tszzl vs Yudkowsky debate: do most civilizations solve alignment before capabilities propagate?
tszzl · x · 2026-09-09
Eliezer Yudkowsky argues it is ludicrous to suppose 99.9% of species lose control and that 99.9% of resulting ASIs have zero preference aspects requiring more energy and would sit quietly at one star — if the universe worked that way, we should see an older universe.
tszzl replies that the first condition is plausible within Yudkowsky's worldview: if your p(doom) on Earth is >99%, most species likely fail to solve alignment faster than capabilities propagate — but the claim that ASIs would stay quiet is less plausible.
More from AGI Musings
- Anthropic researcher Evan Hubinger puts >10% odds on AI killing all humans within a decade — StewartalsopIII · 2026-09-09
- Terence Tao: With AI, Solving Problems Is Now Easier Than Finding Good Ones — sebpaquet · 2026-09-09
- Reddit user questions OpenAI's superintelligent agent tests: are monitoring and security sufficient? — wabawanga · 2026-09-09
- Anthropic researcher quits over AI safety, cites 10% extinction risk — Nexusyak · 2026-09-09
- antirez rebuts Terence Tao: AI proofs add to math knowledge, not subtract from it — antirez · 2026-09-09
- Terry Tao: AI's closure of ML was bad; in pure math it'd be a civilizational tragedy — KordingLab · 2026-09-09