Tag
3Blue1Brown (Grant Sanderson) argues in a post: even if AI completely surpasses humans in proof generation, the value chain and funding chain of mathematics will remain largely intact. What really changes are culture and the allocation of credit — making now exactly the best time to enter mathematics.
A guest post by leading formalization researchers (Jeremy Avigad, Kevin Buzzard, Johan Commelin, Rémy Degenne, Kim Morrison, Oliver Nash, Adam Topaz) reflecting on AI's rapid theorem-proving advances, cautioning against rushed large-scale formalization efforts like Math, Inc.'s sphere-packing project and Anthropic's Fermat's Last Theorem verification, and outlining a positive vision for high-quality, community-driven formalization of mathematics.
In a Numberphile interview, Fields Medalist Terence Tao discusses mathematics entering an industrialized era, moving from “proof scarcity” to “proof abundance,” the misalignment of incentives between AI companies and academia, and the risk of human understanding being sidelined.
Derek Thompson 探讨 OpenAI 在数学领域取得突破性成果后，数学界所经历的震惊与悲伤，反思人工智能对人类智力与自我价值的冲击。
A commentary questioning whether AI proof generation alone can replace mathematicians, contrasting Anthropic's Fermat project and OpenAI's math release with the harder challenge of reliably explaining and connecting new results.
帖子指出，Terrence Tao 在加州理工的演讲被社区误读为反 AI，实际上他呼吁为 AI 数学研究投入更多算力，并鼓励对前沿模型进行更具创造性的应用。
文章抱怨AI实验室在利用数学家验证AI生成的定理和证明，却不提供相应的免费API token资源，呼吁实验室为数学验证工作提供支持。
一条推文称 Anthropic 的 Opus 5.5 仅凭一个提示和论文页面 URL，便将 OpenAI 的 722 篇 AI 数学论文浓缩成一段 92 秒的动态图形解说视频。
OpenAI 发布了一个新仓库，收录其未公开内部模型产生的 719 篇数学手稿（归为 372 个结果族），涵盖黎曼 zeta 函数零点区域、Mahler 猜想、自由群因子同构等前沿问题，其中约 42% 的结果已用 Lean 形式化验证，并附带部分推理过程摘要。
作者探讨 AI 在数学证明中的角色，质疑为何 AI 不能在生成证明之外，进一步理解、解释并推广证明背后的数学思想，而非仅停留在验证层面。
The article explores how soon AI might effectively 'solve' mathematics and what obstacles remain, framing the discussion around the frontier of AI-driven mathematical reasoning.
OpenAI's so-called 'slop drop' of hundreds of AI-generated results on open math and computer science problems has left mathematicians both impressed and skeptical, with groundbreaking proofs emerging even as questions persist about how the results were produced.
This opinion piece argues OpenAI is justified in publishing its AI-generated mathematical results through its own channels rather than traditional peer review, and that mathematicians' complaints stem from self-preservation rather than academic integrity concerns.
A commentary highlighting the perceived hypocrisy of mathematicians who spent decades justifying taxpayer funding on societal benefit grounds, but now decry AI solving famous open problems as harmful to the field and its community.
OpenAI released 372 major mathematical breakthroughs, including a machine-generated proof of Subhash Khot's Unique Games Conjecture with a Lean certificate — the first results for which human mathematicians have not yet understood the proofs.
A report claims AI systems have solved roughly 90 of the 500 most important open problems in mathematics, with a leading mathematician calling it the most significant moment in mathematical history.
UT Austin Math Chair Francesco Maggi says OpenAI appears to be preparing to release around 400 AI-generated mathematical proofs at once, signaling that mathematical discovery may no longer be the scarce resource — human understanding is.
Guest post by Jeremy Avigad on Terence Tao's blog reflecting on how recent AI capabilities are disrupting mathematical research workflows, and arguing that mathematicians must adapt by pursuing harder problems and deeper understanding in the age of AI.
A commentary on the rapidly changing field of AI for mathematics, mapping mathematicians' varied reactions — denial, anger, bargaining and acceptance — to recent advances such as OpenAI's internal models reportedly solving hard problems.
A workshop paper presenting a single-instance case study in which an AI agent is asked to rediscover a hidden invariant of degree-four Blaschke curves from numerical data, yielding a homogeneous cubic that predicts unseen configurations with machine-precision residuals, while honestly noting limitations such as a deterministic polynomial-fitting baseline matching the agent's result.