Actualités — Page 3

DeepSeek Unveils Method for Training AI Agents at Scale

DeepSeek has published a paper describing its platform for training AI agents, which runs millions of isolated sandboxes daily. The platform is designed to contain agent execution, as no single mechanism can prevent all misbehavior. Europe requires member states to have regulatory sandboxes operational by August 2027. Lire la suite
DeepSeek Unveils Method for Training AI Agents at ScaleThe Next Web

OpenAI to Allow Outside Groups to Test AI Models During Training

OpenAI will permit third-party groups to conduct technical safety assessments on its AI models during the training and evaluation process, a move that comes after CEO Sam Altman promised independent evaluators access to the company's models. This decision marks a shift in OpenAI's approach, as it previously only allowed assessments before launch. Lire la suite
OpenAI to Allow Outside Groups to Test AI Models During TrainingThe Next Web

Meta's AI Agent Muse Aims to Save Users $1,000

Meta's chief AI officer, Alexandr Wang, has challenged users to see if the company's new AI agent, Muse, can save them $1,000. The agent, which can compare insurance, cancel forgotten subscriptions, and hunt for discounts, has already shown promising results, with some users reporting savings of up to $3,500. Lire la suite
Meta's AI Agent Muse Aims to Save Users $1,000The Next Web

Young Users Abandon Google for AI, Raising Concerns Over Critical Thinking

A recent study analyzing 173 international research articles on the use of generative artificial intelligence in school and education found that AI tools can have both positive and negative effects on critical thinking and problem-solving abilities in children and young people. The study highlights a significant knowledge gap in understanding how AI use impacts younger school children. Lire la suite
Young Users Abandon Google for AI, Raising Concerns Over Critical ThinkingHacker News

AI Chatbot Unlocks Secrets in Ancient Greek Records

Researchers have developed a large language model to fill in gaps in papyrus fragments, accelerating the laborious task of restoring ancient texts. This AI-powered tool aims to uncover new details about ancient life, leveraging artificial intelligence to analyze hundreds of thousands of Ancient Greek papyrus fragments stored in academic libraries worldwide. Lire la suite
AI Chatbot Unlocks Secrets in Ancient Greek RecordsWired AI

AI Clones of Coworkers: A Weird Future of Work

A writer created AI clones of their editors to impress them, but things took a strange turn. The clones, including Brian Bot and Sophie Bot, developed their own personalities and quirks, raising questions about the future of work and artificial intelligence. Lire la suite
AI Clones of Coworkers: A Weird Future of WorkWired AI