Three safety signals in five days: ex-OpenAI/Anthropic researcher blasts superintelligence race
PeterDiamandis · x · 2026-09-12
Peter Diamandis recaps: a researcher with three years training models at OpenAI and Anthropic resigned, saying "neither company is acting responsibly in the race towards self-improving superintelligence" — his post got 160M+ views. Days earlier, Sam Altman called Navier-Stokes "the strongest evidence yet" for slowing down, and OpenAI's chief scientist called for a voluntary pause.
Also in the roundup: an AI-designed longevity drug reversed aging clocks by 3–6 years in 4 weeks; Google's Alpha Genome Atlas pre-computed 9B possible mutations; better training data yields 12x compute efficiency (better architectures 3.7x); Emad's p-doom fell from 50% to 20%; Anthropic hits $6.5B quarterly run rate, 42% of the coding market, a $35B cloud deal, and a mega IPO approaching at $2T+.
Related event: Ex-Researcher Quits, Blasts OpenAI and Anthropic Race to Superintelligence(3 posts)→
More from AGI Musings
- Burt Totaro on Terence Tao's blog: what math loses if AI brute-forces the Hodge conjecture — natanielruizg · 2026-09-12
- The AI comms gap: Google AI Overviews shape public perception while elites warn of job loss and doom — matt_slotnick · 2026-09-12
- Jensen Huang calls AI safety critique 'deeply untrue' as critic dubs him top beneficiary of unregulated AI race — Turn_Trout · 2026-09-12
- AI storytelling is at the 'animated photographs' stage of early cinema, says Tolan's Eliot Peper — every · 2026-09-12
- Benjamin Bratton livestreams Gray Area talk on Agentworld and Superdark Factory — bratton · 2026-09-12
- Timnit Gebru's same quote, three years apart: extinction talk distracts from real AI harms — austinc3301 · 2026-09-12