
AI Safety Risks and Alignment

Anthropic co-founder: AGI predictions, leaving OpenAI, what keeps him up at night | Ben Mann
Topic 9 of 14
Published Jul 20, 2025 · Clip 34:23–45:44 (11:22)
Ben's safety focus stems from sci-fi, Bostrom's Superintelligence highlighting alignment challenges. Path clarified via language models; risks include bioweapons, deception. Anthropic's responsible scaling defines ASL levels; shares failures transparently for policy trust. Odds of misalignment 0-10%, but marginal work crucial.