OpenAI宣布其下一代主要模型系列Astra的内部版本成功解决了十个长期悬而未决的开放问题[2]。这些问题涵盖数学、量子复杂性和理论计算机科学等领域[2]，其中包括计算永久式的新型电路下界[2]。OpenAI认为这代表了科学推理领域的重大进步[2]。
这一成就引发了关于人工智能未来角色的深层思考。有观点认为，AI将在数学领域超越人类数学家的品味、直觉和教学能力[1]。与国际象棋等领域不同，数学是科学的核心动力，其成果会对下游应用产生实质性影响[1]。此类AI进步可能改变数学研究的社会语境，使人们难以长期维持纯内在动机从事数学工作[1]。不过也有观点指出，技术进步并非不可避免，人类可以选择是否采用通用人工智能[1]。
OpenAI announced that an internal version of its next-generation Astra model has solved ten major open problems spanning mathematics, quantum complexity, and theoretical computer science.[2] Among these breakthroughs is a novel circuit lower bound for computing the permanent, which OpenAI described as a significant advance in scientific reasoning.[2]
The implications of this development extend beyond pure mathematics.[1] Unlike other domains where AI has demonstrated superiority—such as chess—mathematics serves as a foundational driver of scientific progress, and its discoveries carry direct consequences for downstream applications in engineering and science.[1] As AI continues to advance in mathematical reasoning, it may eventually exceed human mathematicians not only in problem-solving ability but also in intuition, taste, and pedagogical skill.[1] This trajectory could fundamentally alter the social context of mathematics, potentially undermining the intrinsic motivation that has traditionally sustained pure mathematical research.[1] However, technological progress is not inevitable; humans retain the choice of whether to adopt artificial general intelligence systems.[1]