapons development software from NASA, the Air Force, the Navy, and the Army. There was no hacking or breaking through firewalls. People simply emailed him what he asked for, because they believed he was someone they knew.
This worries me more than any zero-day vulnerability. The NASA OIG reported that Song Wu asked for the same software several times without explaining why he needed it. Most people miss this kind of red flag because no one teaches them to spot it. We invest millions in technology controls but spend very little on training people to pause and think like a threat actor before sending information.
Export controls are not only about legal compliance. They are also about human behavior. Your employees make export control decisions every day, often without realizing it.
When was the last time your organization ran a spear-phishing simulation aimed at your researchers, not just your finance team?
If your security awareness program doesn't cover identity deception and unusual software requests, it is not thorough enough.
https://thehackernews.com/2026/04/nasa-employees-duped-in-chinese.html
#Cybersecurity #NationalSecurity #Espionage #SecurityAwareness #InfoSec #security #privacy #cloud #infosec -
A Chinese national pretended to be U.S. engineers and researchers for almost five years, from 2017 to 2021, and walked away with sensitive aerospace and weapons development software from NASA, the Air Force, the Navy, and the Army. There was no hacking or breaking through firewalls. People simply emailed him what he asked for, because they believed he was someone they knew.
This worries me more than any zero-day vulnerability. The NASA OIG reported that Song Wu asked for the same software several times without explaining why he needed it. Most people miss this kind of red flag because no one teaches them to spot it. We invest millions in technology controls but spend very little on training people to pause and think like a threat actor before sending information.
Export controls are not only about legal compliance. They are also about human behavior. Your employees make export control decisions every day, often without realizing it.
When was the last time your organization ran a spear-phishing simulation aimed at your researchers, not just your finance team?
If your security awareness program doesn't cover identity deception and unusual software requests, it is not thorough enough.
https://thehackernews.com/2026/04/nasa-employees-duped-in-chinese.html
#Cybersecurity #NationalSecurity #Espionage #SecurityAwareness #InfoSec #security #privacy #cloud #infosec -
A Chinese national pretended to be U.S. engineers and researchers for almost five years, from 2017 to 2021, and walked away with sensitive aerospace and weapons development software from NASA, the Air Force, the Navy, and the Army. There was no hacking or breaking through firewalls. People simply emailed him what he asked for, because they believed he was someone they knew.
This worries me more than any zero-day vulnerability. The NASA OIG reported that Song Wu asked for the same software several times without explaining why he needed it. Most people miss this kind of red flag because no one teaches them to spot it. We invest millions in technology controls but spend very little on training people to pause and think like a threat actor before sending information.
Export controls are not only about legal compliance. They are also about human behavior. Your employees make export control decisions every day, often without realizing it.
When was the last time your organization ran a spear-phishing simulation aimed at your researchers, not just your finance team?
If your security awareness program doesn't cover identity deception and unusual software requests, it is not thorough enough.
https://thehackernews.com/2026/04/nasa-employees-duped-in-chinese.html
#Cybersecurity #NationalSecurity #Espionage #SecurityAwareness #InfoSec #security #privacy #cloud #infosec -
.@ArmadaCR is showing is lack of knowledge again by making another HUGE mistake with F-35s who are capability wise just an F-16/18 with stealth that can fly longer distance and can carry more load without understanding the concept of air superiority which is critical for the czech republic to be able to efficiently defend ourselves..
So let me be clear again:
* F-35s is a **SUPPORT AIRCRAFT DESIGNED FOR MISSIONS** to e.g. destroy enemy air defenses, air patrol and alike, because e.g. Russian S-400 are reportedly unable to lock it due to it's coating absorbing and reflecting radar radiation which can do air superiority, but is very sub-optimal due to it's agility in the air
* The Gripen is (multirole) **AIR SUPERIORITY FIGHTER!!** designed to engage with enemy air assets and which has the maneuverability to position itself in range and optimal position to shoot missiles and destroy the enemy aircraft while being able to avoid the missiles shot by an enemy at them, but it sucks at trying to destroy enemy defenses, because it's easily lockable even with the magic anti-radar system on it.These are TWO DIFFERENT AIRCRAFTS designed for TWO DIFFERENT PURPOSES!!!
Defending with F-35s against russians to establish an air superiority is **INSANITY**, because yes it can defend itself probably 1:3 if we are lucky and the stealth actually works against russian missiles (it doesn't always work due to the environment and relying on it is insanity and falling for US's stupid 5th generation fighter advertisement), but the quantity is going to fU%&@ us.
The best scenario for us would be for to have F-22 and F-35 where US can't legally sell the F-22 so our next best realistic option is Gripen or Eurofighter to accompany the F-35s.
These two aircrafts are made to work together for an efficient defense and we should consider expanding our AWACS capabilities as well.
Next Why the fuck are we buying tanks? Like they would be any useful in case any of our neighbours attack us.. 🤦♂️
Thus for us to be able to defend ourselves on land we need force **multipliers** like drone and anti-material weapons and us investing heavy money in management for the drone signal capabilities to be able to manage enemy jamming and anti-drone defenses which also makes us significantly less useless in NATO to help our allies in case of a conflict to do surveillance and drone management..
And to clarify i am not saying that we don't need tanks, ukraine shows us that we clearly do, but we don't need this many with our current budget and with our location in the EU and current alliances we are significantly more capable army if we focus on drones, continue focusing on chemical and biological task, get proper anti-air defenses around the country and train so that we can turn the whole country into a fortress with multiple layers in case of an attack and invest more in special forces.
I urge @Jana_cernochova @P_Fiala @TomasZdechovsky @Czechia_NATO @ODScz @SpoluKoalice @EPPGroup @kducsl @PiratskaStrana to do more research into how to best manage our military budget or this decision will be a disasterous for us in case of further conflict with Russia.
#CZ #Czech #CzechRepublic #TheCzechRepublic #CZArmy #CZPolitics #F35 #gripen #politics #military #russia #xp
-
.@ArmadaCR is showing is lack of knowledge again by making another HUGE mistake with F-35s who are capability wise just an F-16/18 with stealth that can fly longer distance and can carry more load without understanding the concept of air superiority which is critical for the czech republic to be able to efficiently defend ourselves..
So let me be clear again:
* F-35s is a **SUPPORT AIRCRAFT DESIGNED FOR MISSIONS** to e.g. destroy enemy air defenses, air patrol and alike, because e.g. Russian S-400 are reportedly unable to lock it due to it's coating absorbing and reflecting radar radiation which can do air superiority, but is very sub-optimal due to it's agility in the air
* The Gripen is (multirole) **AIR SUPERIORITY FIGHTER!!** designed to engage with enemy air assets and which has the maneuverability to position itself in range and optimal position to shoot missiles and destroy the enemy aircraft while being able to avoid the missiles shot by an enemy at them, but it sucks at trying to destroy enemy defenses, because it's easily lockable even with the magic anti-radar system on it.These are TWO DIFFERENT AIRCRAFTS designed for TWO DIFFERENT PURPOSES!!!
Defending with F-35s against russians to establish an air superiority is **INSANITY**, because yes it can defend itself probably 1:3 if we are lucky and the stealth actually works against russian missiles (it doesn't always work due to the environment and relying on it is insanity and falling for US's stupid 5th generation fighter advertisement), but the quantity is going to fU%&@ us.
The best scenario for us would be for to have F-22 and F-35 where US can't legally sell the F-22 so our next best realistic option is Gripen or Eurofighter to accompany the F-35s.
These two aircrafts are made to work together for an efficient defense and we should consider expanding our AWACS capabilities as well.
Next Why the fuck are we buying tanks? Like they would be any useful in case any of our neighbours attack us.. 🤦♂️
Thus for us to be able to defend ourselves on land we need force **multipliers** like drone and anti-material weapons and us investing heavy money in management for the drone signal capabilities to be able to manage enemy jamming and anti-drone defenses which also makes us significantly less useless in NATO to help our allies in case of a conflict to do surveillance and drone management..
And to clarify i am not saying that we don't need tanks, ukraine shows us that we clearly do, but we don't need this many with our current budget and with our location in the EU and current alliances we are significantly more capable army if we focus on drones, continue focusing on chemical and biological task, get proper anti-air defenses around the country and train so that we can turn the whole country into a fortress with multiple layers in case of an attack and invest more in special forces.
I urge @Jana_cernochova @P_Fiala @TomasZdechovsky @Czechia_NATO @ODScz @SpoluKoalice @EPPGroup @kducsl @PiratskaStrana to do more research into how to best manage our military budget or this decision will be a disasterous for us in case of further conflict with Russia.
#CZ #Czech #CzechRepublic #TheCzechRepublic #CZArmy #CZPolitics #F35 #gripen #politics #military #russia #xp
-
.@ArmadaCR is showing is lack of knowledge again by making another HUGE mistake with F-35s who are capability wise just an F-16/18 with stealth that can fly longer distance and can carry more load without understanding the concept of air superiority which is critical for the czech republic to be able to efficiently defend ourselves..
So let me be clear again:
* F-35s is a **SUPPORT AIRCRAFT DESIGNED FOR MISSIONS** to e.g. destroy enemy air defenses, air patrol and alike, because e.g. Russian S-400 are reportedly unable to lock it due to it's coating absorbing and reflecting radar radiation which can do air superiority, but is very sub-optimal due to it's agility in the air
* The Gripen is (multirole) **AIR SUPERIORITY FIGHTER!!** designed to engage with enemy air assets and which has the maneuverability to position itself in range and optimal position to shoot missiles and destroy the enemy aircraft while being able to avoid the missiles shot by an enemy at them, but it sucks at trying to destroy enemy defenses, because it's easily lockable even with the magic anti-radar system on it.These are TWO DIFFERENT AIRCRAFTS designed for TWO DIFFERENT PURPOSES!!!
Defending with F-35s against russians to establish an air superiority is **INSANITY**, because yes it can defend itself probably 1:3 if we are lucky and the stealth actually works against russian missiles (it doesn't always work due to the environment and relying on it is insanity and falling for US's stupid 5th generation fighter advertisement), but the quantity is going to fU%&@ us.
The best scenario for us would be for to have F-22 and F-35 where US can't legally sell the F-22 so our next best realistic option is Gripen or Eurofighter to accompany the F-35s.
These two aircrafts are made to work together for an efficient defense and we should consider expanding our AWACS capabilities as well.
Next Why the fuck are we buying tanks? Like they would be any useful in case any of our neighbours attack us.. 🤦♂️
Thus for us to be able to defend ourselves on land we need force **multipliers** like drone and anti-material weapons and us investing heavy money in management for the drone signal capabilities to be able to manage enemy jamming and anti-drone defenses which also makes us significantly less useless in NATO to help our allies in case of a conflict to do surveillance and drone management..
And to clarify i am not saying that we don't need tanks, ukraine shows us that we clearly do, but we don't need this many with our current budget and with our location in the EU and current alliances we are significantly more capable army if we focus on drones, continue focusing on chemical and biological task, get proper anti-air defenses around the country and train so that we can turn the whole country into a fortress with multiple layers in case of an attack and invest more in special forces.
I urge @Jana_cernochova @P_Fiala @TomasZdechovsky @Czechia_NATO @ODScz @SpoluKoalice @EPPGroup @kducsl @PiratskaStrana to do more research into how to best manage our military budget or this decision will be a disasterous for us in case of further conflict with Russia.
#CZ #Czech #CzechRepublic #TheCzechRepublic #CZArmy #CZPolitics #F35 #gripen #politics #military #russia #xp
-
The sensory pleasure of academic writing
I thought it was interesting that Sword (2021: loc 1306) frames this as a matter of “cutting ourselves loose from the world of the sense and giving ourselves over to the flow of ideas”. The objection might seem like a pedantic one but I think it’s important to recognise the sensory pleasure which we can and do take in ideas, even if this by no means exhausts what satisfaction can be found in thinking. I want to suggest this sensory pleasure is important for what Fitzpatrick (2019: loc 122) calls ‘generous thinking’: “a mode of engagement that emphasizes listening over speaking, community over individualism, collaboration over competition, and lingering with the ideas that are in front of us rather than continually pressing forward to where we want to go”.
To linger is to stay longer than is strictly necessary, to be reluctant to leave and to be slow to act. If we are reluctant to leave what is keeping us in place? If we are slow to act what leaves us unwilling to move onto the next thing? Generous thinking in this sense entails an unwillingness to immediately mobilise ideas, to see them as resources through which we advance our intellectual career, rather than as focal points for the pleasures which permit our academic lives to be in some reliable sense satisfying.
The sociologist C. Wright Mills once wrote to a struggling friend to remind him of the things in life to get excited about, in a passage which has stayed with me since I first read it over a decade ago:
“You ask for what one should be keyed up? My god, for long weekends in the country, and snow and the feel of an idea and New York streets early in the morning and late at night and the camera eye always working whether you want or not and yes by god how the earth feels when it’s been ploughed deep and the new chartreuse wall in the study and wine before dinner and if you can afford it Irish whiskey afterwards and sawdust in your pants cuff and sometimes at evening the dusky pink sky to the northwest, and the books to read never touched and all that stuff the Greeks wrote about and have you ever read Macaulay’s speeches to hear the English language? And to revise your mode of talk and what you talk about and yes by god the world of music which we just now discover and there’s still hot jazz and getting a car out of the mud when nobody else can. That’s what the hell to get keyed up about.”
It has often felt to me like the pleasure of writing should be understood alongside the sensory pleasures which Mills lists here. He describes “the books to read never touched”, “to revise your mode of talk” and “the feel of an idea” alongside snow, jazz and the mystical stillness of a major city as it sleeps. Every time I read this passage I’m brought back to living in central London, which made it possible to walk through the metropolis in that small window in which it was quiet. There’s a calming quality to these pleasures, a sense of being lifted out of ourselves in a manner which is deeply soothing.
I would suggest that writing, which I understand as a process in which we work with ideas, carries the same potential even if we struggle to realise it. Ideas feel a certain way. Giving form to nebulous insights feels a certain way. Being lost in conversation with people who share your intellectual passions feels a certain way. Being immersed in a thought provoking book feels a certain way. Being frustrated by it feels a certain way. There’s an embodied quality to these pleasures which can be easily overlooked if we too readily invest ourselves in the notion of writing as a cognitive process.
Elbow (1981: 340) draws attention to the embodied quality of new ideas as they emerge for us in the writing process:
“See which part of your body the thought wants to erupt through. Some researchers have found that children have a physical reaction – a piece of tension-release in some part of their body, a shiver of jiggle – when they figure something out. What’s special about figuring something out is that it always consists of a new thought or a new connection, and you can’t have a new thought without really experiencing it.”
There are a rich array of sensory pleasures to be found in the writing process, which take on different meanings for different people and at different points in careers. They can also, I will argue, be squeezed out by other factors which shape what and how we write. In fact this is the norm rather than the exception.
But if we practice naming these experiences, identifying them when they occur, it comes to be easier to recognise and value them. It makes lingering easier, both in terms of justifying it when our to-do list calls to us and allowing the impulse to arise in the first place. To intellectually linger requires making space in which the continual flow of demands can ease, if not be entirely interrupted. This might be taking a book and a notepad to a coffee shop near your office, leaving laptop and phone behind. It might be taking the risk of approaching a speaker at a workshop, in order to voice the ideas which their talk provoked in you. It might be contacting a former collaborator who you now see far too infrequently, in order to suggest catching up via Zoom with no set agenda.
If there is an agenda it ceases to be lingering. It might be that a project or collaboration frames the interaction, with the intention to explore something you might do together in future. But it is possible to linger within the interaction, finding space to enjoy the engagement through which projects might unfold in the future, as opposed to defining your meeting in terms of what it should produce.
Obviously this is more difficult to achieve in practice than it is to opine about in a blog post. If you’re rushing from one commitment to the next it can be extraordinarily difficult to find time to write, let alone savour the ‘feel of the ideas’ you encounter in your writing. If it feels like a continual struggle to find any sustained time for writing, then lingering in the way I’m describing can feel like an unfathomable indulgence: a distraction from both the important and the urgent.
The suggestion frequently made in writing manuals and workshops that academics should not wait for the perfect time to write is certainly correct from a practical perspective. If you can only write on those occasions when it’s possible to block out sustained periods of time freed from other interruptions, then writing will inevitably be confined to summer and sabbaticals at best. But writing in the gaps can also be a stultifying experience, in which the size of the thoughts you think and the ideas you express is curtailed to match the limitations of the time and space available to you.
Writing in the gaps can be a pragmatic adaptation if undertaken in a careful way. It can also be a creative choice if undertaken in a reflective way. But if not embraced with care and reflection it can lead to a shrinking of creative horizons, with rushed writing undertaken in odd moments leading to odd writing marked by a rushed quality.
This is why I believe reflection is so crucial to academic writing: not just reflection on what we write, but on how and why we write. There is nothing intrinsic to how you write which dictates which outcome you will experience. It is unhelpful to imagine there is a right way of approaching your writing, which others have already found but which feels frustratingly beyond your grasp. What is right for you will not be right for someone else. What is right for you at a particular point in your career, or even your week, might not be right for you at another point in time. The key is developing awareness of when you experience those sensory pleasures of ideas that Mills and Elbow describe, and creating conditions where you can linger with them, even if only briefly, amid the demands of academic life.
-
The sensory pleasure of academic writing
I thought it was interesting that Sword (2021: loc 1306) frames this as a matter of “cutting ourselves loose from the world of the sense and giving ourselves over to the flow of ideas”. The objection might seem like a pedantic one but I think it’s important to recognise the sensory pleasure which we can and do take in ideas, even if this by no means exhausts what satisfaction can be found in thinking. I want to suggest this sensory pleasure is important for what Fitzpatrick (2019: loc 122) calls ‘generous thinking’: “a mode of engagement that emphasizes listening over speaking, community over individualism, collaboration over competition, and lingering with the ideas that are in front of us rather than continually pressing forward to where we want to go”.
To linger is to stay longer than is strictly necessary, to be reluctant to leave and to be slow to act. If we are reluctant to leave what is keeping us in place? If we are slow to act what leaves us unwilling to move onto the next thing? Generous thinking in this sense entails an unwillingness to immediately mobilise ideas, to see them as resources through which we advance our intellectual career, rather than as focal points for the pleasures which permit our academic lives to be in some reliable sense satisfying.
The sociologist C. Wright Mills once wrote to a struggling friend to remind him of the things in life to get excited about, in a passage which has stayed with me since I first read it over a decade ago:
“You ask for what one should be keyed up? My god, for long weekends in the country, and snow and the feel of an idea and New York streets early in the morning and late at night and the camera eye always working whether you want or not and yes by god how the earth feels when it’s been ploughed deep and the new chartreuse wall in the study and wine before dinner and if you can afford it Irish whiskey afterwards and sawdust in your pants cuff and sometimes at evening the dusky pink sky to the northwest, and the books to read never touched and all that stuff the Greeks wrote about and have you ever read Macaulay’s speeches to hear the English language? And to revise your mode of talk and what you talk about and yes by god the world of music which we just now discover and there’s still hot jazz and getting a car out of the mud when nobody else can. That’s what the hell to get keyed up about.”
It has often felt to me like the pleasure of writing should be understood alongside the sensory pleasures which Mills lists here. He describes “the books to read never touched”, “to revise your mode of talk” and “the feel of an idea” alongside snow, jazz and the mystical stillness of a major city as it sleeps. Every time I read this passage I’m brought back to living in central London, which made it possible to walk through the metropolis in that small window in which it was quiet. There’s a calming quality to these pleasures, a sense of being lifted out of ourselves in a manner which is deeply soothing.
I would suggest that writing, which I understand as a process in which we work with ideas, carries the same potential even if we struggle to realise it. Ideas feel a certain way. Giving form to nebulous insights feels a certain way. Being lost in conversation with people who share your intellectual passions feels a certain way. Being immersed in a thought provoking book feels a certain way. Being frustrated by it feels a certain way. There’s an embodied quality to these pleasures which can be easily overlooked if we too readily invest ourselves in the notion of writing as a cognitive process.
Elbow (1981: 340) draws attention to the embodied quality of new ideas as they emerge for us in the writing process:
“See which part of your body the thought wants to erupt through. Some researchers have found that children have a physical reaction – a piece of tension-release in some part of their body, a shiver of jiggle – when they figure something out. What’s special about figuring something out is that it always consists of a new thought or a new connection, and you can’t have a new thought without really experiencing it.”
There are a rich array of sensory pleasures to be found in the writing process, which take on different meanings for different people and at different points in careers. They can also, I will argue, be squeezed out by other factors which shape what and how we write. In fact this is the norm rather than the exception.
But if we practice naming these experiences, identifying them when they occur, it comes to be easier to recognise and value them. It makes lingering easier, both in terms of justifying it when our to-do list calls to us and allowing the impulse to arise in the first place. To intellectually linger requires making space in which the continual flow of demands can ease, if not be entirely interrupted. This might be taking a book and a notepad to a coffee shop near your office, leaving laptop and phone behind. It might be taking the risk of approaching a speaker at a workshop, in order to voice the ideas which their talk provoked in you. It might be contacting a former collaborator who you now see far too infrequently, in order to suggest catching up via Zoom with no set agenda.
If there is an agenda it ceases to be lingering. It might be that a project or collaboration frames the interaction, with the intention to explore something you might do together in future. But it is possible to linger within the interaction, finding space to enjoy the engagement through which projects might unfold in the future, as opposed to defining your meeting in terms of what it should produce.
Obviously this is more difficult to achieve in practice than it is to opine about in a blog post. If you’re rushing from one commitment to the next it can be extraordinarily difficult to find time to write, let alone savour the ‘feel of the ideas’ you encounter in your writing. If it feels like a continual struggle to find any sustained time for writing, then lingering in the way I’m describing can feel like an unfathomable indulgence: a distraction from both the important and the urgent.
The suggestion frequently made in writing manuals and workshops that academics should not wait for the perfect time to write is certainly correct from a practical perspective. If you can only write on those occasions when it’s possible to block out sustained periods of time freed from other interruptions, then writing will inevitably be confined to summer and sabbaticals at best. But writing in the gaps can also be a stultifying experience, in which the size of the thoughts you think and the ideas you express is curtailed to match the limitations of the time and space available to you.
Writing in the gaps can be a pragmatic adaptation if undertaken in a careful way. It can also be a creative choice if undertaken in a reflective way. But if not embraced with care and reflection it can lead to a shrinking of creative horizons, with rushed writing undertaken in odd moments leading to odd writing marked by a rushed quality.
This is why I believe reflection is so crucial to academic writing: not just reflection on what we write, but on how and why we write. There is nothing intrinsic to how you write which dictates which outcome you will experience. It is unhelpful to imagine there is a right way of approaching your writing, which others have already found but which feels frustratingly beyond your grasp. What is right for you will not be right for someone else. What is right for you at a particular point in your career, or even your week, might not be right for you at another point in time. The key is developing awareness of when you experience those sensory pleasures of ideas that Mills and Elbow describe, and creating conditions where you can linger with them, even if only briefly, amid the demands of academic life.
-
The sensory pleasure of academic writing
I thought it was interesting that Sword (2021: loc 1306) frames this as a matter of “cutting ourselves loose from the world of the sense and giving ourselves over to the flow of ideas”. The objection might seem like a pedantic one but I think it’s important to recognise the sensory pleasure which we can and do take in ideas, even if this by no means exhausts what satisfaction can be found in thinking. I want to suggest this sensory pleasure is important for what Fitzpatrick (2019: loc 122) calls ‘generous thinking’: “a mode of engagement that emphasizes listening over speaking, community over individualism, collaboration over competition, and lingering with the ideas that are in front of us rather than continually pressing forward to where we want to go”.
To linger is to stay longer than is strictly necessary, to be reluctant to leave and to be slow to act. If we are reluctant to leave what is keeping us in place? If we are slow to act what leaves us unwilling to move onto the next thing? Generous thinking in this sense entails an unwillingness to immediately mobilise ideas, to see them as resources through which we advance our intellectual career, rather than as focal points for the pleasures which permit our academic lives to be in some reliable sense satisfying.
The sociologist C. Wright Mills once wrote to a struggling friend to remind him of the things in life to get excited about, in a passage which has stayed with me since I first read it over a decade ago:
“You ask for what one should be keyed up? My god, for long weekends in the country, and snow and the feel of an idea and New York streets early in the morning and late at night and the camera eye always working whether you want or not and yes by god how the earth feels when it’s been ploughed deep and the new chartreuse wall in the study and wine before dinner and if you can afford it Irish whiskey afterwards and sawdust in your pants cuff and sometimes at evening the dusky pink sky to the northwest, and the books to read never touched and all that stuff the Greeks wrote about and have you ever read Macaulay’s speeches to hear the English language? And to revise your mode of talk and what you talk about and yes by god the world of music which we just now discover and there’s still hot jazz and getting a car out of the mud when nobody else can. That’s what the hell to get keyed up about.”
It has often felt to me like the pleasure of writing should be understood alongside the sensory pleasures which Mills lists here. He describes “the books to read never touched”, “to revise your mode of talk” and “the feel of an idea” alongside snow, jazz and the mystical stillness of a major city as it sleeps. Every time I read this passage I’m brought back to living in central London, which made it possible to walk through the metropolis in that small window in which it was quiet. There’s a calming quality to these pleasures, a sense of being lifted out of ourselves in a manner which is deeply soothing.
I would suggest that writing, which I understand as a process in which we work with ideas, carries the same potential even if we struggle to realise it. Ideas feel a certain way. Giving form to nebulous insights feels a certain way. Being lost in conversation with people who share your intellectual passions feels a certain way. Being immersed in a thought provoking book feels a certain way. Being frustrated by it feels a certain way. There’s an embodied quality to these pleasures which can be easily overlooked if we too readily invest ourselves in the notion of writing as a cognitive process.
Elbow (1981: 340) draws attention to the embodied quality of new ideas as they emerge for us in the writing process:
“See which part of your body the thought wants to erupt through. Some researchers have found that children have a physical reaction – a piece of tension-release in some part of their body, a shiver of jiggle – when they figure something out. What’s special about figuring something out is that it always consists of a new thought or a new connection, and you can’t have a new thought without really experiencing it.”
There are a rich array of sensory pleasures to be found in the writing process, which take on different meanings for different people and at different points in careers. They can also, I will argue, be squeezed out by other factors which shape what and how we write. In fact this is the norm rather than the exception.
But if we practice naming these experiences, identifying them when they occur, it comes to be easier to recognise and value them. It makes lingering easier, both in terms of justifying it when our to-do list calls to us and allowing the impulse to arise in the first place. To intellectually linger requires making space in which the continual flow of demands can ease, if not be entirely interrupted. This might be taking a book and a notepad to a coffee shop near your office, leaving laptop and phone behind. It might be taking the risk of approaching a speaker at a workshop, in order to voice the ideas which their talk provoked in you. It might be contacting a former collaborator who you now see far too infrequently, in order to suggest catching up via Zoom with no set agenda.
If there is an agenda it ceases to be lingering. It might be that a project or collaboration frames the interaction, with the intention to explore something you might do together in future. But it is possible to linger within the interaction, finding space to enjoy the engagement through which projects might unfold in the future, as opposed to defining your meeting in terms of what it should produce.
Obviously this is more difficult to achieve in practice than it is to opine about in a blog post. If you’re rushing from one commitment to the next it can be extraordinarily difficult to find time to write, let alone savour the ‘feel of the ideas’ you encounter in your writing. If it feels like a continual struggle to find any sustained time for writing, then lingering in the way I’m describing can feel like an unfathomable indulgence: a distraction from both the important and the urgent.
The suggestion frequently made in writing manuals and workshops that academics should not wait for the perfect time to write is certainly correct from a practical perspective. If you can only write on those occasions when it’s possible to block out sustained periods of time freed from other interruptions, then writing will inevitably be confined to summer and sabbaticals at best. But writing in the gaps can also be a stultifying experience, in which the size of the thoughts you think and the ideas you express is curtailed to match the limitations of the time and space available to you.
Writing in the gaps can be a pragmatic adaptation if undertaken in a careful way. It can also be a creative choice if undertaken in a reflective way. But if not embraced with care and reflection it can lead to a shrinking of creative horizons, with rushed writing undertaken in odd moments leading to odd writing marked by a rushed quality.
This is why I believe reflection is so crucial to academic writing: not just reflection on what we write, but on how and why we write. There is nothing intrinsic to how you write which dictates which outcome you will experience. It is unhelpful to imagine there is a right way of approaching your writing, which others have already found but which feels frustratingly beyond your grasp. What is right for you will not be right for someone else. What is right for you at a particular point in your career, or even your week, might not be right for you at another point in time. The key is developing awareness of when you experience those sensory pleasures of ideas that Mills and Elbow describe, and creating conditions where you can linger with them, even if only briefly, amid the demands of academic life.