Rob Leclerc bets P=NP will never be proven — otherwise AI could do anything
robleclerc · x · 2026-10-27
Responding to jenzhuscott's claim that "P=NP will be the ultimate test for AI," Rob Leclerc says he's "pulling a Gary Marcus" by confidently predicting it will never be proven. Turning an exponential 2^n algorithm into polynomial O(n^c) would be a relativity-scale discovery — and if proven, it would be hard to name anything AI couldn't do.
jenzhuscott's original post argues P=NP pits AI's polynomial-time architecture against an inherently exponential problem, demands reasoning about computation's limits, and its resolution would be the clearest signal of general, creative mathematical intelligence beyond pattern matching.
More from AGI Musings
- Daniel Kokotajlo debates a Chinese AI safety startup on whether CoT monitorability evals prove enough — S_OhEigeartaigh · 2026-10-27
- Mintlify: doc traffic up 10x in a year, agent traffic up 135x, 60% of edits agent-written — handotdev · 2026-10-27
- Hot take: OpenAI spends compute on unsolved problems while Anthropic saves it for next-gen models — iruletheworldmo · 2026-10-27
- Dev job market: it's not fewer jobs overall, it's fewer easy jobs, says founder — yenkel · 2026-10-27
- Vox: Anthropic should quit the AI race unless it can prove its safety strategy works — GarrisonLovely · 2026-10-27
- Stanford's 457-page AI Index: cost down ~30%/yr, top-model gap narrows to 5.4% — mdancho84 · 2026-10-27