LeCun: 'LLMs Are Not the Future' — AMI Labs Raises $1 Billion

AI pioneer Yann LeCun made his position clear again yesterday, delivering a lecture at Brown University. The message was unambiguous: large language models are not the path to human-level intelligence, and his new company AMI Labs is working to prove it.

AMI Labs, officially launched in March 2026, has raised over $1 billion in seed funding and is now valued at $3.5 billion, making it the largest seed round in European history.

LeCun argues that today's AI "sucks" in its current LLM-centric form. The problem is fundamental: language models cannot understand the physical world, causality, or plan ahead in meaningful ways. The solution lies in "world models," systems that learn from diverse sensory data including images, video, audio, and scientific information, not just text.

His approach is embodied in the Joint Embedding Predictive Architecture, or JEPA. The system builds abstract representations of the physical world, reasons about cause and effect, and predicts future states based on imagined actions. The latest research advance from AMI Labs is LeWorldModel, a JEPA trained end-to-end directly from raw pixel data.

LeCun estimates human-level intelligence is still five years away, possibly more. But he is convinced the path there does not run through more LLM scaling. AMI Labs is now the heaviest bet in the industry that he is right.