A proof is only intelligible when each part is graspable 'holistically'
francoisfleuret · x · 2026-09-17
François Fleuret argues that for a math proof to be intelligible, being a valid sequence of symbols is not enough: it must be recursively composed of elements each simple enough to be grasped holistically by the mind, so that the subjective sense of correctness aligns with actual formal validity. The point also sheds light on why LLM-generated formal proofs often feel unreadable.
More from AGI Musings
- Engineer's interview ended midway after admitting he hadn't used AI — CtrlAltDwayne · 2026-09-17
- Beyond "stochastic parrots": Wigner's unreasonable effectiveness now applies to LLMs — anselm · 2026-09-17
- EA 'cult' debate reignites as Jesse Singal calls critics reverse-engineering their priors — mark_k · 2026-09-17
- Toby Walsh: 'kill us all' scenario fantastical, dangerous AI capabilities real — TobyWalsh · 2026-09-17
- AI scientist Toby Walsh: 'kill us all' scenario fantastical, real dangers remain — TobyWalsh · 2026-09-17
- Ethics of Simulation: Extend to Others the Complexity You'd Want for Yourself — basedjensen · 2026-09-17