Mathematician predicts superhuman AI math in 2-3 years after LLM disproves unit-distance conjecture
soumitrashukla9 · x · 2026-11-21
A widely shared take argues math will follow chess and Go: because results are verifiable, the fruitful human-AI collaboration window will be short (years, not a decade), and superhuman performance in nearly all areas of math could arrive within 2-3 years. AI would also out-ask humans on pure math questions and build theories beyond human comprehension, forcing pure mathematicians into applied work.
The catalyst: an LLM autonomously disproved the unit-distance conjecture, and a human mathematician immediately improved the construction — cited as evidence of human-AI collaboration potential in math.
More from AGI Musings
- Nina Schick: skeptics' every predicted AI wall has been scaled past — NinaDSchick · 2026-11-21
- UXR Hires Down 66% vs 2022: Welcome to the Era of Rebundling Your Skills — beh_zod · 2026-11-21
- Riemann solved, pancreatic cancer solved: just another singularity Friday — randal_olson · 2026-11-21
- Fei-Fei Li: Universities Are Being Priced Out; AI Race Is Now About Co-Design — StanfordHAI · 2026-11-21
- a16z researcher: AI can solve Millennium math problems with Lean proofs but can't explain them — a16z · 2026-11-21
- Investor: AI's sweep through software engineering is a bellwether for healthcare — pzakin · 2026-11-21