Jul 24, 2025, 12:00 AM
Jul 22, 2025, 12:00 PM

Humans outperform AI models in prestigious math competition

Highlights
  • Humans outperformed AI at the International Mathematical Olympiad held in July 2025.
  • AI models from Google and OpenAI reached gold-level scores for the first time, scoring 35 points.
  • The competition raises questions about the future of human capabilities compared to advancing AI.
Story

In Queensland, Australia, the International Mathematical Olympiad (IMO) recently showcased a significant achievement for human competitors, as they outscored AI models developed by Google and OpenAI. Held in July, this prestigious yearly competition invites young mathematicians under the age of 20 to solve complex mathematical challenges. This year, five participants achieved perfect scores of 42, surpassing the best performance from AI models, which scored 35 points each — an impressive feat for artificial intelligence but not enough for full marks. Google's Gemini chatbot was noted for its clear and precise solutions to five out of the six problems presented, marking a notable improvement from its previous year's silver-medal performance where it managed to solve only four. OpenAI’s experimental reasoning model mirrored these results, achieving gold-level status for the first time as well. Their capabilities drew positive recognition from the IMO organizers, who noted the clarity of their solutions. However, there remain concerns regarding the methods of evaluation, as the actual computational power used and the involvement of humans in the AI's processes have not been disclosed. The progress of AI in mathematics invites a complex set of reflections among educators and the broader society, concerning how to best prepare future generations for a landscape increasingly influenced by advanced technologies. It raises a philosophical question: can human virtues like wisdom, justice, courage, and self-discipline offer a persistent advantage over AI? Analysts are keen to see if AI will continue to develop and reach parity or even exceed human capabilities in specific domains. Experts in AI, such as Google DeepMind CEO Demis Hassabis, predict significant advancements in computing capabilities in the next decade, potentially leading to artificial general intelligence (AGI). Such predictions imply that AI technology might evolve to exhibit not only problem-solving skills but also a semblance of creativity and imagination. In light of these developments, there is a pressing need for human beings to cultivate their virtues through education and practice, ensuring they remain relevant and superior in their unique capabilities against the backdrop of advancing AI technologies.

Opinions

You've reached the end