Toby Ord and Tom Davidson debate whether the AI research-spend curve breaks
tobyordoxford · x · 2026-11-10
AI safety researchers Toby Ord and Tom Davidson debate a key curve in intelligence-explosion models: as AECI (effective cognitive inputs) goes to infinity, research spend (mainly compute dollars) must also go to infinity or the curve breaks.
- Davidson: it depends on how you operationalise "research spend." If improving the quality of AIs doing AI research counts, the curve can hold and spend goes to infinity; if not, it breaks—you'd hold spend constant while massively increasing cognitive inputs.
- Ord, counting actual $ compute spend: interestingly the curve hasn't broken so far, despite claims of a doubling in AI pace from more-than-doubled research pace.
- Ord adds: if the curve must break for an explosion, monitoring it could yield useful signal.
More from AGI Musings
- I built an early-cancer-detection AI in 2015 — it still runs in hundreds of hospitals — itsOmSarraf_ · 2026-11-10
- Ofir Press: coding isn't 'solved' until agents no longer need human oversight — OfirPress · 2026-11-10
- The Second Station: why 5x-ing one factory station doesn't save the line — saranormous · 2026-11-10
- AI chatbot referrals surge but can't offset lost search traffic, Axios reports — gaganghotra_ · 2026-11-10
- RSI is just an engineering problem — ASI is a different story entirely — haider1 · 2026-11-10
- Claude-built visualization: insects may hold 62 million times more experience-years than humans — javirandor · 2026-11-10