Godfather of AI Yann LeCun tells tech industry everything it knows about AI is wrong

Yann LeCun, widely regarded as the godfather of artificial intelligence and Meta’s former chief AI scientist, has issued a stark warning to the technology industry, stating that large language models, the dominant foundation of today’s AI boom, are a dead end when it comes to achieving superintelligence.

In an interview with the Financial Times, LeCun said large language models are useful but fundamentally limited because they are constrained by language and lack an understanding of the physical world. He stated that achieving human-level or superhuman intelligence requires systems that can model how the real world works, rather than relying solely on text-based data, and added that LLMs are effectively a dead end in the pursuit of superintelligence, according to the report.

This is not the first time LeCun has criticised large language models. The veteran researcher has repeatedly spoken against the idea that LLMs, even at massive scale, can lead to superhuman intelligence, and has been vocal about what he sees as their structural limitations.

LeCun has instead argued for a different approach centred on what he calls world models, a class of AI systems designed to learn from physical reality rather than language alone. His preferred architecture, known as V-JEPA, aims to build internal representations of the world by learning from videos and spatial data. These systems, he has said, are capable of planning, reasoning and maintaining persistent memory, forming the basis of what he refers to as Advanced Machine Intelligence, or AMI.

While LeCun did not explicitly state the reason for his departure from Meta, he said that remaining at the company had become politically difficult. He noted that Meta chief executive Mark Zuckerberg was supportive of his world model research, but said the teams hired for the company’s new superintelligence push were heavily focused on large language models, describing them as being entirely LLM-centric, as reported by the Financial Times.

LeCun was referring to the cohort that joined Meta from Scale AI. In June 2025, Zuckerberg announced the formation of Meta Superintelligence Labs, a new division focused on developing advanced AI systems. The initiative is led by senior executives including former Scale AI chief executive Alexandr Wang. Earlier that month, Meta said it would bring Wang on as its chief AI officer along with several of his colleagues as part of a $14.3 billion investment into Scale AI, according to news reports.