
EPISODE DETAILS
About this episode
David Silver and Richard S. Sutton argue that experience generated through interaction could become a dominant source of AI learning.
Their proposal connects persistent streams of experience with richer actions, observations, and feedback from consequences.
The discussion follows their account of rewards, planning, world models, and the limits of imitation.
Human guidance remains part of the proposal, including feedback that can change which outcomes a system pursues.
The episode examines the gap between measuring a successful outcome and choosing an outcome worth pursuing.
Author: David Silver and Richard S...