Meta’s chief AI scientist Yann LeCun is preparing to leave the company to establish a new AI startup focused on world models, prompting investor attention as Meta relies on his leadership for core artificial intelligence research.
The Financial Times, citing people familiar with the matter, reported that LeCun is in early discussions to raise capital for the venture, which would advance his longstanding research agenda on computational world models—systems that learn structured representations of the physical and conceptual environment to support prediction and autonomous decision-making. News of his prospective departure coincided with a 1.5% decline in Meta’s share price during pre-market trading, reflecting market scrutiny of changes to the company’s AI leadership.
LeCun’s standing within the field is rooted in work from the late 1980s and early 1990s, when he developed LeNet, among the first successful convolutional neural networks for handwritten-digit recognition. That architecture demonstrated how layered convolutional filters, shared weights, and pooling operations could extract hierarchical visual features, enabling robust automated reading of bank checks and postal codes. LeNet’s design directly informed modern convolutional neural networks that support large-scale image classification, facial recognition pipelines, and perception systems used in autonomous vehicles.
LeCun joined Facebook in December 2013 as founding director of Facebook AI Research (FAIR), tasked with building a dedicated fundamental research organization. In his current position as Meta’s chief AI scientist, he oversees long-horizon work in self-supervised learning algorithms that reduce reliance on labeled data, world-model research aimed at more general machine understanding, and architectures for autonomous AI systems that can learn, reason, and act under uncertainty across complex tasks.
Meta has linked its strategic direction to large-scale AI investment, stating that it plans to invest more than $600 billion in the United States by 2028, spanning AI technology development, data center and compute infrastructure, and workforce initiatives. In 2023, the company committed $14.3 billion to acquire a 49% stake in Scale AI. As part of this transaction, Scale AI’s former chief executive Alexandr Wang joined Meta and now holds a leadership role to whom LeCun reports, according to the FT account.
In October 2023, Meta eliminated around 600 positions within its AI organizations. The company described the move as an effort to streamline operations and reduce decision-making layers. The cuts affected FAIR and the AI product and infrastructure division, while TBD Labs—an elite unit within the Wang-led Meta Superintelligence Labs focused on next-generation foundation models—remained unaffected.
LeCun has consistently challenged claims that current generative AI poses catastrophic risks. He has characterized such warnings as “ridiculous” and “complete B.S.” and has expressed skepticism that large language models, in their present trajectory, will produce artificial general intelligence or AI superintelligence.
