Inside AMI Labs, Yann LeCun’s Big New AI Venture

The latest startup of AI researcher Yann LeCun, AMI Labs has already attracted attention from around the world. The company has now confirmed that the work is in fact on “world models,” a type of artificial intelligence that is built to comprehend and interact with the messy real world — not just text data.

The concept of World Models is to design systems that can learn from physical environments, remember past events, plan actions and take reliable decisions. This methodology separates AMI Labs from the bulk of current AI work, which is dominated by large language models.

The AMI Labs stand for Advanced Machine Intelligence, a name that was suggestive of its direction before any details were known. The start-up enters a competition of sorts in a fast-growing space that has already received heavy investment. One of MMT’s closest competitors, World Labs, which was founded by Fei-Fei Li and quickly became a unicorn after its launch, is reportedly targeting new financing at an even higher valuation.

AMI Labs is also attracting strong interest from investors. The company is in discussions with multiple venture firms, according to reports; some estimates placed the worth of the company at $3.5 billion but no funding round has been announced.

LeCun is the public face of the company, but he’s working as executive chairman, not chief executive. CEO is Alex LeBrun, a founder of former health AI company Nabla. LeBrun’s departure follows a collaboration between Nabla and AMI Labs, which provided Nabla with research from AMI while letting LeBrun step away to launch the new venture.

LeBrun and LeCun have worked together before at Meta’s AI research unit after Meta acquired LeBrun’s earlier startup, Wit.ai. Other ex-Meta leaders are expected to join AMI Labs, fueling speculation that Meta could end up being one of its first customers.

AMI Labs stands in contrast to language-first AI systems. LeCun has long warned about problems like hallucinations in language models, especially in domains such as health care. The start-up says its technology will prioritize safety, control and precision, with planned applications in medicine, robotics, industrial systems and wearable devices.

The company will look to license its technology with industry partners and publish research as well as contributing back to the open source community. LeCun will keep his day job at New York University, but AMI Labs will have a home in Paris, along with planned offices in Montreal, New York, and Singapore.

By deciding to base itself in Paris, AMI Labs gives momentum to the city’s burgeoning status as a global hub for AI R&D.