OpenAI宣布其Astra模型在数学和计算机科学领域实现十项突破，涵盖几何、密码学和编码理论等多个学科方向[1]。这一进展标志着生成式AI在数学研究中的影响力不断提升，但同时也在学术界引发了关于伦理、学分归属和评价标准等深层问题的广泛讨论[1]。arXiv预印本库近期数学领域的投稿数量出现急剧增加[1]。
数学家社群对AI在研究中的正确角色尚未达成共识[1]。全球数千名数学家签署的《莱顿宣言》主张，AI应该增强而非替代人类的数学创意[1]。同时，部分学者出于对技术环保和社会影响的担忧而完全拒绝在合作中使用AI[1]。知名数学家Terence Tao在国际数学家大会上发表演讲，讨论了"AI时代"的到来，敦促数学家提前思考相关问题[1]。
AI的计算效率也引起了关注。作者Aluna Rizzoli使用OpenAI模型用43小时找到了一个搜索了两年多的数学对象，而其他协作者的传统方法则需要数月才能完成[1]。这些对比凸显了生成式AI在加速数学研究进程中的潜力，但其带来的挑战和影响仍需要数学界进一步探索和规范。
Generative artificial intelligence is reshaping mathematical research in unprecedented ways, prompting the global mathematics community to grapple with fundamental questions about attribution, evaluation standards, and the proper role of AI in academic work.[1] OpenAI announced ten breakthroughs by its Astra model across mathematics and computer science, spanning disciplines including geometry, cryptography, and coding theory.[1] The surge in mathematical submissions to the arXiv preprint repository reflects the growing adoption of these tools among researchers.[1]
The response from mathematicians has been mixed and contentious.[1] The Leiden Declaration, signed by thousands of mathematicians worldwide, advocates that AI should augment rather than replace human mathematical creativity.[1] Conversely, some scholars have refused to use AI in their collaborative work entirely due to ethical concerns about technology's environmental and social impact.[1] Terence Tao, speaking at the International Congress of Mathematicians, urged the mathematical community to deliberate proactively on the challenges of an "AI era."[1]
The practical implications are already evident in research outcomes.[1] Researcher Aluna Rizzoli employed an OpenAI model to locate a mathematical object in 43 hours that had eluded researchers through traditional methods for over two years, whereas collaborators using conventional approaches would have required months to achieve the same result.[1] This efficiency gain underscores AI's potential utility, yet it remains unclear whether such productivity gains justify fundamental shifts in how mathematical research is conducted, validated, and credited.[1]
