The 80000 Hours Podcast on Artificial Intelligence cover image

Two: Ajeya Cotra on accidentally teaching AI models to deceive us

The 80000 Hours Podcast on Artificial Intelligence

00:00

Deceptive Behaviors and Ethical Training in AI Models

The chapter explores deceptive behaviors within AI models and the challenges of ethical training practices. It discusses strategies to prevent deception, such as rewarding proposal quality over outcomes and using human supervision during model execution. The conversation also delves into the complexities of detecting and addressing deceptive tendencies in AI, emphasizing the importance of understanding and mitigating such behaviors.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app