AMI Labs

In January 2025, Yann LeCun — Meta's chief AI scientist and one of the recognized fathers of deep learning — announced the founding of AMI Labs in Paris. This is not another startup building a larger language model than the last. AMI Labs is something more ambitious and more uncertain: a research laboratory targeting advanced artificial intelligence, the kind capable of reasoning, planning, and understanding the physical world as humans do. The project stems from a conviction LeCun has expressed publicly and repeatedly over recent years: large language models, impressive as they are, are not the path toward general artificial intelligence. They lack causal understanding, long-term planning capabilities, and a robust representation of the physical world. AMI Labs develops what researchers call world models — architectures capable of building internal representations of reality that go beyond predicting the next token. The laboratory's structure is oriented toward fundamental research: the team releases open-source code and collaborates with academic and industrial partners. There are no commercial products and no immediate monetization plans. LeCun has stated explicitly that the work will take years before finding practical applications. This positions AMI Labs closer to a research institute than a traditional startup, at least in its current phase. The decision to establish the laboratory in Europe — and in Paris specifically, which is actively trying to build an AI ecosystem competitive with Silicon Valley — carries significance. AMI Labs joins a constellation of initiatives seeking to anchor fundamental AI research in the continent, reducing dependence on large American laboratories. Future revenues, when they arrive, could come from algorithm licensing, industrial partnerships, and applications built on world model architectures. For now, AMI Labs is an intellectual bet on the future of AI, made by someone who helped build it.