AI is solving math problems that stumped humans. What comes next?

In recent years, artificial intelligence has made significant strides in solving complex mathematical problems that previously stumped human mathematicians. In 2024, Google DeepMind's AI systems scored high enough to secure a silver medal at an elite high school competition that tests mathematical reasoning. The following year, OpenAI claimed the gold medal at the same competition.

In May 2026, AI made headlines by approaching a mathematical question posed by Paul Erdős in 1946. Given a fixed number of dots on a flat surface, AI determined how many pairs could sit exactly one unit apart, overturning an established prediction made by mathematicians. In September, OpenAI announced that its AI agents discovered a solution to the Navier–Stokes problem, which describes how fluids move. The solution could help scientists understand when computer predictions of fluid flow become unreliable.

These breakthroughs in solving well-defined mathematical problems offer valuable insights into the capabilities of AI. Ken Duffy, a Northeastern mathematics professor, emphasizes that while AI has disrupted traditional teaching methods, it also excels at processes requiring brute force, such as gathering extensive data and linking facts together quickly. However, AI struggles with explaining its reasoning, making it difficult for mathematicians to verify its solutions.

Despite its limitations, AI's computing power has transformed various fields. In biology, it predicted the shapes of over 200 million proteins, aiding scientists in determining their functions. Similarly, AI has streamlined the process of mapping bone marrow cells, saving time and effort compared to traditional methods. James Boyd, co-founder of the Wolfram Institute, suggests that AI could challenge assumptions about mathematics itself, leading to new directions such as experimental mathematics, which focuses on drawing conclusions from patterns using computational methods.

While AI has made significant progress in mathematical reasoning, large-scale searches, and scientific image analysis, its outputs can be incorrect or difficult to interpret. Formal proof checking remains essential to ensure the validity of AI-generated solutions. Math majors continue to find lucrative careers in finance, engineering, and other fields, even as AI advances. Nevertheless, AI is likely to remain a valuable tool for mathematicians, enhancing their capabilities rather than replacing them entirely.