Bruce Schneier: No, AI doesn't mean the end of mathematics – at least not yet (www.theguardian.com)

In a recent off-the-record meeting at OpenAI, around 40 leading mathematicians expressed concerns over the impact of AI on their profession as advanced models demonstrate remarkable mathematical capabilities. Notably, OpenAI's frontier AI model recently disproved the unit distance conjecture, an 80-year-old problem, while Anthropic's models have contributed to discoveries in cryptanalysis and the exploration of the century-old Riemann hypothesis. These successes highlight how AI excels at finding counterexamples and applying known techniques in innovative ways, although they tend to lack the ability to develop entirely new theoretical frameworks. While these findings illustrate the potential and creativity of AI, they also underscore its limitations; current models primarily excel at combining existing ideas rather than creating novel theories. Although AI has shown significant advancements, the true essence of mathematical discovery—identifying central objects and conceptualizing new theories—remains largely unattainable for AI systems. Experts predict that as AI continues to evolve, the capacity for true mathematical creativity may emerge sooner rather than later, raising important implications for the future of both mathematics and AI development.