Scientific American: after AI upended math in 2026, is science next in line?
HistedLab · x · 2026-11-20
A Scientific American feature reviews how AI upended mathematics in 2026 and asks whether science is next:
- Timeline: in March, Harvard physicist Matthew Schwartz warned at the APS meeting that "AI is going to do lots of science" — and was met with derision. Months later the picture reversed.
- What already hit math: in May, OpenAI models guided by humans disproved a long-standing Erdős conjecture; in August, AI models effectively solved the Navier-Stokes problem amid controversy; in October, OpenAI released 722 AI-generated preprints on 370 open problems, including a partial attack on the Riemann hypothesis.
- Science remains murky: unlike proofs, experimental science is harder to automate. Neuroscientist Histed argues AI will struggle where data is scarce — the brain and biology need better measurements, not more intelligence, since nonlinear extrapolation eventually requires more data than can ever be collected.
More from AGI Musings
- The economics of intelligence have shifted: it now pays to apply heavy 'thought' to mundane decisions — jamesdouma · 2026-11-20
- Gemini Neon's advice on decoupling self-worth from economic output wins over a top AI voice — paraschopra · 2026-11-20
- Neuroscientist pushes back on 'AI will solve biology': nonlinear extrapolation data can't fix — HistedLab · 2026-11-20
- 2021 forecast of 2026 by Daniel Kokotajlo goes viral for being eerily accurate — zetalyrae · 2026-11-20
- Chimpanzee tea parties to ASI: a viral thread on how AI reframes humanity — ZeroStateReflex · 2026-11-20
- Rethinking gaming after Civ VII and GTA VI: build agent-friendly, customizable games — bingxu_ · 2026-11-20