In a massive vote of confidence for a less-traveled path in artificial intelligence, AMI Labs has raised $1.03 billion. The startup, co-founded by Turing Award laureate Yann LeCun after his departure from Meta, is now valued at $3.5 billion. Its mission is to build 'world models'—systems that learn from reality itself, not just text.

CEO Alexandre LeBrun, who also chairs digital health company Nabla, notes the field is heating up. 'My prediction is that ‘world models’ will be the next buzzword,' he told TechCrunch. 'In six months, every company will call itself a world model to raise funding.' He believes AMI is distinct, aiming for a fundamental understanding of the physical world, starting with a healthcare partnership with Nabla. The move stems from recognizing the risks of hallucinations in large language models for critical applications.

The company's approach, based on LeCun's 2022 Joint Embedding Predictive Architecture (JEPA), is a long-term research play. 'It’s not your typical applied AI startup that can release a product in three months,' LeBrun said. This hasn't deterred investors. The round was co-led by Cathay Innovation, Greycroft, and Bezos Expeditions, among others, with backing from NVIDIA, Toyota Ventures, and individuals like Tim Berners-Lee and Mark Cuban.

The capital will fund compute power and a carefully built team across Paris, New York, Montreal, and Singapore. While revenue is years away, AMI plans early engagement with partners like Nabla to test models in real-world scenarios. In a departure from many modern AI labs, AMI commits to publishing its research and releasing open-source code. 'We think things move faster when they’re open,' said LeBrun, a veteran of Meta's FAIR lab. The bet is that understanding the world requires building in the open, too.

Source: TechCrunch