1. Home
  2. / Science and Technology
  3. / OpenAI’s AI solves a mathematical problem that spanned decades, surprises experts, and raises a frightening question: have machines already started making scientific discoveries on their own?
Reading time 5 min of reading Comments 0 comments

OpenAI’s AI solves a mathematical problem that spanned decades, surprises experts, and raises a frightening question: have machines already started making scientific discoveries on their own?

Written by Noel Budeguer
Published on 22/05/2026 at 12:17
Updated on 22/05/2026 at 12:18
Be the first to react!
React to this article

The turnaround that toppled a historical expectation

A discovery attributed to an internal model of OpenAI has shocked the world of mathematics. The artificial intelligence is said to have found a new construction for one of the most famous problems in discrete geometry, linked to the legendary mathematician Paul Erdős, challenging an assumption that had resisted for decades. According to information released by The Guardian, the breakthrough involves the so-called unit distance problem in the plane.

The case gained attention because it is not just about a machine assisting human researchers. The model is said to have produced original mathematical ideas, sparking an explosive discussion about how far autonomous artificial intelligence can go when it ceases to be just a tool and starts acting as a kind of independent researcher.

The problem that seemed simple but stumped mathematicians for decades

The central question seems almost childish: if we place several points on a sheet, how many pairs of these points can be exactly one unit apart?

Behind this simple formulation lies one of the great challenges of combinatorial geometry. Paul Erdős, one of the most productive mathematicians of the 20th century, believed that this number would not grow much faster than the number of points placed on the plane.

In practical terms, the conjecture suggested that even by adding many points, the total number of pairs separated by exactly one unit would remain relatively controlled. However, OpenAI’s AI is said to have shown a different, much more aggressive path.

Visual representation of a network of connected points in the plane, a concept linked to the unit distance problem that gained prominence after an OpenAI artificial intelligence found a new mathematical solution.
Visual representation of a network of connected points in the plane, a concept linked to the unit distance problem that gained prominence after an OpenAI artificial intelligence found a new mathematical solution.

The turnaround that toppled a historical expectation

The model found a construction capable of generating, for infinite values of n, a number of unit distance pairs greater than imagined. The most impactful detail is that this growth would not be just slightly above expected, but significantly superior.

This difference changes the weight of the discovery. Instead of just improving an old calculation, the AI would have pointed to a new mathematical structure, capable of directly contradicting the expectation associated with the Erdős problem.

For an area accustomed to slow advances, careful reviews, and years of attempts, the episode sounds like a watershed. The message is direct: AI models may already be entering territories that previously seemed exclusive to human intuition.

The most frightening point: the machine not only calculated

The most impressive aspect is not only in the result but in the way it would have been achieved. OpenAI claims that the model did not receive a partial solution nor was it guided step by step by mathematicians.

The AI would have started from a formulation of the problem and arrived at a new construction on its own. This completely changes the debate about the role of machines in science.

For decades, computers have been used to verify cases, test hypotheses, and speed up huge calculations. But there is a brutal difference between verifying possibilities and discovering a new mathematical strategy. It is precisely this frontier that now seems to have been crossed.

Paul Erdős and the symbolic weight of this fall

Paul Erdős was not just any name. He published or collaborated on about 1,500 articles and became known for posing problems capable of mobilizing entire generations of mathematicians.

Having a conjecture associated with Erdős challenged by an AI is something symbolically gigantic. It is not just a technical victory. It is a powerful message that generative artificial intelligence can begin to compete in one of the most abstract areas of human thought.

If the demonstration is fully confirmed by the mathematical community, the episode could go down in history as one of the first great signs that machines not only repeat patterns but can also generate unexpected intellectual paths.

The impact goes beyond mathematics

This type of advancement also affects the technology market, especially sectors related to AI, advanced computing, blockchain, verifiable proofs, and decentralized infrastructure.

If artificial systems start creating complex demonstrations, an inevitable question arises: who will verify all this? Human mathematicians will still be essential, but perhaps they will increasingly act as validators, interpreters, and refiners of discoveries produced by machines.

In this scenario, technologies of formal verification, computational proofs, and even systems based on zero knowledge may gain importance. After all, a world where artificial intelligences produce proofs that are difficult to manually audit requires new forms of trust.

Caution still exists, and it is necessary

Despite the enthusiasm, caution is needed. The history of mathematics is full of supposed brilliant solutions that ended up collapsing after detailed analysis. A demonstration only becomes truly indisputable after passing the scrutiny of experts.

Even so, the signal is strong. OpenAI is not alone in this race. Google DeepMind, Anthropic, and other laboratories are also seeking models with increasingly deep reasoning capabilities.

The lingering question is uncomfortable: are we facing a brilliant tool to help scientists or the beginning of a new era where machines also start making original scientific discoveries?

A new phase of artificial intelligence may have begun

The case of the planar unit distance shows that AI is no longer seen just as a producer of texts, images, and codes. Now, it begins to appear in a much more challenging territory: that of autonomous mathematical creation.

If this advancement is confirmed, the impact will not be small. It could change the way universities, laboratories, and companies view scientific research in the coming years.

The machine that answers questions may be transforming into something much greater: a force capable of proposing answers that even the greatest experts had not found. And this, for science, for the market, and for the future of human knowledge, is simply gigantic.

Sign up
Notify of
guest
0 Comments
most recent
older Most voted
Built-in feedback
View all comments
Noel Budeguer

I am an Argentine journalist based in Rio de Janeiro, focusing on energy and geopolitics, as well as technology and military affairs. I produce analyses and reports with accessible language, data, context, and strategic insight into the developments impacting Brazil and the world. 📩 Contact: noelbudeguer@gmail.com

Share in apps
0
I'd love to hear your opinion, please comment.x