'Math hasn't fallen to AI': Nate Witkin pushes back on the viral claim with a reality check
sebkrier · x · 2026-09-11
Responding to benjtodd's claim that 'maths is the intellectual domain that's fallen first to AI' (easiest to set up verification loops, with other domains to follow), Nate Witkin argues the online discourse is unmoored from reality:
- Mathematical practice in universities and classrooms looks remarkably similar to the pre-AI world; the vast majority of open problems remain unsolved, without substantial AI-driven progress — and not for lack of trying
- Even those close to AI advocacy persistently fail to take 'extreme jaggedness' seriously: models are superhuman at some tasks, but no academic or enterprise domain shows models flatly superior to humans
- Verdict: the field is changing, but 'math has fallen' is false, and human mathematicians will have plenty to do for the near future
More from AGI Musings
- Sophos CISO: AI agents are functionally insiders, monitor them like insider threats — TechNadu · 2026-09-11
- AI researcher: coherent dystopia is unlikely — expect extinction or near-utopia after ASI — adam_dorr · 2026-09-11
- Demis Hassabis Wins RSA Albert Medal, Says Nobody Knows What Happens Next — TorturedPoet30 · 2026-09-11
- Ex-OpenAI/Anthropic researcher warns AI could copy itself beyond unplugging — rohanpaul_ai · 2026-09-11
- Hot take: Consumer AI assistants won't be the next iPhone — only 1-10% actually want AI bots — ivan_bezdomny · 2026-09-11
- AI could add 0.3-0.4 points to Europe's productivity growth, but the EU holds under 5% of global compute — rohanpaul_ai · 2026-09-11