Debate: does it matter if AI proves millions of theorems no human understands?
anshulkundaje · x · 2026-09-13
Stanford professor Anshul Kundaje argues that AI autonomously solving millions of math theorems in a day means little if no one understands them and humans lose the ability to comprehend intellectual work. sytelus counters that math is universal and not "by humans, for humans": proofs force acceptance of truth regardless of human understanding, as it would matter if aliens proved P vs NP.
Related event: Debate: do AI-proved theorems count if humans can't understand them?(3 posts)→
More from AGI Musings
- Should AI labs be nationalized instead of IPOing? AI circles debate — MatthewBerman · 2026-09-13
- UK GDP beats forecasts at 0.4% in July, with programming and info services adding 0.12pp — krishnan · 2026-09-13
- AI Takes Over the Drudgery: Humans Return to Thinking and Appreciating Theorems — deepakns · 2026-09-13
- 'In Weeks to Months': Building Apps Will Be Like Driving a Car, Professor Argues — deepakns · 2026-09-13
- Professor: AI-Driven Fast Execution Frees Students for Real Brain Training — deepakns · 2026-09-13
- Dario Amodei: AI is on an exponential, no wall in sight — kimmonismus · 2026-09-13