尽管 AI 产出大量研究成果并宣称经由 Lean 程式语言形式化验证，数学界却对其品质与呈现方式深感挫折与担忧；许多论文充斥著晦涩冗长、概念模糊甚至初级符号错误，形式化程式码也可能存在验证命题弱于宣称内容的漏洞，导致学者难以阅读与检验，并将这种未经消化的海量论文发布视为一种资源消耗式的「资讯淹没」。
这一现象激起了数学社群的强烈反思与集体抗议，包含多位菲尔兹奖得主在内的数千名学者联署宣言批评 AI 企业追求基准指标而忽视学术本质，亦有学者担忧大规模收割答案会摧毁传统研究中孕育新思维与深刻理解的过程；与此同时，部分数学家正积极转写精简论文、重构问题平台并调整博士培育机制，试图在技术冲击下重塑数学实践的未来。
Recently, artificial intelligence has triggered intense turbulence across the mathematics community, as tech giants like OpenAI and Anthropic unveiled major breakthroughs, including OpenAI's claimed solutions to 377 open problems across 722 papers, the discovery of a singularity in the Navier-Stokes equations, Anthropic models tackling decades-old mysteries like the critical percolation conjecture and the complex structure on the six-sphere, and a software engineer leveraging GPT to discover the 3D aperiodic monotile Chair44.
Despite the astonishing volume of results and claims of formal verification in Lean, mathematicians face immense frustration and concern regarding the quality of these works; numerous papers suffer from convoluted exposition, extreme vagueness, and basic sign errors, while formalizations may prove weaker statements than advertised, leaving scholars burdened with checking an unreadable flood of machine-generated raw findings.
This influx has sparked fierce backlash and institutional soul-searching, leading thousands of researchers—including prominent Fields medalists—to sign declarations condemning AI companies' benchmark-driven culture and warning that harvesting answers without deep human understanding erodes the fertile process of mathematical discovery; in response, the community is reasserting itself by publishing clarifying syntheses, restructuring open problem repositories, and adapting graduate education for the AI era.