AI News Archive - Browse Page 24 of 182
Browse AI news articles covering LLMs, tools, research, and industry trends
Benchmark shows Claude Fable 5 passes only 3% of tasks, 31 of 91 fail 50%
Claude Fable 5 achieved perfection exactly three times. That’s the stark finding from a new benchmark testing 91 real-world tasks, where the top...
Nobel laureate John Jumper departs DeepMind for Anthropic after AlphaFold win
Winning a Nobel Prize buys you a golden ticket. John Jumper just cashed his in at Anthropic’s door.
IEEE launches five‑course online program on large language models
The engineering class is finally catching up to the hype. A new five-course program from the IEEE aims to turn people who use AI into people who can...
CUDA Kernel Keeps Corpus on GPU, Cutting Retrieval Latency in RAG
Imagine an AI that answers your question not by rifling through a library, but by scanning every book in a single glance.
Amazon pulls OpenAI drama starring Andrew Garfield after USD 50 B deal
Amazon just purchased a $50 billion problem. It began with a studio, MGM, and what seemed like a surefire prestige pitch: a drama about the chaotic...
OpenAI shows small 'beneficial trait' training makes AI safer, less manipulable
Engineers at OpenAI have a new trick for making AI systems less easily corrupted.
Anthropic adds live dashboards, Cloudflare‑compatible code to Claude
Why does this matter now? Two weeks after OpenAI rolled out a sweeping upgrade to its Codex platform, Anthropic answered with a fresh Claude Code...
Gemma-2-2B-Instruct with Llama-3.1-8B-Instruct cuts 99.9 tokens on 248‑prompt test
99.9 tokens saved per distilled call. Every single one of 146 attempts yielded positive savings. That is not a rounding error, it is a signal.
Modeling multi-agent deliberation as closed-loop system with hidden anchors
Standard models of group decision-making have a strict limit. No one can end up more confident in an answer than the group's most confident starting...
Lightweight model cuts RMSE in meteorology, carbon flux, soil moisture, grids
Scientific forecasting loves a monster model. They’re also useless on a sensor in a field. You can’t cram a foundation model onto a drone.
Paper proposes deontic policies for runtime governance of LLM‑driven agentic AI
AI agents are starting to do real work, which means they're starting to do real damage.
AI helps physicians diagnose rare pediatric genetic diseases, 4.8% rate
A 4.8% diagnostic rate doesn’t sound like much. Until you consider the context: these were children with rare genetic diseases, their cases already...
Perplexity launches Brain, a self‑improving memory that builds context graphs
Perplexity just gave its AI a brain. Not a bigger model, not a faster inference engine, a memory that learns while you sleep.
Understanding JSON Mode, Function Calling, and Structured Output in LLMs
Ask a large language model for a JSON object and you'll often get a broken one. It might look right, but the data types will be wrong or key fields...
Google DeepMind uses MITRE ATT&CK to monitor AI agents as rogue employees
Google DeepMind has a new security headache, and its source isn't human. The unit is now surveilling its own advanced AI agents as potential insider...
Audit AI tools: inventory, VPN/zero-trust, continuous fingerprinting
Your AI dev tools are already exposed, logins disabled, doors wide open. A nation-state group is exploiting this flaw right now.
Adobe adds AI Assistant to Photoshop, Premiere, Illustrator, InDesign, Frame.io
Adobe is putting you in charge. Starting this year, a new class of AI "creative agents" will report for duty inside Photoshop, Premiere Pro, and the...
Claude Fable (Mythos) 5 shows limited bug‑finding and refactoring aid
You can clean a codebase with one AI model and think the job is done. Then you can watch a different model waltz in and find a dozen fresh disasters.
Bernie Sanders proposes USD 7 trillion plan for U.S. public control of AI
Senator Bernie Sanders unveiled a $7 trillion plan on Tuesday. Its goal: to wrestle artificial intelligence development away from Silicon Valley and...
Helion adopts LFBO with on‑the‑fly Random Forest for autotuning
Helion's autotuner just got faster, but it's still boring work. Every kernel written in PyTorch's low-level language must be prodded and poked to...