What is Yann LeCun planning with his startup AMI Labs?

He is considered one of the godfathers of artificial intelligence, he left Meta last November, he has just created a unicorn with his first fundraising round, and he is French: let's take a look at Yann LeCun's ultra-ambitious project.

After studying at the Sorbonne and then in Canada before moving to the United States, Yann LeCun became interested in 'deep learning' at the very beginning of his career, a branch of AI based on neural networks capable of learning from large amounts of data. In 2013, he joined Meta, where he became Chief AI Scientist and took charge of the group's artificial intelligence research. In 2018, along with two other researchers, he received the Alan Turing Award (the Nobel Prize of computing) for his work on deep neural networks.

Last November, he left Meta on good terms with Mark Zuckerberg despite differences over their respective ambitions for the future of AI. Since then, he has been distilling information about a particularly ambitious project, to the point of attracting the attention of the biggest names in global tech. All this led him to a $1bn fundraising round, even though he was initially aiming for half that amount.

Advanced Machine Intelligence, a deliberate breakthrough

With a team that includes leading profiles from the research world, Yann LeCun is advocating for a break from the dominant trajectory of AI so far.

Contrary to the current craze for Large Language Models (LLMs), which, although rich in knowledge, in his opinion, show intellectual weakness, he is betting on another approach: 'World Models'. Hie aims to make machines truly autonomous, capable of reasoning, memorizing, and planning actions in real and complex systems, while maintaining a strong commitment to open-source development.

According to him, language models alone will never achieve intelligence close to that of humans. He believes the path forward first requires an understanding of the physical world.

A different architecture, but also a different development model

Beyond a simple technological bet, Yann LeCun also defends another way of developing AI.

The goal of AMI is to learn the laws of physics from videos and spatial data, in order to allow the machine to perceive reality better than any current system. The challenge is to bring about an intelligence capable of evolving and acting in the real world.

The envisioned applications are numerous. Yann LeCun and his team have already mentioned opportunities in heavy and demanding sectors, such as robotics, industry, transport, or healthcare, but also, eventually, in domestic robotics. Specifically, he has, for example, mentioned the possibility of AMI modeling an aircraft engine and collaborating with a manufacturer to optimize its efficiency, based on precise industrial issues.

To launch AMI Labs, Yann LeCun has surrounded himself with former Meta employees. These include Michael Rabbat, former head of scientific research, Laurent Solly, former vice-president of the group, and Pascale Fung, former head of AI research.

The table of co-founders also includes Saining Xie, a former Google DeepMind researcher, called upon to lead the scientific strategy, as well as Alexandre Lebrun, former CEO of Nabla, who will lead the company. The first partner of AMI will also be Nabla, an AI applied to healthcare that serves as a clinical assistant.

A record fundraising for a project still in the experimental stage

The young startup has above all made an impression by signing the largest first funding round in European history. AMI raised $1bn for a valuation now estimated at $3.5bn, even though the project is still at the stage of scientific experimentation.

About twenty investors participated in the operation. On the tech side, we find Nvidia, former Google CEO Eric Schmidt, as well as Jeff Bezos' fund. Several European funds are also present, as well as a particularly visible French contingent with Bpifrance, but also some of the largest fortunes in France, including Xavier Niel, the Mulliez family, Dassault, Arnault and Saadé.

Asia is also represented, with Toyota Ventures, Samsung, the Singaporean fund Temasek, and a fund linked to SoftBank.

The next challenge: transforming a scientific vision into a global company

The goal is now to recruit to bring this ambition to life. AMI still needs to expand its Paris headquarters, but also its offices in Singapore, New York, and Montreal, with engineers and researchers. Although they are few in number, sought-after brains have undoubtedly paid attention to this project.