Yann LeCun's New Venture, AMI Labs, Secures Over $1 Billion to Develop "World Models"

Yann LeCun's New Venture, AMI Labs, Secures Over $1 Billion to Develop "World Models"

AMI Labs, a new company co-founded by the renowned AI researcher Yann LeCun, has announced a staggering $1.03 billion seed funding round, valuing the company at $3.5 billion pre-money. This marks one of the largest seed rounds in European history and signals a significant bet on a new AI paradigm known as "world models," a departure from the currently dominant Large Language Models (LLMs).

Led by a team of top AI talent, with LeCun as chairman and serial entrepreneur Alexandre LeBrun as CEO, AMI Labs has assembled a formidable force. The company's mission is to overcome the limitations of LLMs, which rely on learning from vast amounts of text data. Instead, AMI Labs aims to build AI that can observe, understand, and predict the workings of the real world—in essence, creating "world models." This approach could enable AI to internalize physical laws and common sense, potentially solving fundamental issues like the "hallucination" problem (the tendency of LLMs to generate plausible-sounding falsehoods).

The record-breaking funding round reflects the market's high expectations for the ambitious vision of world models. The list of investors includes prominent venture capital firms like Cathay Innovation, Greycroft, and Bezos Expeditions, as well as corporate giants such as NVIDIA, Samsung, and Toyota Ventures, indicating strong interest from the industrial sector. The raised capital will be used to secure the immense computational resources necessary for AI development and to recruit top-tier researchers and engineers. The company is establishing a global R&D presence with offices in Paris, New York, Montreal, and Singapore.

At the core of AMI Labs' technology is the Joint Embedding Predictive Architecture (JEPA), a concept proposed by LeCun in 2022. This architecture learns the structure of the world and the relationships between data points through self-supervised learning by predicting missing information from observed data (e.g., predicting surrounding frames in a video from a single frame). This technology is expected to have applications in fields where interaction with the real world is crucial, such as healthcare, manufacturing, and robotics.

AMI Labs' endeavor represents a significant shift in the landscape of AI research and development. Its long-term approach, starting with fundamental research that may take years to commercialize, stands in stark contrast to many AI startups focused on short-term profitability. Furthermore, the company plans to actively publish its research findings and open-source its code, aiming to accelerate the overall progress of world model research by fostering an open ecosystem. The world is watching to see if this grand experiment can unlock a new frontier for artificial intelligence.