Why AI Goes Rogue: Turing Award Winner Explains Reward Failure | Andrew Barto

Andrew Bartow is one of the founders of reinforcement learning.

It goes beyond the title with direct discussion of it's, know, reward, including: He and Rich Sutton wrote the textbook on it.

1. 01It's how today's large models learn to reason through problems
2. 02It's how agents learn to carry out tasks on their own
3. 03Agents chase the reward in ways their designers never intended