菲尔兹奖获得者 William Timothy Gowers 分析了近期人工智能（AI）在数学领域的突破，得出结论：虽然 AI 解决了长期存在的难题，但它尚未达到顶尖人类数学家的水平。其主要优势在于寻找反例——识别出反驳普遍猜想的单一对象，而非直接证明定理。

William Timothy Gowers 确定了形成大型语言模型（LLM）数学风格的两种独特能力：广阔的知识覆盖和高速的尝试执行。虽然人类数学家每天可能彻底评估极少量的路线，但 AI 处理了巨大数量的潜在解决方案，承受了巨大比例的失败以实现单一的成功组合。这表明 AI 的优势源于搜索规模和组合能力，而不是单一的深刻见解。

目前，人类在具有极深的搜索树、众多的分支路径以及需要立即消除错误的极少数量的可行路线的问题上保持优势。然而，考虑到快速的发展速度，William Timothy Gowers 预计会出现范式转变。当 AI 生成一种单一的、出乎意料的新颖方法（该方法不是现有文献的微小变体，优雅地统一了概念，并启动了多个后续研究轨迹）时，真正的精通将被证明。

Fields Medalist William Timothy Gowers analyzed recent artificial intelligence (AI) breakthroughs in mathematics, concluding that while AI solves long-standing problems, it has not reached top human mathematician levels. Its primary advantage lies in finding counterexamples—identifying a single object that disproves a universal conjecture, rather than proving a theorem directly.

William Timothy Gowers identified two distinct capabilities forming a Large Language Model (LLM) mathematical style: vast knowledge coverage and high-speed trial execution. While human mathematicians might thoroughly evaluate a minimal number of routes per day, AI processes an immense volume of potential solutions, absorbing a massive percentage of failures to achieve a single successful combination. This indicates AI's strength stems from search scale and combinatorial power rather than singular profound insights.

Currently, humans maintain an edge in problems featuring extremely deep search trees, numerous branching pathways, and a scarce number of viable routes requiring immediate error elimination. However, given the rapid developmental velocity, William Timothy Gowers anticipates a paradigm shift. True mastery will be evidenced when AI generates a singular, unanticipated novel method that is not a minor variant of existing literature, unifies concepts elegantly, and initiates multiple subsequent research trajectories.