saved
Scott Aaronson: The Mathocalypse, Will Intelligence Matter Anymore?
Hraness wrote this summary from a saved copy of the source. Quotations are taken word for word from the source.
gist
Ryan Peterman interviews quantum computing researcher Scott Aaronson about the summer AI models overtook humans at theorem proving. Aaronson argues the singularity has already begun for math and theoretical computer science, explains why AI excels at counterexamples, describes how publication and credit norms are changing, and asks what is left for humans once intelligence is commoditized. The publisher hosts the full transcript at developing.dev/p/e070-scott-aaronson.
ideas
- 03:53 The singularity has begun for math. Months, not decades, separated AI's unit-distance result and its Navier–Stokes solution; Aaronson says humans are no longer the main theorem provers.
- 08:30 Humans keep the agenda and the explanation. Chess is the hopeful precedent; AI proofs are badly written, so understanding and re-explaining them remains human work for now.
- Credit follows understanding. Put your name on AI-assisted work only if you understand and defend it; disclosures belong up front because experts care how results were found.
- Counterexamples and P vs NP. Parallel search unbothered by expert priors suits AI; he expects AI to be central to any P vs NP resolution.
- Intelligence gets commoditized. As industry did to strength, AI may do to intelligence; his advice is to study math and CS only if they genuinely interest you.
quotes
“we have been dethroned as sort of the main theoremroving entities on planet earth.”
“it's not uh uh blinded by preconceptions the way that we are”
“This write up is garbage, right? That you know, why is that great news? Cuz it means there's still something for you to do.”
“you take away intelligence and you take away verbal ability, what is left, right?”