#workslop — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #workslop, aggregated by home.social.
-
I have a 4 month teaching assistant job coming up in September and I'm absolutely dreading the use of AI by students. Last year, grading the exams was absolutely dreadful. The exams were take home, and a 1-mark question would get a whole paragraph answer. Obviously AI generated, but no way to proof it (except in the cases where the students were sloppy enough to paste in "ChatGPT said:" into their answer).
Talking to one of my cohort mates who is active in our union, we are now thinking of raising the issue inflation of work required by us because of AI slop. We have a contract negotiation round coming up soon, and so this could get VERY interesting.
#AIslop #workslop #academia #union -
Reliance on AI results in both worse output and cognitive atrophy.
This article suggests steps to avoid both problems, but I feel very sure that we're in for more polarisation of cognitive abilities. A small minority will flourish while an overwhelming majority will not. Not good.
https://theconversation.com/offloading-work-tasks-to-ai-comes-with-a-cost-to-our-brains-289384
-
#KnowledgeByte: "#AI #Workslop" is a term used to describe AI-generated work content that appears polished on the surface but lacks the necessary substance, context, or quality to meaningfully advance a task.
It is low-effort output that often creates more work for colleagues who have to correct, edit, or redo it.
https://knowledgezone.co.in/posts/AI-Workslop-690980f082e0c0efef5321dc
-
@elasticsoul Very true, there're some good use cases for LLMs, but frankly for them to work based on company-specific trained data, there first needs to be a substantial amount of prep work of that data, which has been a real mess for decades, as any good KMer out there may confirm.
Somehow, feels like we've forgotten to do the proper prep work (Usually, done by humans) thinking it'll all work out vs. GIGO, which is what we currently enjoy with #Workslop 😔
.cc @pluralistic
-
@elasticsoul Obvious mistakes, indeed, we + the planet keep paying for with our very own lives and well-being, which is why I question their so-called leadership skills.
It's starting to feel like we're heading far too fast into the slaughterhouse vs. asking for some accountbility and responsibility of their own actions that affect us all 😤😔
-
RE: https://mastodon.social/@elsua54/116839989797150322
Never forget: executives get paid millions to make obvious mistakes.
"A July 2025 MIT Media Lab report found that 95 percent of organisations saw no measurable return on their generative AI investments, despite billions in spending. Goldman Sachs reached a similar conclusion in March 2026, finding no meaningful relationship between AI adoption and productivity gains at the economy-wide level, even as 70 percent of S&P 500 management teams discussed AI on earnings calls."
-
What's the best way to deal with a colleague sending me #workslop, when I know they're under enormous pressure to use LLMs for everything?
-
"Companies that pushed hardest to adopt generative AI are now contending with a problem the technology was supposed to prevent: their work is getting worse. Two articles published by Harvard Business Review this month describe a feedback loop in which AI-generated low-quality output degrades the information companies rely on to make decisions, a phenomenon the authors call “knowledge decay.”
The June 2026 HBR article, written by Oxford operations management professor Matthias Holweg and Babson College professor Thomas Davenport, argues that the damage goes beyond individual errors. When employees use AI to produce work that looks polished but contains mistakes or lacks substance, colleagues downstream waste time verifying, correcting, or redoing it. As those errors compound across teams and departments, the organisation’s collective knowledge base deteriorates.
The term for this low-quality AI output already has a name. BetterUp Labs and Stanford’s Social Media Lab coined “workslop” in a September 2025 HBR article to describe AI-generated content that masquerades as good work but lacks the substance to advance a task. Their survey of 1,150 US full-time workers found that 41 percent had received workslop in the preceding month, with each incident requiring an average of one hour and 56 minutes to sort out."
https://thenextweb.com/news/ai-workslop-knowledge-decay-harvard-business-review-productivity
-
Не лопнет. Сдуется. И наконец начнут считать
Все ждут, что ИИ-пузырь хлопнет. Картинка в голове простая: однажды утром рынок рухнет, как доткомы, и всё закончится. Но это неправильная метафора. Лопается не способность моделей и не «ИИ вообще». Сдувается финансовая архитектура вокруг них - медленно, в другом слое, чем тот, на который направлены глаза. Аргумент про то что растёт выручка можно в 2026 читать ровно наоборот. Пока подписки были щедро субсидированы, вопрос «а что мы с этого получаем» можно было не задавать. В первом квартале 2026-го фронтир-компании перевели корпоративных клиентов на оплату по токенам - субсидия кончилась, вопрос задали, ответ вернулся пустым. Uber сжёг весь годовой бюджет на токены за квартал, и его операционный директор честно признал, что не может провести линию от красивых метрик к отгруженной пользе. SemiAnalysis показал экономику на одного пользователя в лоб: на подписке за $200 в месяц можно сжечь токенов на $8–14 тысяч - провайдер доплачивает за то, что вы им пользуетесь. Meta через пару недель после того, как сама подстёгивала сотрудников жечь побольше токенов, ввела лимиты. Обе компании, по данным WSJ, обсуждают резкое снижение цен на и без того убыточный сервис. А отчёт KPMG про триумф агентного ИИ тихо сняли, когда выяснилось, что десятки ссылок в нём - галлюцинации модели, которой и поручили этот отчёт написать. По сути - это схлопывание схемы финансирования. И показательнее всего то, что миф о продуктивности рушится даже у тех, кто продаёт лопаты. 9 июня официальный аккаунт AWS - да, того самого Amazon, который зарабатывает на каждом вашем токене, - написал , что больше ИИ-кода не делает команду быстрее, а может и замедлить. Шесть миллионов просмотров. Когда поставщик инфраструктуры публично сдаёт главный тезис собственного маркетинга, это не оговорка - это «покажи окупаемость», пришедшее из самого замка. Свежий NBER подтверждает арифметику: строк кода стало больше, а реально отгруженных приложений - нет.
https://habr.com/ru/articles/1050102/
#ИИпузырь #искусственный_интеллект #окупаемость_ИИ #токеномика #workslop #когнитивный_долг #LLMинги #человек_в_контуре #продуктивность
-
KPMG-Bericht über KI-Vorteile bei SBB ist voller KI-Halluzinationen
> KPMG hat sich bei einem Bericht über die Vorteile von Künstlicher Intelligenz anscheinend allzu sehr auf die Künstliche Intelligenz (KI) verlassen.
-
The idea of workslop is that you will want to read what they didn't want to write.
Surely this is a great opportunity for business and productivity.
-
Wie sich manche krampfhaft an KI Assistenten klammern, die "schnell" Antworten liefern, aber das konstant ungenau machen.
Ich meinte, wenn das Ergebnis des KI Assistenten schon seit zwei Wochen immer wieder fehlerhafte Abfragen liefert (und damit varierende, nicht vergleichbare Ergebnisse), sollte man evtl doch zu stabilen selbstprogrammierten Abfragen wechseln.
Die Reaktion war, dass AI in Zukunft funktionieren muss und man sich nicht an Legacy Tools aufhalten sollte.
Okay. Das ist dann so der Punkt, wo ich argumentativ auch nicht mehr weiter weiß.
-
Es gibt nicht nur Vibe Coding. Es gibt auch Menschen, die machen Vibe Working. Sie versuchen jegliche Tätigkeit mit LLM zu lösen, weil sie glauben, keine Zeit zu finden, sich selbst in das Thema einzuarbeiten.
Am Ende stehen sie vor einem Resultat, von dem sie nicht mehr wissen, wie sie dahingekommen sind. Sie können es nicht mal erklären, wenn man Fragen dazu hat.
Manchmal ist es frustrierend.
-
Heute eine mit ChatGPT generierte Mail bekommen (also fein mit Emoticon Bulletpointlisten), ohne Corporate Design Footer.
Als Phishing gemeldet.
-
Schon zehn Minuten mit einem KI-Chatbot reduzieren unsere Denkleistung - KI
> Wenn wir Rechenaufgaben oder Übungen zum Textverständnis an KI-Tools abgeben, verlernen wir, sie selbst zu lösen, zeigt eine britisch-amerikanische Studie
-
Seufz, Kollege nutzt LLM um Guidelines und Best Practice zur Diagrammerstellung zu überprüfen. Best practice ist in einem Confluence Space, Diagramm ist ein Draw.io (in unserem EAM).
40% der Findings sind false negative. Ein Finding ausgedacht.
Der Text, den die LLM dazu schreibt hat mehr Buchstaben, als das XML des Diagramms.
Auf die Frage, wie ich sowas mache: Ich benutze mein Gehirn.
Man muss Microsoft ernst nehmen. LLMs sind für das Entertainment, nicht für den produktiven Einsatz gedacht. Nur ist das Entertainment allein auf der Prompt-Seite. Die Menschen, die das Ergebnis reviewen müssen, leiden darunter.
-
:notAI: 💩 "The researchers found that 40% of workers had encountered workslop within a month, and then spent an average of 3.4 hours a month dealing with it – which the study estimates adds up to $8.1m in lost productivity for a 10,000-person organization."
https://www.theguardian.com/technology/2026/apr/14/ai-productivity-workplace-errors
-
Bosses say AI boosts productivity – workers say they’re drowning in ‘workslop’
https://www.theguardian.com/technology/2026/apr/14/ai-productivity-workplace-errors
-
2026 workers say they’re drowning in ‘workslop’ *
https://www.theguardian.com/technology/2026/apr/14/ai-productivity-workplace-errors* WORKSLOP refers to AI-generated work that seems polished but is flawed and in need of heavy corrections
https://www.betterup.com/workslop -
"Workslop is an unintended consequence of the AI boom. It’s what happens when employees use AI to quickly generate work that seems polished – at least superficially – but is in fact so flawed or inaccurate that it needs to be heavily corrected, cleaned upor even completely redone after it’s passed on to colleagues."
Seems like the usual slop, just in the context of work. No objections to "workslop" being a more narrow term, though.
https://www.theguardian.com/technology/2026/apr/14/ai-productivity-workplace-errors
-
"[A] recent survey of 5,000 white-collar US workers found that 40% of non-managers say AI saves them no time at all at work, while 92% of high-level executives say it makes them more productive."
Good insight into which jobs can be safely automated.
https://www.theguardian.com/technology/2026/apr/14/ai-productivity-workplace-errors
#news #technology #TechNews #LLMs #workslop #AI #work #automation
-
New one from Cal Newport. New research shows that using #AI increases shallow efforts while decreasing the amount of time spent on tasks that move the needle. It follows history or other productivity tools that haven't necessarily produced stronger work or a better work environment.
-
Die Schatten IT, die durch agentische AI Tools aufgebaut wird, bricht uns irgendwann mal das Genick.
-
"At Microsoft, managers are including questions about AI use in performance discussions. Employees are supposed to quantify how they are using AI tools in their workflows."
Solidarity with all workers affected by this ✊
-
Die KI-Falle für Nachwuchs-Programmierer: Schneller coden, weniger verstehen
> I-Werkzeuge versprechen höhere Effizienz in der Softwareentwicklung. Einer dieser Anbieter ist Anthropic mit Claude Code. Der Hersteller hat nun untersucht, welchen Preis diese Produktivität haben könnte. In einem kontrollierten Experiment mit 52 Softwareentwicklern hat sich gezeigt: Wer beim Lernen neuer Programmierfähigkeiten auf KI setzt, schneidet bei Verständnistests deutlich schlechter ab. Man könnte auch sagen, durch die Delegation an KI wird der Entwickler nicht schlauer.
Ach...
Captain Obvious hat heute Sonderschichten
