Yann LeCun, known as one of the godfathers of modern artificial intelligence and one of the first AI visionaries to join the company then known as Facebook, is leaving the company. meta.
In a LinkedIn post on Wednesday, LuCun said he plans to found a startup focused on a type of AI technology that researchers describe as world models, which analyze information beyond web data to better represent the physical world and its properties.
“I am founding a startup to continue the Advanced Machine Intelligence Research Program (AMI) that I have been pursuing for the past several years with colleagues at FAIR, NYU, and other universities,” LeCun wrote. “The startup’s goal is to bring about the next big revolution in AI: systems that understand the physical world, have persistent memory, can reason, and can plan complex courses of action.”
Meta will partner with LeCun’s startup.
The departure comes at a time of turmoil within Meta’s AI division. The division underwent a major overhaul this year after the company released the fourth version of its open source large-scale language model, Llama, following a disappointing response from developers. That spurred CEO Mark Zuckerberg to spend billions of dollars recruiting top AI talent, including investing $14.5 billion in Scale AI in June to bring in Alexander Wang, the company’s 28-year-old CEO and now Meta’s new chief AI officer.
Mr. LeCun, 65, joined Facebook in 2013 as director of FAIR AI research while maintaining an adjunct professorship at New York University. “Founding FAIR is my proudest non-technical accomplishment,” he said in a LinkedIn post.
“We are extremely grateful to Mark Zuckerberg, Andrew Bosworth, Chris Cox, and Mike Schropfer for their support of FAIR and the AMI program over the past several years,” LeCun said. “Their continued interest and support will make Meta a partner in the new company.”
At that time, Facebook google was hiring a large number of high-level academics like LeCun to spearhead efforts to produce cutting-edge computer science research that could benefit the company’s core business and products.
LeCun, along with other AI luminaries like Yoshua Bengio and Geoffrey Hinton, have put a type of AI technique known as deep learning at the center of their academic research. Deep learning involves training large software systems called neural networks to discover patterns within a set of data. The researchers helped popularize deep learning approaches and won the prestigious Turing Award from the Association for Computing Machinery in 2019.
Since then, LeCun’s approach to AI development has strayed from the direction taken by Meta and other Silicon Valley companies.
Technology companies like Meta and OpenAI have spent billions of dollars developing so-called foundational models, specifically LLMs, as part of their efforts to advance cutting-edge computing. But LeCun and other deep learning experts say that while powerful, current AI models have a limited understanding of the world and that new computing architectures are needed for researchers to create software that can match or surpass humans at certain tasks, a concept known as artificial general intelligence.
“My imagination is that AMI will have widespread applications in many sectors of the economy, some of which overlap with meta’s commercial interests, but many of which do not,” LeCun said in the post. “Pursuing AMI’s goals in an independent organization is a way to maximize its broad impact.”
In addition to Wang, recent high-profile names Zuckerberg has brought in to revamp Meta’s AI division include former GitHub CEO Nat Friedman, who leads the division’s product team, and Shengjia Zhao, co-developer of ChatGPT and the group’s chief scientist.
In October, Meta fired 600 employees from its Superintelligence Research Institute division, including some who had been part of the FAIR division that LeCun helped launch. These years of cuts, along with other reductions at FAIR and the creation of a new AI leadership team, played a large role in LeCun’s decision to step down, said the people, who requested anonymity because they were not authorized to speak publicly.
Additionally, Mr. LeCun had little interaction with Mr. Wang or the TBD Labs division, and Mr. Zuckerberg had been compromised in many of his headline-grabbing jobs over the summer. TBD Labs is overseeing the development of Meta’s Llama AI model, which was originally developed within FAIR, the people said.
While LeCun has always promoted sharing AI research and related technologies with the open source community, Wang and his team prefer a more closed approach amid intense competition from rivals such as OpenAI and Google, the people said.
Attention: The meta is table pounder here.
