Published Nov 30, 2024
Beyond Preference Alignment: Teaching AIs to Play Roles & Respect Norms, with Tan Zhi Xuan
MIT PhD student Tan Zhi Xuan critiques traditional AI alignment approaches, proposing innovative role-based systems and Bayesian rule induction to teach AI social norms, while exploring AI's potential transformations, political and ethical considerations, and the complexities of aligning AI with societal values.
Topics covered
Popular Clips
Questions from this episode
- Asked by 204 people
- Asked by 184 people
- Asked by 145 people
- Asked by 142 people
- Asked by 124 people
- Asked by 116 people
- Asked by 107 people
- Asked by 103 people
- Asked by 97 people
- Asked by 92 people
- Asked by 92 people
- Asked by 90 people
- Asked by 90 people
- Asked by 83 people
- Asked by 78 people
Episode Highlights
Related Episodes
Empathy for AIs: Reframing Alignment with Robopsychologist Yeshua God
Answers 383 questions
Zooming Out on AI, from the Nick Halaris Show
Answers 383 questions
The AI Reasoning Revolution with Ought's Jungwon Byun and Andreas Stuhlmüller
Answers 383 questions
Seeing is Believing with MIT’s Ziming Liu
Answers 383 questions
The AI Safety Debates with Zvi Mowshowitz
Answers 383 questions
The AI Multimodal Revolution with Junnan Li and Dongxu Li of BLIP & BLIP2
Answers 383 questions
Understanding AI "Understanding" with Robert Wright of Nonzero Newsletter & Podcast
Answers 383 questionsAI AMA – Part 2: AI Utopia, Consciousness, and the Future of Work
Answers 383 questions
AI Safety Regulations: Prudent or Paranoid? with a16z's Martin Casado
Answers 383 questions
