Fez – For decades, one of mathematics’ most famous unsolved puzzles resisted the efforts of researchers around the world.
Now, an artificial intelligence model developed by OpenAI has achieved what many experts once believed would take years of human work, successfully solving the long-standing “unit distance problem.”
The breakthrough marks a major moment in the growing relationship between artificial intelligence and advanced mathematics.
Until recently, even the most sophisticated AI systems struggled with complex mathematical reasoning.
Over the past year, however, their capabilities have improved rapidly, allowing them to tackle increasingly difficult problems that require deep logical thinking.
According to reports, OpenAI’s model solved the puzzle without human assistance and without relying on the traditional methods mathematicians use, such as writing lengthy equations or conducting extensive manual calculations.
The result surprised not only AI researchers but also leading mathematicians.
The problem, known as the “unit distance problem,” was first proposed nearly 80 years ago by the renowned Hungarian mathematician Paul Erdős, one of the most influential figures in modern mathematics.
Erdős devoted significant attention to the puzzle and even offered cash rewards to encourage researchers to find a solution.
In simple terms, the problem asks how many pairs of points can be placed on a plane so that the distance between each pair is exactly one unit.
In 1946, Erdős demonstrated a particular arrangement of points and suggested that no significantly better arrangement could exist.
For decades, mathematicians attempted to prove his idea.
Instead, OpenAI’s model took a different path. Rather than trying to confirm Erdős’s assumption, it discovered a new arrangement that outperformed the one he had proposed, effectively disproving the long-standing conjecture.
The achievement has drawn praise from several leading mathematicians.
Noga Alon, a professor at Princeton University, described the result as the first truly independent mathematical discovery produced by AI.
He said the breakthrough was remarkable not only for the result itself but also for what it could signal about the future of AI research.
Daniel Litt, an assistant professor at the University of Toronto, called the solution a milestone in AI-driven mathematics.
He noted that if a human mathematician had submitted the work to one of the world’s leading mathematics journals, he would have strongly supported its publication.
Meanwhile, Fields Medal winner Timothy Gowers said the development suggests it may become increasingly difficult for humans to compete with artificial intelligence in solving certain mathematical problems.
Researchers at OpenAI admitted that they originally presented the puzzle to the model merely as a test of its capabilities and did not expect such a significant result.
After obtaining the solution, they spent considerable time checking for mistakes, consulting outside experts, and using additional AI tools to verify the proof.
The verification process ultimately confirmed that the solution was both correct and highly innovative. What makes the achievement particularly notable is the model’s unconventional approach.
While many human researchers focused on proving Erdős’ original assumption, the AI explored alternative possibilities and identified a completely different route to the answer.
This willingness to challenge established thinking appears to have been key to its success.
The breakthrough shows the rapid evolution of AI from a tool that assists researchers to one that may increasingly contribute original ideas and discoveries.
As AI systems continue to improve, their role in scientific and mathematical research is likely to expand, opening new possibilities while also raising important questions about the future relationship between human expertise and machine intelligence.
