5 articles · Last covered
In a new interview, the Meta Chief AI Scientist and AMI Labs founder doubles down on world models, arguing that the massive data requirements of current humanoid demos prove the Vision-Language-Action (VLA) approach is a dead end.
A new research paper co-authored by Yann LeCun introduces LeWorldModel (LeWM), a streamlined, stable architecture that trains on a single GPU and plans nearly 50 times faster than foundation-model alternatives.
With a massive seed round and a valuation of $3.5 billion, AMI Labs is building a global research powerhouse to prove that "world models" are the true path to human-level intelligence.
Meta’s former AI chief is taking his 'contrarian' bet to Paris, founding AMI Labs to build non-generative world models. Between a rumored $3.5 billion valuation and high-profile spats with Elon Musk, LeCun is betting that the path to true robotic intelligence doesn't start with language.
Speaking at MIT, Yann LeCun argued the "big secret" of the humanoid industry is that companies lack the fundamental AI breakthroughs—specifically "world models"—needed to make robots "generally useful" in domestic settings.
