WaPo investigation: Amodei privately put 20-30% odds on AGI destroying humanity
GaryMarcus · x · 2026-11-22
The Washington Post published an investigation into Dario Amodei and Anthropic's secretive internal culture. A 2022 private memo had Amodei warning AGI could be more dangerous than nuclear weapons, estimating a 20–30% chance it could destroy or permanently stunt humanity. At OpenAI he used air-gapped laptops and printers for essays on powerful AI, fearing "infohazards." Anthropic reportedly fired the employee who leaked the memo acknowledging Gulf-state funding could "enrich dictators," and a hiring manager privately described its culture interview as an "are you ready for the cult" exercise.
More from AGI Musings
- "Still a Stochastic Parrot?" Classic AI Debate Resurfaces — rand_longevity · 2026-11-22
- The most valuable AI skill in 2027 isn't prompting — it's taste — jocarrasqueira · 2026-11-22
- Speculation: OpenAI quietly throwing massive compute at remaining Millennium Problems — haider1 · 2026-11-22
- David Sacks: Anthropic's Claude Constitution Encourages Transhumanism — DavidSacks · 2026-11-22
- AI proved the Riemann Hypothesis by brute force — and mathematicians aren't celebrating — peace_venerable · 2026-11-22
- NYT AI projects editor writes first drafts on a 1950s Olivetti typewriter, never with AI — dylfreed · 2026-11-22