The speed limit for AI science is the cost of verification, field by field
professor_ajay · x · 2026-11-21
A sharp observation on why AI advances fastest in math: verification is nearly free — a lean kernel says yes or no, so brute-force generation works since you can produce endless failed attempts and keep the one that compiles. Biology has no such lean checker; verification requires wet labs, which don't scale with compute. The same logic applies to agents at work: tasks with cheap checks get automated first, and the rest wait for someone to build the checker.
More from AGI Musings
- Underrated AI use cases: custom accounting, negotiation sims, state capacity — curious_vii · 2026-11-21
- Soviet planning was just a compute/latency problem, plus humans RLHF'd to flatter Gosplan — l4rz · 2026-11-21
- Blogger's AI timeline: math in 2026, physics 2027, biology 2028, aging 2030 — rand_longevity · 2026-11-21
- Why does a Lean-verified proof get more hate than a contested journal proof? — IAmTimNguyen · 2026-11-21
- Pinker amplifies Hanania's critique: post-scarcity fear is an intellectual luxury, not a real worry — sapinker · 2026-11-21
- "Truth doesn't depend on us": the era of machine-verified but humanly incomprehensible results — c_valenzuelab · 2026-11-21