Tom Davidson: my emotional orientation to AI takeover risk shifted from intellectual to visceral
eli_lifland · x · 2026-10-31
AI safety researcher Tom Davidson describes two phases in how he relates to AI takeover risk. Phase one (years): purely intellectual — he believed takeover probability exceeded 10% but felt no fear, treating it like a game of accurate forecasting and neglected-cause selection, while cataloging many ways doomers might be wrong (misaligned AGI struggles to align its successor, reward seekers may not seize power, militaries might not deploy AI). Phase two (recently): his orientation has become visceral and emotional — he is now genuinely scared and angry. A notable window into how safety researchers cope with their own forecasts.
More from AGI Musings
- Protocol Labs' Juan Benet: BCI is where AI was in 2019, before ChatGPT changed everything — juanbenet · 2026-10-31
- New paper quantifies ineffable concepts: 65 of 1,270 can never be taught to LLMs via language — sanmikoyejo · 2026-10-31
- A "STOP AGI" protester spotted in San Francisco with handmade extinction-risk signs — aniketmaurya · 2026-10-31
- Google models hold all the world's books but can't read them, dev argues — gnukeith · 2026-10-31
- Is AI the real-world sophon from Three-Body? Reddit debates cognitive lock-in — Original-East-8 · 2026-10-31
- Dev pushes back on scaling mindset: closed-source giant models benefit no one — gnukeith · 2026-10-31