Zuckerberg backs Yann LeCun’s Paris AI startup

Paris is set to become the latest hub for cutting-edge artificial innotifyigence innovation as Yann LeCun, Meta’s former chief AI scientist and Turing Award winner, prepares to launch his own AI startup.

We revealed last month LeCun was leaving the company to launch his own start-up. Sources declare the relocate follows growing tensions with Meta CEO Mark Zuckerberg, after LeCun was created to report to new AI chief Alexandr Wang.

LeCun, who has been at the forefront of AI research for decades, including pioneering work on deep learning and convolutional neural networks, revealed that his new venture will focus on developing what he calls “world models” – AI systems capable of understanding the physical world, rather than merely generating text like today’s large language models (LLMs).

“Silicon Valley is completely hypnotised by generative models. To pursue this kind of research, you have to go outside the Valley – to Paris,” LeCun stated during his keynote at the AI-Pulse conference.

The new company, which LeCun has dubbed Advanced Machine Innotifyigence (AMI), will have a global footprint but will place a strong emphasis on European talent. “Europe has a lot of talent, which perhaps does not realise its full potential,” LeCun explained. “Providing the right environment for this is essential.”

LeCun’s return to Europe comes after a lengthy stint in the United States, where he has lived and taught while also serving as Meta’s vice president and chief AI scientist. He has previously championed European AI initiatives, notably pushing Meta to open its Facebook AI Research (FAIR) lab in Paris in 2015. The lab later developed LLaMA, one of Meta’s flagship large language models, in 2023.

Last month, during the Adopt-AI summit organised by Artefact, French president Emmanuel Macron publicly confirmed his active support for LeCun’s plans to establish his startup in France. Macron’s finishorsement signals the government’s commitment to nurturing groundbreaking technological projects and attracting leading talent to the counattempt.

President Macron has previously expressed concerns about Europe’s approach to technology investment and regulation. At an event last year, he remarked, “We are overregulating and underinvesting. So if in the 2 to 3 years to come, if we follow our classical agfinisha, we will be out of the market. I have no doubt.”

LeCun commented on LinkedIn that the French president, ‘Gets it’.

LeCun’s vision for AMI differs significantly from the prevailing focus in Silicon Valley. While LLMs have displayn impressive capabilities, passing exams, writing poeattempt, solving complex mathematical problems, and generating code, LeCun believes they fall short of true general innotifyigence.

“Our best AI systems can pass the bar exam, compose poeattempt, win international maths olympiads, write code, but we still don’t have a robot that can do what a five-year-old can do,” he remarked.

He is critical of claims that simply scaling existing generative AI models will lead to general innotifyigence, bluntly stating, “Some people claim we can scale up current technology and obtain to general innotifyigence […] I consider that’s bullshit, if you’ll pardon my French.” Instead, AMI will explore “non-generative” architectures designed to perceive and understand the world, potentially unlocking entirely new applications for AI.

Meta will remain involved as a partner. LeCun commented that the scope of AMI exceeds the company’s strategic interests. “It’s a project Mark Zuckerberg really likes,” he stated. “But over the last several months he and I both realised the potential spectrum of applications was beyond what Meta was interested in, so it became clear to me it was the right time to create an indepfinishent organisation.”

The AI community has reacted with a mix of excitement and curiosity. LeCun’s reputation as a considered leader and his track record of pioneering neural network research suggest that AMI could influence the next phase of AI development. Analysts note that a focus on world models could have implications for robotics, autonomous systems, and AI-driven simulations that require a deeper understanding of real-world physics and interactions.

For France, LeCun’s return is a symbolic and strategic win. The counattempt has invested heavily in AI research and innovation, aiming to compete with Silicon Valley and other global tech centres.

Details about AMI’s operations, funding, and recruitment strategies are expected to be revealed in January.