Meta loses AI ‘godfather’ Yann LeCun as he quits to launch new firm

Yann LeCun, one of the so-called “godfathers of AI,” is leaving Meta after 12 years to build a new company focused on what he calls “advanced machine intelligence,” the BBC reported. His exit marks a sharp philosophical split within the sector over the future direction of artificial intelligence.

LeCun announced his departure after days of speculation, saying on Threads that he planned to step down as Meta’s chief AI scientist while maintaining a partnership with the company. He thanked founder Mark Zuckerberg and described the creation of Meta’s Fundamental AI Research (FAIR) lab as his “proudest non-technical accomplishment.”

During his tenure, LeCun earned the Turing Award and helped shape Meta’s early investment in machine learning. He also witnessed the industry’s shifting tides, from early enthusiasm for deep learning to the surge in generative AI that followed OpenAI’s launch of ChatGPT in late 2022. Yet he has increasingly diverged from the dominant approach driving the boom.

The BBC reported that LeCun has been openly critical of large language models, the technology underpinning the current wave of AI chatbots and image generators. While Meta—like most of the industry—has channelled vast resources into LLM development, LeCun argues the systems cannot deliver human-level intelligence. Instead, he plans to pursue visual-first learning methods that mimic how children and young animals interpret the world.

His departure comes as some investors warn of an overheating AI market. The BBC noted that analysts and executives, including Google chief executive Sundar Pichai, have cautioned that a correction could ripple across the global economy if the “AI bubble” bursts.

LeCun has also set himself apart from fellow AI pioneers Geoffrey Hinton and Yoshua Bengio, who have raised alarms about existential AI risks. He has repeatedly rejected those fears, calling them “preposterously ridiculous” in a BBC interview last year. “Will AI take over the world? No,” he said. “This is a projection of human nature on machines.”

Not all experts view LeCun as a lone visionary. AI researcher Gary Marcus wrote in a recent blog that while LeCun’s technical contributions are significant, he has “systematically dismissed and ignored the work of others for years,” including critics who question deep learning’s long-term limits.

As he steps away from Meta’s formal hierarchy, LeCun’s next venture is set to intensify a defining argument in AI: whether the future belongs to ever-larger language models or to new architectures that rethink how machines perceive and learn. The outcome could shape the trajectory of a trillion-dollar industry.