AI researcher proposes harder test: no O(n²) randomized algorithm for n-bit primes conjecture
geoffreyirving · x · 2026-11-21
AI safety researcher Geoffrey Irving: now that the Riemann hypothesis apparently is too easy for AIs, we should switch to harder conjectures. He offers a 2020 conjecture as an example:
> Conjecture (2020): There is no O(n²) expected-time randomized algorithm to generate ≥n-bit primes.
A notable signal of the AI-math arms race — researchers are now picking targets harder than Millennium Prize Problems as new benchmarks for AI.
More from AGI Musings
- Hot take: Millennials struggling to 'learn AI' are the new boomers with the internet — vxnuaj · 2026-11-21
- Terence Tao blogs on the Riemann hypothesis, weighs AI-generated proofs and Masullo manuscript — burny_tech · 2026-11-21
- 'Humans have no business doing mathematics any more' — investor on AI's Ridiculous Gap — birchlse · 2026-11-21
- Stanford's Omar Khattab coins 'virality law': humans and LLMs don't read, they average k-nearest patterns — lateinteraction · 2026-11-21
- Benjamin Bratton: Solving good writing is harder than frontier math, yet humanities call AI bullshit — bratton · 2026-11-21
- Dahlia Ohara: even very smart engineering minds struggle to grasp the fundamental nature of AI — repligate · 2026-11-21