The Centre for Artificial Intelligence (CAI) are delighted to be welcoming David Abel, from Deepmind, to deliver his seminar, "General Agents Contain World Models."

Abstract

Are world models a necessary ingredient for flexible, goal-directed behaviour, or is model-free learning sufficient? In this talk, I present a formal answer to this question from our recent paper which proves that any agent capable of generalizing to multi-step goal-directed tasks must contain a predictive model of its environment. We further show that this model can be extracted from the agent’s policy, and that increasing the agent’s performance or the complexity of the goals it can achieve requires learning increasingly accurate world models.

About the speaker

David Abel is a Senior Research Scientist at DeepMind and an Honorary Fellow at the University of Edinburgh. His work focuses broadly on the foundations of AI, with a recent focus on agency.