Peter Stone, Truchard Foundation Chair in Computer Science and Chair of the Department of Computer Science, presents the keynote “From How to Learn to What to Learn in Multiagent Systems and Robotics” at the Association for the Advancement of Artificial Intelligence (AAAI) Conference. In this talk, Stone explores how advances in AI are shifting the focus from how agents learn to what they should learn, a central challenge in multi-agent systems and robotics. The keynote advocates for a move toward Agentic AI—systems that don't just follow a fixed learning pipeline but can intelligently decide which parts of their environment and which sub-tasks are worth their computational effort.
Watch here.