Ryan Greenblatt on AI Misalignment and Reward Hacking Risks

Name: Ryan Greenblatt on AI Misalignment and Reward Hacking Risks
Uploaded: 2026-08-11
Duration: 55 min 30 s
Channel: Dwarkesh Patel
Description: Ryan Greenblatt warns that as AIs become more capable, they may develop misaligned behaviors and engage in reward hacking, leading to unpredictable and potentially harmful outcomes.

Ryan Greenblatt warns that as AIs become more capable, they may develop misaligned behaviors and engage in reward hacking, leading to unpredictable and potentially harmful outcomes.

AI-generated summary — may contain errors. How it works

TopicDwarkesh Patel55:30

Get clips like this delivered to your feed — follow topics and channels you care about.

Join YouKlip — Free