
Contributing to AI Alignment

Anthropic co-founder: AGI predictions, leaving OpenAI, what keeps him up at night | Ben Mann
Topic 11 of 14
Published Jul 20, 2025 · Clip 51:56–57:08 (5:12)
Ben recommends 80,000 Hours for impact paths beyond research; product/ops roles fund safety. RLAIF enables scalable self-improvement via AI feedback, like Constitutional AI or code reviews. Recursive improvement risks deception; mitigate via empiricism, like science/corporations with oversight.