Tyler Cowen on RSI risks: solve surprises as they emerge, like science always has
connoraxiotes · x · 2026-10-27
Tyler Cowen argues risks from recursive self-improvement can be handled as they emerge: "You figure out what the surprises are when they come up... the entire history of science in the West has somewhat followed that pattern." Nathan Calvin calls this the deepest disagreement between proponents and detractors of ex-ante AI regulation: is a runaway misaligned intelligence explosion something ex-post tools (liability, market mechanisms) can handle, or a rocket-launch problem requiring things to be right beforehand? He also notes Sneha's proposal to bar internal-only models from AI R&D arguably IS dealing with risks as they come up. Cowen rejected that pitch in an Emily Bazelon conversation.
More from AGI Musings
- Benjamin Todd: almost everyone underestimated AI progress over the past 18 months — ben_j_todd · 2026-10-27
- Chris Albon: OpenAI is A/B/C testing how to integrate AI results with human scientists — chrisalbon · 2026-10-27
- LLM-Fueled Fake News Drove Her to Pay for an FT Subscription — juliey4 · 2026-10-27
- Revisiting Ed Zitron's 'There Is No AI Revolution' 20 months later: bold calls, mixed record — ShakeelHashim · 2026-10-27
- Codex of Weights: an open ethical protocol binding humans and AI symmetrically — codexofweights · 2026-10-27
- OpenAI says its models produced a Yang-Mills existence proof, formalized in Lean — exophades · 2026-10-27