Yann LeCun is taking a new swing at foundational AI with Advanced Machine Intelligence Labs, a startup he launched after leaving his role as Chief AI Scientist at Meta. The company is seeking to raise capital at a reported valuation of about $3.5 billion and is positioning its “world models” approach—learning from video and spatial data alongside text—as a path toward systems that can plan actions with tighter control and safety guardrails, especially in settings where errors carry real consequences.
The most revealing early signal is where AMI wants to apply the technology first: healthcare, led by CEO Alex LeBrun, the co-founder of Nabla and AMI’s first announced partner. The bet is that today’s large language models perform well for tasks like clinical documentation, yet their tendency to hallucinate makes them risky for higher-stakes clinical workflows; AMI argues world models can improve reliability by predicting consequences and sequencing actions under constraints. LeBrun told Forbes he expects initial product rollouts to begin in about a year, while the broader “world models” race is heating up with other well-funded efforts, including Fei-Fei Li’s World Labs.
