Yann LeCun’s AI Lab Raises $1B for World Models

A new artificial intelligence research lab, co-founded by Turing Award recipient Yann LeCun following his departure from Meta, has secured a massive $1.03 billion in funding. AMI Labs is focusing its efforts on developing “world models,” a form of AI designed to learn from the physical reality of the world rather than solely from text and language data. This substantial financial backing, which values the company at $3.5 billion before the investment, signals strong investor confidence in this emerging field of AI research.

The CEO of AMI Labs, Alexandre LeBrun, anticipates a surge of interest in this approach. He predicts that “world models” will soon become the next major industry buzzword, with numerous companies adopting the label to attract investment. However, LeBrun emphasizes that his company’s mission is fundamentally distinct: to create AI that genuinely comprehends how the real world operates. This foundational understanding could unlock transformative applications, beginning with a partnership in digital healthcare.

LeBrun’s perspective is informed by his dual role as chairman of the digital health startup Nabla, which will be AMI Labs’ first official partner. In healthcare, the tendency of current large language models to generate incorrect information, known as hallucinations, carries serious risks. This limitation drove both LeBrun and LeCun to pursue an alternative path. Their work is based on a theoretical framework called the Joint Embedding Predictive Architecture (JEPA), proposed by LeCun in 2022. Translating this theory into practical, commercial products, however, is a long-term endeavor that may take years, distinguishing AMI Labs from typical fast-moving AI startups.

Despite the lengthy development timeline, investor appetite for world models is remarkably robust. AMI Labs is not alone; other ventures like SpAItial and Fei-Fei Li’s World Labs have also attracted significant capital. The French lab’s final raise far exceeded initial rumors, a feat LeBrun attributes to the strength of its founding team. The group includes high-profile figures such as Meta’s VP for Europe, Laurent Solly, as COO, alongside renowned researchers Saining Xie, Pascale Fung, and Michael Rabbat in key scientific roles.

This high level of interest allowed AMI Labs to be selective with its investors, ensuring alignment in vision and expectations. The funding round was co-led by several prominent venture firms including Cathay Innovation and Bezos Expeditions, with participation from a notable list of individual backers like Tim Berners-Lee and Eric Schmidt. The capital will primarily fuel two major expenses: advanced computing power and top-tier research talent.

The company plans to build its team strategically across four global hubs: Paris, New York, Montreal, and Singapore. This geographic spread aims to attract the best AI researchers and position the lab close to future clients in key markets. Although revenue generation is not an immediate goal, AMI Labs intends to engage with potential partners early in the development process. LeBrun stresses that authentic world models cannot be built in isolation; they must eventually be tested with real-world data and evaluations.

Nabla is the first disclosed partner expecting access to these early models, but more collaborations are expected. This commercial potential likely explains the participation of several strategic corporate investors in the funding round, such as NVIDIA, Samsung, Toyota Ventures, and French industrial groups. True to its academic roots and LeCun’s philosophy, AMI Labs commits to publishing its research findings and releasing a significant portion of its code as open source. The founders believe this open approach accelerates progress and helps build a vital research community around their ambitious project.

(Source: TechCrunch)