Towards AI that Learns, Plans, and Acts with World Models

Mike Rabbat - AMI Labs

Nov. 27, 2026, 3:30 p.m. - Nov. 27, 2026, 4:30 p.m.

ENGMC 11

Hosted by: Hsiu-Chin Lin

In this talk, I will explore the next frontier: world models. I’ll start by providing an overview and motivation for why I believe world models are a necessary component of future AI systems. The term “world model” is overloaded in the broader discourse. For the purposes of this talk I’ll argue that a world model: (a) should construct an agent’s internal representation of the world through perception, and (b) should be able to predict the evolution of this representation through time, including in response to the agent’s actions. Such a world model enables planning. I will discuss how our work on joint-embedding predictive architectures offers a path beyond next-token prediction toward systems that learn from observations of the physical world. By learning directly from video without the need for human-labeled data, this self-supervised approach provides a more efficient path toward machines that learn predictive representations of the physical world. I will outline how this direction will allow us to build the next generation of intelligent agents—systems that don't just talk, but can reason, plan, and act in the physical world.

Michael Rabbat is Co-founder and VP World Models at Advanced Machine Intelligence (AMI Labs), where he is building a new breed of AI systems focused on understanding and reasoning about the world. He is also an Associate Industry Member at Mila.  Previously, Michael was a Research Director at Meta, where he was a founding member of FAIR Montréal, contributing to the company's fundamental AI research efforts. Prior to that, he spent over a decade as a professor at McGill University in Electrical and Computer Engineering. He earned his PhD from the University of Wisconsin, Madison. Throughout his career he has worked on a variety of topics across signal processing, learning, and optimization, including distributed optimization and federated learning. His current research focuses on methods for training world models and using them for planning and control.