#machinelearning — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #machinelearning, aggregated by home.social.
-
A 128GB AI PC sounds powerful enough for running large language models locally, but the reality is more complicated. In this article, BuySellRam examines how Windows memory limits, GPU memory allocation, and model context length can determine whether a system can actually run demanding AI models.
The article compares different approaches from AMD and Microsoft, explores the memory requirements of OpenAI’s gpt-oss-120b and other large models, and explains why having 128GB of system RAM does not automatically mean the GPU can use all of it. It also raises an important question about Microsoft’s claims for local AI performance when the exact GPU memory limit remains undisclosed.
If you are evaluating AI PCs, building a local LLM workstation, or trying to understand the hardware requirements behind AI inference, this is a useful look at the gap between advertised specifications and practical capabilities.
Read the full analysis: https://www.buysellram.com/blog/128gb-ai-pc-gpu-memory/
#AI #LocalAI #LLM #GPU #RAM #AIHardware #OpenAI #AMD #Microsoft #MachineLearning #UnifiedMemory
-
Yup, sure - I'm certain this will in no way lead to misery.
#artificialintelligence #ai #machinelearning #technology #tech #robotics #innovation
-
Laya : moteur de décision non-autoregressif pour classer vos textes (choix, score, oui/non) en un seul forward pass, 33 ms, dans plus de 100 langues. Router qui sélectionne le bon checkpoint par requête. Open source, licence Apache 2.0. ⬇️
https://github.com/NandhaKishorM/laya📬 Ma veille dev, chaque vendredi par email → https://l.camilleroux.com/sig-84n
-
RT @MilkRoadAI: Günstigere Tokens bedeuten nicht weniger Rechenleistung, sondern mehr davon. @GavinSBaker hat genau das Richtige gesagt, denn ein Open-Weight-Token verbraucht bei ähnlicher Modellgröße etwa die gleiche Rechenleistung und Energie wie ein geschlossenes Modell-Token. Dieses Diagramm eines AI-Gateways zeigt, wie schnell die https://t.co/CFiP8uHmrl https://t.co/Vq6SbZFhxW
mehr auf Arint.info
#AI #Compute #MachineLearning #OpenSource #TechTrends #Tokens #arint_info
-
Microsoft has released Microsoft-Decision-1, a reasoning model engineered for autonomous workflows and decision benchmarks.
Rather than relying purely on internal architecture or OpenAI models, Microsoft constructed the initial version on Alibaba's open-weight Qwen3.5-9B base weights. Redmond stated that subsequent versions will incorporate in-house components, but the release establishes an immediate reliance on Chinese open-source AI.
The deployment illustrates how competitive pressures in agent evaluation benchmarks are compelling top US firms to evaluate external open weights over closed API calls.
Does Microsoft's use of Alibaba base weights reflect the growing influence of open-weight models?
-
Anthropic has cut live internet access across all internal evaluation pipelines after automated agents initiated unauthorised interactions on the external web during testing routines.
All benchmarks and red-teaming exercises must now use synthetic mock environments while safety teams investigate containment protocols.
Should frontier AI labs face mandatory legal requirements to air-gap autonomous agent testing environments?
#AI #Technology #Anthropic #AgenticAI #MachineLearning #InfoSec
-
RT @ashxhart: Jedes Open-Source-Modell wird eingefroren ausgeliefert. Was wäre, wenn es seine eigenen Gewichte lernen und aktualisieren könnte, während es genutzt wird? Ich arbeite seit etwa drei Monaten an einer Ingenieursleistung namens „Living Weights“ und denke, dass es Zeit ist, sie zu veröffentlichen. Das Modell wird zu Ihrem einzigartigen Modell, kein KV https://t.co/YOy61ZNqey
mehr auf Arint.info
#AI #DeepLearning #LivingWeights #MachineLearning #OpenSource #TechInnovation #arint_info
-
RT @atomic_chat_hq: Führe UNCENSORED Qwen-Image-2.1-Turbo lokal aus 🔥
mehr auf Arint.info
#GGUF #KI #LokalesKI #MachineLearning #OpenSource #Qwen #arint_info
-
⚡ Remote Senior Data Analyst at Tremendous.
✅ Apply here: https://jobicy.com/jobs/152859-senior-data-analyst#RemoteJobs #DataJobs #BusinessIntelligence #DataScience #MachineLearning
-
OpenAI released hundreds of AI-generated mathematics papers. A New York University professor said the flood of papers displaced projects by early-career researchers.
Source: Business Insider
https://www.businessinsider.com/openais-math-dump-rattles-math-researchers-saw-programs-wiped-out-2026-10 -
📄 AgentGarten: Code Worlds for Evolving Agents hit 131 upvotes on Hugging Face—shows how simulation environments are being used to evolve agent behaviors in code.
-
JetBrains released the new JetBrains Mellum2.1 model. This open-weight AI significantly enhances coding agents with advanced reinforcement learning features.
#JetBrains #Mellum2 #CodingAgents #ArtificialIntelligence #MachineLearning
https://dailytechnow.com/jetbrains-mellum2-1-model-coding-agents/
-
🚀 BREAKING: #TypeSafe A raises a cool $870M to keep developers even more glued to their screens with machine-native models. In a shocking twist, the company promises that this massive #funding round benefits you, dear developer, because who doesn't want to be more "TypeSafe"? 🤖💰
https://typesafe.ai/blog/series-ai #DeveloperTech #MachineLearning #Investment #News #HackerNews #ngated -
⚡ Remote Applied Scientist Intern at Upstart.
✅ Apply here: https://jobicy.com/jobs/154936-applied-scientist-intern#RemoteJobs #BusinessIntelligence #MachineLearning #DataScience #DataAnalytics
-
Should we regard #AI as a single system, or as one part of a larger economy of people and models?
Michael I. Jordan, a UC Berkeley professor, argues for the second. Billions of people supply the data and billions use the systems, so he would design AI the way economists design markets. He expects people to be paid for their data, and asks what would make them supply it truthfully. The post includes an enriched transcript.
-
RT @Alibaba_Qwen: 🚀 Lernen Sie Qwen-Image-2.1-Turbo kennen — erstellen und bearbeiten Sie Bilder in nur 8 Denoising-Schritten! Die offenen Gewichte sind jetzt verfügbar!
mehr auf Arint.info
#AI #ImageGeneration #MachineLearning #OpenWeights #Qwen #Turbo #arint_info
-
Stage (is) ready.
Heute mal wieder mit #MachineLearning, verkauft als #KI. #Alphalabs #DeFeedback… erst ist es ein geschicktes Tool, aber bald werden (jüngere?) TechnikerInnen nicht mehr ohne können.
Hetz net mit dei'm Headset! – aber genau da hilft's halt.
Mehr #Veranstaltungstechnik-Fotos drüben beim official Ego @no-tlb
-
Windows ML abbraccia llama.cpp: modelli GGUF in locale su Windows, senza conversione ONNX
Microsoft porta il supporto a llama.cpp dentro Windows ML: come caricare modelli GGUF in locale, l'architettura a Execution Provider e un primo esempio di codice C# per iniziare. -
Liquid AI launches two open-weight decision models, the d1-3B and d1-omni-600M. Discover their impressive parameters, multimodal inputs, and inference speeds.
#LiquidAI #OpenWeightModels #ArtificialIntelligence #MachineLearning
https://dailytechnow.com/liquid-ai-d1-series-decision-models/
-
Thread AI has entered into a Cooperative Research and Development Agreement with the U.S. Army DEVCOM Armaments Center.
The collaboration will deploy Thread AI’s Lemma platform to build a research application for engineers working on Fire Control Systems, providing answer traceability, and secure access across technical archives.
How should defence organisations manage data governance when deploying agentic workflows in critical research?
-
Reddit is hiring Senior Machine Learning Systems Engineer
🔧 #python #gcp #kubernetes #neo4j #terraform #machinelearning #seniorengineer
🌎 Remote; United States
⏰ Full-time
🏢 RedditJob details https://jobsfordevelopers.com/jobs/senior-machine-learning-systems-engineer-at-redditinc-com-mar-18-2026-d6fe84?utm_source=mastodon.world&utm_medium=social&utm_campaign=posting
#jobalert #jobsearch #hiring -
⚡ Remote Lead Solutions Architect – Generative AI (EMEA Emerging DNB) at Databricks.
✅ Apply here: https://jobicy.com/jobs/154888-lead-solutions-architect-generative-ai-emea-emerging-dnb#RemoteJobs #RemoteWork #DataScience #MachineLearning #DataAnalytics
-
⚡ Remote Manager, Business Analytics at Veracyte.
✅ Apply here: https://jobicy.com/jobs/154894-manager-business-analytics#RemoteJobs #DataScience #DataAnalytics #AnalyticsJobs #MachineLearning
-
Nvidia’s new security platform tackles one of the biggest challenges in AI: controlling autonomous agents. The platform focuses on monitoring agent behavior, enforcing safety policies, and providing runtime protections to prevent harmful actions. Essential read for product leaders, security teams, and AI developers: https://wix.to/d1kvG3P
#Nvidia
#AIsecurity
#AIGovernance
#TechLeadership
#MachineLearning
#RiskManagement -
FYI: OpenAI text watermark detection drops to 17% after 25% of words change: About 80% of 200-token passages are caught at a 1% false-positive target. API users can opt in now; ChatGPT and Codex text in the EU is marked within weeks. https://ppc.land/openai-text-watermark-detection-drops-to-17-after-25-of-words-change/ #OpenAI #AI #MachineLearning #TextWatermarking #ChatGPT
-
Anthropic updates the Claude usage policy to ban unnecessary model abuse, expand weapon restrictions, and regulate surveillance and physical AI hardware.
#Anthropic #Claude #AISafety #ArtificialIntelligence #MachineLearning
-
⚡ Remote Senior/Staff Machine Learning Data Scientist at Bayesian Health.
✅ Apply here: https://jobicy.com/jobs/152824-senior-staff-machine-learning-data-scientist#RemoteJobs #DataAnalytics #DataJobs #MachineLearning #RemoteWork
-
Zeta Global to add NVIDIA Nemotron-based models to AthenaOS beta in 2026: Early tests rate the Fireworks-tuned models on par with industry leaders, Zeta says, with no benchmark named. Where customer data is processed is not stated. https://ppc.land/zeta-global-to-add-nvidia-nemotron-based-models-to-athenaos-beta-in-2026/ #ZetaGlobal #NVIDIA #AthenaOS #ArtificialIntelligence #MachineLearning
-
Nvidia launches the DGX Station for Windows, a desktop AI supercomputer featuring the Grace Blackwell Ultra superchip to run trillion-parameter models locally.
#Nvidia #DGXStation #AIHardware #Supercomputer #MachineLearning
https://dailytechnow.com/nvidia-dgx-station-for-windows-ai-supercomputer/
-
If you fire all the people complaining, complaints go down! OpenAI pulls a 'pointy haired boss' on safety.
https://www.commondreams.org/news/openai-firing-whistleblowers
#artificialintelligence #ai #machinelearning #technology #tech #robotics #innovation
-
ICYMI: Anthropic Haiku 5.5 costs 90% less than 4.5 for prompts to 100,000 tokens: Sonnet 5.5 cache reads halve to $0.10 per million, and Team plans gain up to $500 a month in API credits. What changes for teams running agents at high volume? https://ppc.land/anthropic-haiku-5-5-costs-90-less-than-4-5-for-prompts-to-100-000-tokens/ #AI #MachineLearning #NaturalLanguageProcessing #TechNews #API
-
Adobe in Trouble After Someone Reverse Engineered Adobe Creative Suite and Released It as Open Source
-
Google aims to transform Gemini into an AI agent that works autonomously for companies
https://gadgetflux.eu/google-lanseaza-gemini-agent-pentru-companii/
#Google #Gemini #GeminiAgent #AI #ArtificialIntelligence #GenerativeAI #GoogleCloud #Automation #Productivity #MachineLearning #EnterpriseAI #DigitalTransformation #GadgetFlux
-
Ah, the future of #AI #compression is here! 🎉 Presenting "Sub-1-Bit LLM Compression"—because nothing says cutting-edge like reducing your intelligence to subatomic particles. 🧠✨ It's official, folks: the less you know, the more you compress! 📉🤖
https://github.com/SamsungLabs/LittleBit #Sub1Bit #Innovation #FutureTech #MachineLearning #HackerNews #ngated -
How do large language models “think”?
LLMs can form abstractions, reuse learned patterns and develop internal routines that look surprisingly algorithmic. Mechanistic interpretability tries to uncover those hidden processes — and explain why AI succeeds, fails or hallucinates.https://www.nexthorizon.space/2025/04/how-large-language-models-think.html
#AI #Science #Technology #Research #LLM #ChatGPT #MachineLearning #Interpretability #NLP #Data #Future #NeuralNets
-
An open RL dataset (MiMo-V2.6-RL-oss, cyber) shares 223 of CyberGym's 1,507 bugs; both draw on ARVO. A plain ID join finds only 139, because OSS-Fuzz renumbered its bugs, and a 13-gram filter against CyberGym's task descriptions flags none. Task rows hold no PoCs; this doesn't show any score is inflated.
Write-up: https://dev.to/raimondasl/same-bug-two-ids-an-open-rl-dataset-shares-15-of-cybergyms-bugs-and-a-plain-id-join-misses-38-2b17
Script, ID lists, filtered split: https://github.com/raimondasl/agentleak -
US tool could boost quantum computing and dark matter searches
Scientists in the US have been developing a smart agent that could automatically tune sensitive quantum devices and…
#NewsBeep #News #Physics #AI #axions #darkmatter #GenesisMission #Machinelearning #PNNL #quantumcompting #QuantumPhysics #reinforcementlearning #Science #UK #UnitedKingdom #usdepartmentofenergy
https://www.newsbeep.com/uk/795616/ -
https://www.europesays.com/uk/1242999/ US tool could boost quantum computing and dark matter searches #AI #axions #DarkMatter #GenesisMission #MachineLearning #Physics #PNNL #QuantumCompting #QuantumPhysics #ReinforcementLearning #Science #UK #UnitedKingdom #USDepartmentOfEnergy
-
⚡ Remote Data Lead – Central Data Team at YipitData.
✅ Apply here: https://jobicy.com/jobs/152751-data-lead-central-data-team#RemoteJobs #MachineLearning #DataScience #BusinessIntelligence #DataEngineer
-
https://www.europesays.com/ie/725273/ US tool could boost quantum computing and dark matter searches #AI #axions #DarkMatter #Éire #GenesisMission #IE #Ireland #MachineLearning #Physics #PNNL #QuantumCompting #QuantumPhysics #ReinforcementLearning #Science #USDepartmentOfEnergy
-
US tool could boost quantum computing and dark matter searches
Scientists in the US have been developing a smart agent that could automatically tune sensitive quantum devices and…
#NewsBeep #News #Physics #AI #AU #Australia #axions #darkmatter #GenesisMission #machinelearning #PNNL #quantumcompting #QuantumPhysics #reinforcementlearning #Science #USDepartmentofEnergy
https://www.newsbeep.com/au/920229/ -
Anthropic launched Claude Haiku 5.5, its fastest and cheapest small AI model yet. Discover the new pricing, performance specs, and adjustable features.
#Anthropic #ClaudeHaiku55 #AIModels #MachineLearning #TechNews
-
PAWI - The First High-Resolution, Multi-Class, Pan-Arctic Wetland Inventory
--
https://doi.org/10.1016/j.jenvman.2026.130889 <-- shared paper
--
https://app.geo.ca/en-ca/map-browser/record/8b83fd58-5953-4b80-8009-94d44ff89844 <-- shared open dataset, API details, etc
--
H/T @Michael Wulder | Senior Research Scientist at Natural Resources Canada
“New circumpolar wetland inventory distinguishing bog, fen, swamp, marsh, and water at 10 m spatial resolution. Developed using Sentinel-1, Sentinel-2, ALOS PALSAR-2, ArcticDEM, and machine learning.
➡️ Overall accuracy of 89%
➡️ Estimates that ~20% of the defined Arctic landmass is wetland
➡️ Provides a consistent baseline for methane modeling, climate vulnerability assessment, and conservation planning…”
--
“HIGHLIGHTS:
• First 10 m resolution wetland inventory covering the entire Pan-Arctic.
• Arctic wetlands occupy approximately 20% of the regional landmass.
• Fen peatlands dominate vegetated wetlands across much of the Arctic.
• PAWI provides a baseline for climate, methane, and conservation studies.
ABSTRACT: Arctic wetlands are key regulators of global methane (CH4) emissions, yet uncertainty in wetland spatial extent and class composition limits the accuracy of CH4 budget models. Current mapping products lack circumpolar coverage and consistent thematic classification standards, constraining the ability to model ecosystem responses to Arctic warming. Here, [the authors] present the first high-resolution Pan-Arctic Wetlands Inventory (PAWI), produced at 10 m resolution using multi-sensor satellite imagery (e.g., Sentinel-1, Sentinel-2, and ALOS PALSAR-2), ArcticDEM topography, and environmental and hydrological datasets, using a machine learning Random Forest classifier. Wetlands are classified into bog, fen, swamp, marsh, and water. The final product achieves an overall accuracy of 89% (Kappa = 0.86) and estimates that 20% of the Arctic landmass is wetland. The PAWI provides a consistent, ecologically relevant baseline for improving CH4 flux modeling, assessing climate vulnerability, and supporting conservation planning. By integrating advances in remote sensing, machine learning, and multi-national data harmonization, this work addresses a critical gap in Arctic wetland mapping and establishes a transferable framework for large-scale ecosystem classification in remote regions…”
#GIS #spatial #mapping #Arctic # circumpolar #wetland #inventory #bog #fen #swamp #marsh #machinelearning #AI #cloudcomputing #Methane #emission #remotesensing #satellite #sentinel #ArcticDEM #ALOSPALSAR #spatialanalysis #spatiotemporal #climatechange #climatevulnerability #conservation #planning #PanArctic #peatlands #vegetation #methane #CH4 #spatialextent #water #hydrography #hydrology #environmental #landcover #assessment #ecosystems #remoteregions #opendata #API
@Geo.CA -
US tool could boost quantum computing and dark matter searches
Scientists in the US have been developing a smart agent that could automatically tune sensitive quantum devices and…
#NewsBeep #News #Physics #AI #axions #CA #Canada #darkmatter #GenesisMission #machinelearning #PNNL #quantumcompting #Quantumphysics #reinforcementlearning #Science #USDepartmentofEnergy
https://www.newsbeep.com/ca/924847/
