@penzlik

gm frens. Paul Christiano (RLHF co-inventor, longtime AI doom voice) just joined OpenAI's Foundation board + Safety and Security Committee, Sept 9. His take, if superintelligence ships without solid alignment, we lose control permanently, and most people could die. Rough odds he gave, ~4% in a year, ~15% in three years, approximate data. The part that matters for CT, every AI-agent token pumping right now assumes smooth scaling. Nobody's pricing the alignment-failure tail risk. Genuine caution or optics before regulators move in. What's your read frens

3 replies

0 recast

7 reactions