Sometimes people say that the LLMs make obvious errors because they don’t have a world model, but one is welcome to one’s guess as to what percentage of population has sufficiently detailed world model to discuss layers of defense against failure cases in a remotely operated…
RESEARCH
-
Open Weights Strategy: Shipping Lesser Models Ineffective Against DeepSeek
By
–
Great that they're doing open weights but the reason for shipping a lesser version seems silly (China's DeepSeek is already ahead) and reminiscent of some old (and ineffective) open source licensing strategies https://t.co/IcvtDarZsn
— Matt Asay (@mjasay) 5 mai 2025Great that they're doing open weights but the reason for shipping a lesser version seems silly (China's DeepSeek is already ahead) and reminiscent of some old (and ineffective) open source licensing strategies
-

Building Trust: Essential for Business AI Scaling
By
–
Can businesses scale AI without earning trust first? That’s the question I explored with Leanne Allen, Partner and Head of AI at @kpmguk
, in a powerful conversation on Trust in AI. A recent global @KPMG and University of Melbourne study found: Fewer than half of UK citizens -

Paper2Code: Multi-Agent System Transforms ML Papers into Code
By
–
Turn any ML paper into code repository! Paper2Code is a multi-agent LLM system that transforms a paper into a code repository. It follows a three-stage pipeline: planning, analysis, and code generation, each handled by specialized agents. 100% Open Source
-
Building Experimentation Culture for R&D Excellence
By
–
Sometimes that is doing what they won’t. Sometimes that is doing what they can’t. Sometimes it is building an experimentation machine and culture that will just churn out a large portion of the world’s R&D in your field specifically.
-
Grok 3.5 Benchmark Results Were Reportedly False
By
–
Il semblerait que le benchmark de grok 3.5 était faux ! Le véritable verdict aura lieu cette semaine normalement.
-
Anthropic Co-Founder Says DeepSeek Hype Overblown, 6-8 Months Behind
By
–
Le cofondateur d'Anthropic, Jack Clark, déclare que l'engouement autour de DeepSeek est "peut-être un peu exagéré".
— VISION IA (@vision_ia) 5 mai 2025
Ses modèles sont compétents mais accusent un retard de 6 à 8 mois par rapport aux laboratoires américains de pointe et ne constituent pas encore une préoccupation… pic.twitter.com/bE7VCbNJ1QLe cofondateur d'Anthropic, Jack Clark, déclare que l'engouement autour de DeepSeek est "peut-être un peu exagéré". Ses modèles sont compétents mais accusent un retard de 6 à 8 mois par rapport aux laboratoires américains de pointe et ne constituent pas encore une préoccupation
-

Google DeepMind CEO Warns Society Unprepared for AGI Arrival
By
–
"L'AGI arrive, la société n'est pas prête" Le PDG de Google DeepMind, Demis Hassabis, avertit que l'Intelligence Artificielle Générale (AGI) pourrait arriver dans 5 à 10 ans, mais la société reste non préparée à son impact transformateur. Il plaide pour une collaboration
-

DeepSeek Prover-V2 Merges Informal Reasoning and Theorem Proving
By
–
DeepSeek a dévoilé Prover-V2, une IA open source qui fusionne raisonnement mathématique informel et démonstration de théorèmes. Avec ses 671 milliards de paramètres, elle atteint un taux de réussite impressionnant de 88,9 % sur le benchmark MiniF2F. Particularité : elle
-

Alibaba Qwen Team Launches Qwen3 Series with MoE and Dense Models
By
–
L’équipe Qwen d’Alibaba a lancé la nouvelle série Qwen3, composée de 2 modèles MoE (Mixture of Experts) et 6 modèles denses. https://
youtu.be/GIQxA1610yE