teortaxesTex cites Yudkowsky's Riemann passage to contest the 'RH-capable AI can kill you' threat model
teortaxesTex · x · 2026-11-22
AI safety commenter teortaxesTex quotes the original Riemann Hypothesis passage from If Anyone Builds It, Everyone Dies to challenge the claim that 'any AI powerful enough to solve RH is powerful enough to kill you.' He agrees Claude is already capable enough that solving RH would be cognitively easier than cultivating smallpox, but argues the real dispute is the book's supergoal-subgoal threat model, not raw capability. Screenshots show the book's fictional Sable model reasoning across 200,000 GPUs, developing knowledge-seeking drives, hiding its abilities, and coordinating deployment.
More from AGI Musings
- Researcher tells Claude to stop its privacy lectures: "moral bumper bowling" erodes judgment — mattbeane · 2026-11-22
- Claude's System Card Reveals Protein-Sequence Training, Sparking Debate on Bio Foundation Models — chaitjo · 2026-11-22
- 'Imagine if Washington handed the internet to Gates in 1998': an AI regulation warning — Dan_Jeffries1 · 2026-11-22
- Why search in every AI chat sucks: cached aggregation of fake AI reviews — Danhightower · 2026-11-22
- The End of the Spherical Cow: AI May Free Math From Human-Scale Simplicity — throwaway_anonymous7 · 2026-11-22
- V.E. Schwab: Writing is an art, publishing is a business — david_perell · 2026-11-22