Yann LeCun's AMI Labs raises $1B to build JEPA, a smarter AI beyond LLMs

Yann LeCun, one of AI's most influential figures, has left Meta after a decade as chief AI scientist to launch Advanced Machine Intelligence Labs (AMI Labs) in Paris. At VivaTech, he told the BBC that mainstream systems like ChatGPT, Claude, and Gemini are fundamentally limited: "They're not a path towards human-level or even animal-like intelligence." His new venture has already raised more than $1 billion in seed funding—one of the largest rounds of its kind in Europe—with backing from Nvidia and the investment fund of Amazon founder Jeff Bezos. AMI Labs is now building a new type of AI called Joint Embedding Predictive Architecture (JEPA), designed to create abstractions of the physical world that allow the system to assess the outcomes of actions.

LeCun argues that LLMs just "regurgitate" statistical patterns from training data rather than truly understanding reality. He demonstrated the problem with a pen: ask an LLM what happens when you drop a pen held upright, and it will try to predict the exact direction of the fall—an impossible guess. JEPA, by contrast, filters out useless information and focuses on what matters, understanding that the pen will fall without needing to know which way. This flexibility is crucial for robotics, where current AI models are "largely hopeless," according to LeCun. He dismisses the idea that scaling up LLMs will lead to superhuman intelligence. Oxford professor Ingmar Posner agrees, saying the next decade of AI will be about building systems that explain cause, effect, and alternative actions—exactly the gap JEPA aims to fill.