Welcome to the AI crisis in math

The intersection of artificial intelligence and mathematics has shifted from a niche curiosity to a full-blown existential crisis for the academic community. According to reporting by Robert Hart for The Verge, the recent publication of AI-generated solutions to longstanding mathematical problems has sent shockwaves through the field. This development forces a difficult conversation about the future role of human mathematicians in an era where frontier models can produce professional-level proofs.

The paradox at the heart of this crisis is stark. Modern AI systems remain notoriously bad at elementary arithmetic, struggling with tasks like counting letters in a word or even telling the current day of the week. Yet, these same models are increasingly capable of navigating complex, abstract mathematical landscapes that have stumped humans for decades. This disconnect suggests that AI is not merely calculating but engaging in a form of high-level reasoning that bypasses traditional step-by-step logic.

This rapid transition from incompetence to competence in advanced domains has compressed years of industry adaptation into a matter of months. Mathematicians are now grappling with questions that software engineers faced earlier. If AI can identify patterns and solve problems at a professional level, what is the unique value proposition of human expertise? The traditional path of training new generations of scholars to identify and solve outstanding questions feels suddenly obsolete.

There is also a growing skepticism about the motives behind these breakthroughs. Some in the community wonder if the intense focus on mathematical prowess is merely a marketing exercise for AI labs. These companies may be showcasing technical capability to attract investment and talent, with little genuine interest in the preservation or advancement of mathematics as a fundamental academic discipline.

The implications for academic funding and university programs are profound. Grants and research positions are often justified by the promise of human-led discovery. If frontier models can answer outstanding questions without human intervention, the economic and intellectual justification for these programs weakens. Universities must now consider whether their role is shifting from discovery to verification and curation of AI-generated insights.

This crisis mirrors the earlier disruptions in software engineering, where AI tools began to automate coding tasks. However, mathematics is arguably more foundational to science and technology. A failure to adapt the mathematical community could have ripple effects across all STEM fields. The ability to reason abstractly is a core human strength, and seeing it replicated by machines challenges our understanding of intelligence itself.

What this means for you: If you work in data science, research, or any field requiring complex problem-solving, you should view AI as a collaborative partner rather than a replacement. Use AI to generate hypotheses or identify patterns, but maintain rigorous human oversight for verification. Try this workflow: Ask your AI assistant to outline potential approaches to a complex problem, then manually critique each step for logical consistency before proceeding. This ensures you retain the critical thinking skills that AI currently lacks.