At a time when much of the world is concerned about artificial intelligence taking over human roles, researchers at the AMU Artificial Intelligence Centre have shown that, at least in mathematics, human expertise remains irreplaceable.
Dr Bartosz Naskręcki, Vice-Dean of the Faculty of Mathematics and Computer Science at Adam Mickiewicz University, Poznań, proved that AI is still just more advanced calculator.
He the only Polish member of an international team of 30 mathematicians, helped design a groundbreaking test that exposed the limitations of the most advanced AI models. Working alongside top minds in number theory and algebraic geometry—such as Prof. Ken Ono and Prof. Ravi Vakil (currently President of the American Mathematical Society) — Dr Naskręcki contributed to a project that set a new benchmark in AI evaluation.
The study, known as FrontierMath, evaluated AI’s ability to tackle 350 carefully constructed mathematical problems designed to test genuine reasoning. The results were clear: artificial intelligence is still far from mastering real mathematics.
Even the most capable AI model tested—OpenAI’s o4-mini—was only able to solve 6.3% of the problems. Other systems, including those from Google DeepMind, Anthropic, and xAI, failed to solve a single problem of the highest level of difficulty.
“This shows that AI is still like a very advanced calculator,” said Dr Naskręcki. “It can perform complex calculations, but it does not understand deep mathematics.”
The significance of the project has drawn the attention of leading figures in the field. Fields Medal laureate Terence Tao believes these problems will remain unsolvable by AI for many years. His fellow Fields Medalist, Timothy Gowers, agrees: the level of difficulty far exceeds anything AI has tackled to date. Prof. Igor Pak from UCLA suggests that some of these challenges may remain out of reach for artificial intelligence for up to 50 years.
Dr. Naskręcki, the sole representative from Poland and one of only five mathematicians from Europe, was part of an elite group tasked with creating a groundbreaking test:
“This is proof that Polish mathematicians remain at a world-class level and that our skills cannot be replaced by machines,” he stated.
The AMU Artificial Intelligence Centre, where Dr Naskręcki is a member of the Quantum Informatics Team, is thus becoming one of the key research centres on the frontiers of AI, not to exceed them, but to discover where these boundaries lie.
While artificial intelligence continues to make headlines across sectors—from healthcare to law—mathematics remains a stronghold of human intellect. The FrontierMath project underscores that real problem-solving, abstract thinking, and innovation are still uniquely human abilities.
Photo credit: Piotr Jabłoński
Source:
