While Large Language Models (LLMs)—AI systems trained on massive datasets to generate human-like text and code—have recently solved major outstanding problems, the signatories claim that using these challenges merely as benchmarks to test machine performance is detrimental to the field.
The mathematicians explain that their profession relies on a "human transmission chain" where complex ideas are simplified, debated, and eventually turned into textbooks for students.
They argue that the rapid-fire production of verified "true or false" solutions by AI often lacks the proper write-ups, citations, and isolation of new methods that allow humans to actually understand the results.
Without this conceptual insight, the mass production of answers risks destroying the intellectual training and collaborative culture required to sustain mathematics as a science.
This warning highlights a broader tension between AI developers and creative professions regarding attribution and the value of the human process.
The signatories state that the decisions made by those in control of AI technology will determine whether it serves as a tool to enhance genuine study or as a mechanism that obfuscates knowledge.
They are calling for an urgent dialogue between the technology companies and the mathematical community to ensure that AI is used to accelerate understanding rather than simply producing answers that humans cannot fully own or learn from.