AMI, the startup founded by a former Meta AI chief, raised about $1.03 billion at a $3.5 billion pre-money valuation to commercialize AI systems built around reasoning, planning, and world models.

The fundraising drew participation from a mix of European, American, and Asian investors, including Cathay Innovation, Hiro Capital, Dassault, Mulliez, Xavier Niel, Nvidia, Jeff Bezos’s fund, Toyota’s group, and Samsung.

AMI’s approach contrasts with large language models that predict text probabilities, highlighting weaknesses in real-world physics and long-term planning, especially for robotics and industrial use.

LeCun’s vision centers on AI that can simulate and reason about complex systems—such as human cells, power plants, and airplanes—to enable planning and control in industry and science, beyond consumer applications.

The JEPA-based architecture aims to learn abstract representations and predict outcomes in latent space, reducing reliance on next-token predictions and potentially curbing hallucinations on the path toward AGI.

LeCun has argued against the hype around LLMs and advocates shifting toward models that integrate with the physical world.

In the near term, AMI envisions consumer-facing uses like a domestic robot with common-sense physical understanding and potential deployment in Meta’s Ray-Ban smart glasses as a shorter-term application.

Early conversations also cover potential consumer applications and the option to deploy AMI technology in Meta’s smart glasses ecosystem.

Industry observers are debating whether world models can redefine AI progress and whether Europe can secure a leadership role in this emerging paradigm.

AMI’s CEO outlined a vision for world models that learn from real-world data to understand reality and surpass the limitations of current LLMs in sectors like healthcare.

AMI Labs is actively recruiting engineers, scientists, and researchers across four hubs in Paris, New York, Montreal, and Singapore to push its long-term research agenda.

There are discussions about collaboration or deployment with Meta, including exploring Ray-Ban glasses as part of AMI’s shorter-term applications.