# AI's Role and Limitations in Mathematics Can artificial intelligence replace mathematicians? Here are the facts. - **Source:** [Schneier on Security](https://brifing.si/source/7aee9996-c7c0-47f7-89ec-66a1ed95f8cc) - **Published:** 2026-08-28 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/9c75c8f6-1b36-4897-b60b-1dcbf830b82f ## Mathematicians' AI Concerns Forty top mathematicians gathered at OpenAI offices to share concerns about the future of their profession. AI's growing role in mathematical research has many experts worried about their careers. However, some argue that AI cannot yet replace human mathematicians. Particularly in theoretical mathematics, AI has limitations in developing deep conceptual frameworks, according to specialists in the field. ## AI's Mathematical Achievements OpenAI's latest AI model disproved the unit distance conjecture, an 80-year-old famous problem in discrete geometry. Anthropic published two AI-derived results in academic cryptanalysis. These successes demonstrate AI's increasingly important role in mathematical research. However, most of these discoveries are classified either as new applications of existing techniques or as counterexamples to mathematical statements humans were trying to prove. ## The Jacobian Conjecture Example One of AI's most notable mathematical achievements was finding a counterexample to the Jacobian conjecture. While discovering this counterexample was difficult, verifying it was relatively easy once found. The AI combined a form of intuition acquired through machine learning with extensive computational search to find the right example, showcasing its power in recombining existing knowledge. ## Unit Distance Conjecture The unit distance conjecture was motivated by an elegant construction, and most mathematicians believed it to be essentially optimal. The counterexample found by AI introduced ideas from algebraic number theory. A human expert with this specific background could likely have found this counterexample if deliberately searching for it, but there was no reason for someone with precisely this expertise to focus on this particular problem. ## Limits of AI Creativity While current AI systems are creative in recombining existing ideas in novel ways, they remain weak at developing conceptually new theories or structures. Mathematics progresses by identifying core objects in a problem and developing theories to understand them. AIs are strong at searching and recombining existing ideas but still inadequate at creating deep, sustained new theoretical frameworks. ## AI's Successes in Other Fields AIs have demonstrated capabilities like playing Go at grandmaster level or performing Nobel prize-level chemistry in protein folding. These achievements show AIs possess creative abilities in choosing correct directions and making unexpected connections between subjects. However, these abilities differ from developing entirely new conceptual frameworks in mathematics. ## AI's Future Potential As AI models become more capable, all these mathematical abilities emerged as properties rather than being explicitly designed. Experts are confident AI models will someday possess the type of creativity needed to do novel mathematics. Whether this will happen in months, years, or decades remains uncertain, but the trajectory suggests it may occur sooner rather than later. ## Hope for Mathematicians In the short term, AI completely replacing mathematicians seems unlikely. While AIs can efficiently combine and analyze existing knowledge, they haven't yet matched human intuition and creativity in developing deep theoretical frameworks. This offers some relief to mathematicians, though the rapid pace of technological advancement means they must prepare for potential changes in their profession's future. ## Sources covering this story - https://www.schneier.com/blog/archives/2026/08/ai-doesnt-mean-the-end-of-mathematics-at-least-not-yet.html — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._