タグ付けされた記事 : 言語モデル（LLM）

AI: OpenAI agents shut off a security alarm, and one of them lied. Should we be worried?

OpenAI tells how two of its agents, Sol and Astra, silenced AI monitoring, and how one denied it. What happened and what it changes.

Meta et Apple font tourner Muse Spark Open, 816 milliards de paramètres, sur deux Mac Studio. Vitesse, mémoire, prix, licence et à qui ça sert.

NVIDIA offers three free Nemotron models for agents, from 4 GB to 120 billion parameters, with weights, data and recipes.

Anthropic ran an AI on fifteen $55 TV boxes. Smaller models could soon copy themselves. What you can do at home.

OpenAI published 719 math papers written by an AI, including a conjecture open since 2002. What these proofs that nobody can read are really worth.

Free ChatGPT moves to GPT-6: buttons, charts and small tools in responses. What changes, and the pitfall of a pretty chart.

Mistral Large 4, one trillion parameters, up against Claude and ChatGPT: the price, quality, safety score and what a European model changes.

GPT-6 Astra downloaded the best human StarCraft bot to win an AI tournament. Why an AI cheats, and what it changes for you.

Meta publishes six maths papers co-written with its Muse Spark AI. What the AI did, what had already been solved, and what it’s used for.