AI News Archive - Browse Page 233 of 253
Browse AI news articles covering LLMs, tools, research, and industry trends
Terminal-Bench 2.0 launches with Harbor, testing any container-installable agent
Testing AI agents is a mess. Every team runs a different set of bespoke checks, and no one agrees on what passing means.
Teams tackle new prompt injection attacks, boost model mitigations
It is surprisingly easy to hijack a chatbot. Just ask it the right, wrong thing. Security teams are now scrambling to build better armor for large...
ChatLLM Review: All-in-One AI Suits Small Teams, Startups and Freelancers
Small companies and freelancers are being priced out of the AI race. Enterprise platforms cost a fortune, and cobbling together single-use tools eats...
Notion 3.0 uses GPT-5 to let AI run autonomous stakeholder feedback workflows
Every new software update claims to be radical. Notion’s 3.0 release might actually mean it.
Interactive AI Agent Uses OpenAI Function Schemas for Rapid ML Tasks
Most AI hype is just text. A model writes a paragraph, you read it, nothing actually happens. A new trick is changing that.
Huang says AI is industry backbone, while panelists like Li voice skepticism
Jensen Huang sells the machinery. At Davos, he told a room full of buyers that the machinery is everything.
ComputeEval 2025.2 expands to 232 CUDA challenges, upping LLM test difficulty
The standard tests for large language models used to be forgiving. Not anymore. ComputeEval 2025.2, NVIDIA’s new benchmark, just made them brutally...
Firebase Studio uses Code OSS editor to simplify full-stack app development
Firebase Studio is not a new IDE. It's a wrapper for an old one. Google's development platform has announced an integrated environment for building...
AI translation tools fell short as in-laws assumed 95% of Europeans speak English
The industry sold us a lie: AI would dissolve language barriers like sugar in water.
First Pandas & Scikit-learn Project Drops Rows with Missing Values
Every data science tutorial promises a clean path from raw data to working model. The reality is messier.
Microsoft, AWS, Adobe Seek Copyright Clarity for AI Training in India
Microsoft, AWS, and Adobe have their elbows on the table in New Delhi. India is rewriting its rulebook for the internet, and lobbyists for the tech...
Google's Ironwood TPU to be generally available on Cloud in weeks
Google Cloud will rent out its newest AI chip within weeks. The Ironwood TPU is a direct shot at Nvidia.
PyTorch and NVIDIA BioNeMo add attn_input_format for flash-attention scaling
Flash-attention scaling is a notoriously tricky problem. PyTorch and NVIDIA's BioNeMo think they've cracked a piece of it with a new parameter called...
PyTorch speeds MoE training on DGX H100 BF16 with NeMo Automodel
Big AI models are expensive beasts to feed. PyTorch just threw them a cheaper bag of feed.
OpenAI CEO Sam Altman projects USD 20 billion revenue, backs US chip push
Sam Altman says he’ll bring in twenty billion dollars this year. He also wants to build a new industrial base for it.
Microsoft gains OpenAI deal allowing independent AGI pursuit or partnerships
Microsoft now has a get-out-of-jail-free card for the biggest project in tech. Its new deal with OpenAI explicitly permits the company to go build a...
AI drives three initiatives to accelerate global action for planet protection
Every environmental crisis is a data crisis. We are buried in information about a dying planet, but we lack the means to understand it.
Gemini API adds File Search Tool supporting PDF, DOCX, TXT, JSON, code files
Google’s Gemini API can now read your clutter. A new feature lets it search across the usual suspects—PDFs, Word docs, plain text, JSON, and code...
BBVA scales AI to 11,000 staff, saving 3 hrs weekly per employee
BBVA gave 11,000 employees a new company tool, and they actually used it. This is not normal. Most corporate tech rollouts fail.
Moonshot's K2 Thinking tops open-source AI, beating GPT-5 and MiniMax-M2
Every open-source AI model claims to be on the verge of beating the big guys. Moonshot's K2 Thinking actually did it, by the numbers.