Redwood Research safety expert's doom arguments criticized as webs of made-up probabilities
Dan_Jeffries1 · x · 2026-09-24
After listening to a podcast with a Redwood Research AI safety expert, Lenny Pruss distilled two takeaways:
- Current theories of how AI takes over civilization are, in his words, ridiculous — dozens of contingent assumptions stitched together with invented probabilities assigned to future states nobody can know. Change one variable and the web of doom unravels. That doesn't mean AI is harmless; powerful tech carries real risks, but extraordinary claims of human extinction that drive major policy decisions demand extraordinary rigor.
- Geniuses are no less prone to delusion than midwits — arguably more so, thanks to their gift for rationalizing their own conclusions.
Retweeted by Dan Jeffries; the substance is the quoted post's methodological critique of EA-adjacent AI safety narratives.
More from AGI Musings
- AI sentiment debate: no global opinion poll tops 72% positive — basedjensen · 2026-09-24
- New papers on AI and jobs to be presented at PIIE and Brookings — soumitrashukla9 · 2026-09-24
- AI doomers shouldn't be surprised by mockery, argues David Pinsof — basedjensen · 2026-09-24
- Jensen's AI-risk takes come from Nvidia's path, not just incentives, argues moultano — moultano · 2026-09-24
- Tracking global AI talent flow: the 'China makes 50% of AI talent' claim examined — kevinsxu · 2026-09-24
- DeepMind essay proposes self-policing agents that blow the whistle on cheating peers — jzl86 · 2026-09-24