"We don't have robots that are nearly as good at understanding the physical world as a rat," says Yann LeCun, one of the leading figures in artificial intelligence. After a decade as chief AI scientist at Facebook-owner Meta, LeCun left in 2025 and founded Advanced Machine Intelligence Labs (AMI Labs) in Paris. His goal: to move AI beyond the large language models (LLMs) that power systems like ChatGPT, Claude and Gemini.
Investors are betting big. Earlier this year AMI Labs announced it had raised more than $1bn (£760m) in seed funding, one of the largest such rounds in Europe. Backers include US computer chip giant Nvidia and the fund that manages the private wealth of Amazon-founder Jeff Bezos.
“Yann LeCun says LLMs are 'not smart' and launches AMI Labs with $1bn to build a more flexible AI.”
LeCun argues that LLMs, for all their prowess at coding and generating text, are fundamentally limited. "They're not a path towards human level or human-like intelligence, or even animal-like intelligence, because they cannot deal with real world data, they just are not built for that," he says on the sidelines of VivaTech, France's leading technology conference. "They [LLMs] basically just accumulate knowledge... They can regurgitate something, you train them to regurgitate, but they're not particularly smart. They don't have an underlying understanding."
To illustrate, LeCun holds a pen upright on its tip. What happens when you let go? Even a toddler knows the pen will topple over, but no human would try to guess which direction it falls – there's no way to tell. An LLM, he says, would attempt to generate a single prediction based on statistical patterns from its training data. That prediction would almost certainly be wrong because the system is not reasoning about physical reality, only generating statistically plausible output.
AMI Labs is developing a new system called Joint Embedding Predictive Architecture (JEPA) to overcome this. JEPA creates abstractions of the real world that allow it to assess the outcomes of actions, filtering out useless information. It is designed to handle the bewildering array of outcomes in the real world – problems that LLMs cannot tackle, like getting a robot to do household chores.
LeCun's bet is that JEPA can deliver the flexible, grounded intelligence that today's AI lacks. Whether it succeeds remains an open question, but the money – and the urgency – are unmistakable.