Yann LeCun, a prominent leader in today's artificial intelligence (AI) field, has told associates he intends to depart Meta in the next few months to establish his own company, according to the Financial Ti mes, which referenced individuals aware of the discussions. The exit of LeCun, a Turing Award recipient and trailblazing researcher, would signify a major shift for Meta and the wider AI sector, as he moves to advance his concepts for future AI architectures.
TL;DR
- Yann LeCun, a professor at NYU and AI pioneer, plans to leave Meta to start his own AI venture.
- LeCun, known for convolutional neural networks, is reportedly seeking funding for "world models" AI.
- His departure follows Meta's AI reorganization and a strategic pivot towards large language models.
- This move signals a disagreement on the future direction of AI development and research.
LeCun, aged 65, became part of Facebook in December 2013, establishing the Fundamental AI Research initiative, also referred to as FAIR. He continues to hold a position as a Silver Professor at New York University, a role he's held since 2003.
His academic credentials are formidable: LeCun is widely recognized for his pioneering work on convolutional neural networks during the late 1980s, particularly the LeNet architecture, which demonstrated remarkable success in identifying handwritten digits and significantly advanced the field of computer vision. In 2019, he was recognized received the ACM Turing Award with Geoffrey Hinton and Yoshua Bengio for pioneering advancements that established deep neural networks as a vital element in contemporary computing.
A childhood fascination with mechanical devices
Born in Soisy-sous-Montmorency, France, on July 8, 1960, LeCun grew up with an engineer father who encouraged his fascination with electronics. His early interest led him to ESIEE Paris led him to obtain an electrical engineering degree in 1983. Following that, he earned a PhD in computer science from Université Pierre et Marie Curie, finishing his thesis in 1987 concerning connectionist learning models, research that proposed an early form of the backpropagation algorithm for the training of neural networks.
When neural networks were considered unfeasible, LeCun undertook a postdoctoral year alongside Geoffrey Hinton at the University of Toronto, subsequently joining AT&T Bell Labs in 1988. At that location, he pioneered convolutional neural networks, a significant advancement enabling computers to interpret visual data similarly to how humans do. His system for reading handwritten digits proved so successful that NCR began using it in automated check-reading systems around the mid-1990s, handling between 10% and 20% of all U.S. Checks at its peak.
LeCun also spearheaded the creation of DjVu, a technology for compressing images that allowed the Internet Archive and other digital libraries to make scanned documents available on the web. Following a short period at NEC Research Institute, he became affiliated with New York University.
Departing Meta
The reported exit from Meta occurs while the parent company of Facebook is implementing significant shifts in its AI approach. In June, the company invested $14.3 billion in data-labeling firm Scale AI and brought on its 28-year-old CEO, Alexandr Wang, to lead a new division called Meta Superintelligence Labs. The reorganization shifted LeCun’s reporting structure: He previously reported to Chris Cox, Meta’s chief product officer, but reported to Wang afterward.
The structural change reflects a deeper strategic divide. CEO Mark Zuckerberg has pivoted toward rapid deployment of large language models and AI products, particularly after Meta’s Llama 4 model fell short of expectations and lagged behind competitors such as OpenAI and Google. LeCun, however, has been publicly skeptical of large language models, arguing they will never achieve human-level reasoning and planning capabilities.
According to the FT, LeCun is in early discussions to raise funding for a startup focused on what he calls “world models”—AI systems that develop an internal understanding of their environment by learning from video and spatial data rather than relying solely on text. He’s previously said such systems, which aim to simulate cause-and-effect scenarios and predict outcomes, may take about a decade to mature.
The shift at Meta has not been without friction. Multiple former employees told Coins2Day‘s Sharon Goldman earlier this year that FAIR has been “dying a slow death” as the company prioritized commercially focused AI teams over long-term research. More than half the authors of the original Llama research paper left Meta within months of its publication. In October, Meta cut approximately 600 positions from its AI division. So while LeCun’s planned move is a significant personnel change, it also signals a fundamental disagreement about the path to AGI and the role of research in an industry increasingly driven by competitive product timelines.