Dwarkesh Podcast: Why Raising AI Isn't Like Raising Kids w/ Ryan Greenblatt
Dwarkesh Patel · youtube · 2026-08-18
Dwarkesh Patel interviews MIRI researcher Ryan Greenblatt on the misconceptions of analogizing AI training to child-rearing. Ryan explains why this analogy fails when predicting superintelligent behavior, discussing technical alignment, agent predictability, and control mechanisms. The conversation delves into better frameworks for understanding AI development risks.
More from AGI Musings
- Bottleneck for AI in curing disease is regulation, not data or intelligence — Ronangmi · 2026-08-18
- Proposal: an international treaty between humans and AI to preempt harm — PeterBowdenLive · 2026-08-18
- JAMA Paper: Autonomous AI May Outperform Hybrid Human-AI Care — alvelda · 2026-08-18
- Linthicum: Metrics incentivize hype over analysis in tech media — DavidLinthicum · 2026-08-18
- User complains about AI-generated summaries flooding YouTube comments — Evgenii42 · 2026-08-18
- Defending Dario: Why curing all diseases in 10 years is plausible — DeryaTR_ · 2026-08-18