Who’s behind AMI Labs, Yann LeCun’s “global model” startup.

Yann LeCun’s new venture, AMI Labs, has attracted a lot of attention since the AI ​​scientist left Meta to found it. This week, the startup finally confirmed what it’s building, and several key details were hidden in plain sight.

On its newly launched website, the startup revealed its plans to develop “universal models” to “build intelligent systems that understand the real world.” The focus on global models has already been hinted at by the name AMI, which stands for Advanced Machine Intelligence, but it has now officially joined the ranks of the hottest startups in AI research.

Building foundational models that connect AI and the real world has become one of the most exciting activities in the field, attracting leading scientists and wealthy investors alike – productive or non-productive.

World Labs, a direct competitor founded by AI pioneer Fei-Fei Li, became a unicorn shortly after coming out of stealth. Following the launch of its first product, Marble, which generates physically intact 3D worlds, World Labs is now reportedly in talks to raise $5 billion in new funding.

There is no doubt that venture capital firms would be equally keen to invest in LeCun, adding credence to rumors that AMI Labs may raise funding at a valuation of $3.5 billion. According to Bloomberg, venture capital firms in talks with the startup include Cathay Innovation, Greycroft and Hiro Capital, for which LeCun serves as an advisor. Other potential investors are said to include 20VC, Bpifrance, Daphni, and HV Capital.

Regardless of who is writing the checks, investors may want to note an important detail: As LeCun pointed out, he is AMI’s CEO, not its CEO. Instead, that role goes to Alex LeBrun, co-founder and former CEO of Nabla, a health AI startup with offices in Paris and New York.

LeBrun’s move from Nabla to AMI is part of a partnership announced last December by Nabla, which develops artificial intelligence assistants for clinical care and for which LeCun was a consultant. In exchange for “privileged access” to AMI’s global models, Nabla’s board of directors supported LeBron’s transition from CEO to chief AI scientist and chairman of the board, paving the way for his new role.

As CEO of AMI Labs, LeBron will be surrounded by familiar faces. After Facebook acquired his previous startup, Wit.ai, the serial entrepreneur and AI engineer worked under LeCun at Meta’s AI research lab, FAIR. According to reports, the duo will also be joined by Laurent Solly, who resigned as Meta’s vice president for Europe last December.

The talent overlap between AMI and Meta likely won’t stop there. LeCun told MIT Technology Review that his former employer could be AMI’s first client. But he also publicly criticized some of the strategic choices Meta made under Mark Zuckerberg. More broadly, the review interprets AMI Labs as a contrarian bet against large language models (LLMs).

The limitations of LLMs pointed out by LeCun include hallucinations, which are a serious concern in contexts such as medicine, as LeBrun also knows firsthand. The CEO of AMI Labs told Forbes that the main reason he took the position was to apply its global models to healthcare. But the startup will also target other high-risk application areas.

“AMI Labs will advance AI research and develop applications where reliability, controllability, and safety are of great importance, especially for industrial process control, automation, wearable devices, robotics, healthcare, and more,” she wrote in her mission statement. “We share one belief: True intelligence doesn’t start with language. It starts in the world.”

Unlike generative methods, which LeCun and his team see as unsuitable for unpredictable data like sensor inputs, the startup promises that its AI systems will not only understand the real world, but will also have a fixed memory, the ability to reason and plan, and will be controllable and safe.

The startup plans to license its technology to industry partners for real-life applications, but says it also plans to contribute to building the future of AI “with the global academic research community via open publications and open source.” LeCun said he plans to keep his position as a professor at NYU, where he teaches one class a year and supervises doctoral and postdoctoral students.

This means the French-born researcher will remain based in New York, but he told MIT Technology Review that AMI Labs “will be a global company.” [that’s] Its headquarters are in Paris.” French President Emmanuel Macron welcomed the news and expressed his pride that Lacon, also a Turing Award winner, had been selected for Paris. “We will do everything in our power to ensure his success from France,” he said.

The startup will also have offices in Montreal, New York and Singapore, but its decision to choose Paris as its headquarters will help cement Paris’s reputation as an AI hub, as it will join the ranks of H, Mistral AI and several international labs, including FAIR. Perhaps it’s appropriate for AMI to be pronounced a-mee — like the French word “ami,” which means “friend,” LeCun noted.