Socher on extinction risk claims: extraordinary claims need extraordinary evidence, not extrapolation
RichardSocher · x · 2026-09-12
Richard Socher commented on a conversation about the safety of recursive self-improving superintelligence, calling it important and timely. His core position: extraordinary claims about extinction probabilities require extraordinary evidence — not extrapolation plus handwaving about progress. He thanked interviewer @nxthompson for asking hard and interesting questions.
More from AGI Musings
- Skeptic: some AI takeover scenarios ignore how computers and networks work — almmaasoglu · 2026-09-12
- AI researcher littmath: talent poaching is driving more secrecy in the field — littmath · 2026-09-12
- Chess analogy reignites AI doom debate: no step-by-step path doesn't mean no checkmate — harris_edouard · 2026-09-12
- AI Safety Debate: Fail-Closed Mechanisms Fail If the Button Stays With the Labs — AryHHAry · 2026-09-12
- Yampolskiy: Superintelligence needs no body to wipe out 99% of humanity — romanyam · 2026-09-12
- Debating the mathematician letter: can AI search be indexed for serendipity instead of goodharted? — curious_vii · 2026-09-12