① 2019 Turing Award winner, hailed as the modern " artificial intelligence" One of the "godfathers" of Meta has officially announced that he will leave Meta by the end of the year; ② In his latest public appearance, LeCun said that Meta will not provide funding for his new artificial intelligence company and hinted that the company may be headquartered in Paris.
Yann LeCun, the 2019 Turing Award winner and hailed as one of the "godfathers of modern artificial intelligence ," has officially announced his departure from Meta Inc. by the end of the year. In his latest public appearance, LeCun stated that Meta will not provide funding for his new AI company and hinted that the company may be headquartered in Paris.
It's worth noting that when Yang Likun, Meta's chief AI scientist, announced his departure last month, he described Meta as a "partner" of his new company, but did not reveal any further details.
At the AI-Pulse conference, Europe’s top artificial intelligence conference, held in Paris on Thursday, he made it clear that Meta was not an investor in his new company.
It is understood that the unnamed startup will focus on the "Advanced Machine Intelligence" (AMI) research program that Yang Likun has been promoting for the past few years —a form of artificial intelligence that uses visual and other sensory information to predict the physical world.
This "world model" offers an alternative to current mainstream generative AI tools (such as large language models), aiming to overcome the limitations of network data and more accurately represent the physical world and its characteristics by analyzing a wider range of information.
When Yang Likun officially announced his departure from Meta last month, he wrote, "The goal of the new company is to drive the next major revolution in artificial intelligence: to create systems that can understand the physical world, have persistent memory, and can reason and plan complex sequences of actions."
He bluntly stated that Silicon Valley is going astray.
Yang Likun also said on Thursday, "Silicon Valley is completely hypnotized by generative models, so this kind of research (world modeling) must be carried out outside of Silicon Valley—for example, in Paris."
These remarks suggest Yang Li-Chun's growing dissatisfaction with the trend of tech companies like Meta focusing excessively on Large Language Models (LLMs). The core principle of LLMs is actually quite simple: predicting the next word in a sequence based on existing information. World models, on the other hand, aim to predict real-world events that might occur, such as the trajectories of objects or people, which is relevant in robotics. It has application value in physical AI systems such as autonomous driving.
“Our most advanced AI systems can pass the bar exam and write code, but to date, no robot can do what a five-year-old child can do. We’ve missed a crucial element,” Yang Likun said.
Yang Likun stated that more information about his startup will be released in January next year. He has only revealed so far that the technology will be applicable to robotics and industrial fields.
Despite the impending parting of ways with Meta, Yang Li-Chun emphasized that Meta founder and CEO Mark Zuckerberg greatly admired the project. "But over the past few months, we've both realized that the technology's applications have expanded beyond Meta's focus," he stated.
(Article source: CLS)
