Robert Wiblin puts 80% death risk on max-speed ASI, 30% with 3-year delay
trevposts · x · 2026-11-13
Robert Wiblin lays out his blunt AI risk math: rushing to RSI/ASI at maximum pace carries maybe an 80% chance of human extinction; companies are already on track to concede only 3 months of "delay + reduced pace," which alone might cut risk to 70%; squeezing in 3 extra years of testing with highly capable AI assistance on alignment and security could bring it down to 30% — noting AIs could do more alignment research in that window than humans have on all topics ever. Reposter trevposts agrees with the overall shape, saying he's somewhat more optimistic about the effect of a 3-month slowdown.
More from AGI Musings
- "Humanity loses control in 3-12 months" claim sparks debate over whether collective will exists — repligate · 2026-11-13
- Armin Ronacher: I started 141 projects in a year, but only maintain 10-15 — TheTuringPost · 2026-11-13
- Daniel Kokotajlo: pretraining loss is a checkable task proxying nearly all downstream skills — nitarshan · 2026-11-13
- Nina Schick: In the 21st century, power is bound to artificial intelligence — NinaDSchick · 2026-11-13
- UW Prof. Abhishek Gupta muses on academia's role in robotics RL research at CSIR workshop — chris_j_paxton · 2026-11-13
- AI wealth's 'third wave' of philanthropy: ~$37B a year, best spent catalyzing markets — RishiBommasani · 2026-11-13