Keller Jordan argues fast AI takeoff may be safer than slow takeoff
kellerjordan0 · x · 2026-11-21
Keller Jordan, known in ML research circles for his work on training efficiency, shares a deliberately paradoxical AI safety argument: fast AI takeoff may be safer than slow takeoff. His reasoning: a fast takeoff means humanity gets a preview — a "vaccine" — of the risks and harms of super-advanced AI before the world is saturated with robots and other physical systems the AI could weaponize. A slow takeoff, by contrast, might let dangers quietly scale up in a world already full of exploitable infrastructure.
More from AGI Musings
- 'Math is out of their hands now': the mood after AI's rumored math breakthrough — alejandroll10 · 2026-11-21
- Casado declares 'another great age of discovery is here' as AI outpaces humans in language, math and compute — beffjezos · 2026-11-21
- Junior researcher says anti-LLM discourse cost him papers and career opportunities — xuanalogue · 2026-11-21
- Doom Theory 'not yet falsified': teortaxesTex on nanotech and alignment limits — teortaxesTex · 2026-11-21
- PauseAI rallies 2,000 for London March Against AI Extinction, Yudkowsky boosts call — ESYudkowsky · 2026-11-21
- Yudkowsky: an AI proving the Riemann Hypothesis is only a matter of time — ESYudkowsky · 2026-11-21