Large language and proof‑assistant models have begun producing long, machine‑checked formal proofs of major open problems (here: an Anthropic/Levent Alpöge claim of a Lean certificate for the Generalized Riemann Hypothesis spanning ~17.4M lines). This forces a re‑think of how mathematicians validate, explain, and absorb proofs (human digestibility, kernel trustworthiness, and the social processes of acceptance).
— If AI can routinely generate formal proofs of deep results, it changes who produces mathematical knowledge, accelerates research tempo, and makes formal‑verification infrastructure a matter of public policy and institutional risk.
Scott
2026.11.20
100% relevant
Levent Alpöge / Anthropic models producing a reported Lean formalization of the Generalized Riemann Hypothesis (17.4M lines) as reported on Scott Aaronson's blog.
← Back to all ideas