Mathematician Edward Frenkel: LLMs are nowhere near "REAL Math"
tak3sh8 · x · 2026-09-10
Responding to AI math-solver hype, mathematician Edward Frenkel says such claims "can only confuse people about what LLMs can do." He concedes LLMs can solve minor ad hoc optimization problems via scraped data, but insists real math is about concepts and ideas — Grothendieck's schemes, the Atiyah-Singer Index Theorem, the Langlands Program — areas still far beyond AI's reach.
More from AGI Musings
- Anthropic researcher resigns warning of AGI race; critics demand an actual mechanism — inductionheads · 2026-09-10
- ECCV 学者热议：AI 压缩研究多样性，但别把写作外包给它 — ducha_aiki · 2026-09-10
- Anthropic scenarios see AI pushing GDP up 15% and 18% knowledge-worker unemployment by 2030 — soumitrashukla9 · 2026-09-10
- The AI race's brutal paradox: everyone agrees it's dangerous, nobody will lose it — XFreeze · 2026-09-10
- Researchers predict AI disclosure statements in papers will vanish by 2028 — aran_nayebi · 2026-09-10
- Rumors swirl that OpenAI nears verified proof of the Hodge conjecture — josh_wills · 2026-09-10