Low‑Dimensional Intelligence Limits Foom

If intelligence is low‑dimensional, progress concentrated in a few highly verifiable skills (math, coding) will not automatically produce broadly capable agents able to replace most human work. That means big metric jumps in narrow domains can be consistent with modest near‑term social and economic impact. — This reframes AI timelines and policy: regulators and investors should distinguish narrow benchmark progress from general capability, altering preparedness, procurement, and labor policy.