Revolutionary Mathematics AI Solves Decades-Old Research Problems in Just Weeks

Just a week after artificial intelligence (AI) astounded mathematicians by disproving an 80-year-old conjecture, another half-century-old conjecture has been debunked, utilizing a similar technique devised entirely by human researchers.

Recently, OpenAI’s unpublished AI model disproved a critical conjecture known as the unit distance problem, first introduced by the esteemed mathematician Paul Erdős. This conjecture, which Erdős regarded as his “most significant contribution to geometry,” questions the maximum number of equal-sized connections that can be formed among points on a plane.

Erdős established a conjectured upper limit for this value, a notion many experts accepted. However, the AI model revealed that this number could be significantly larger. By employing intricate techniques from algebraic number theory, researchers could construct complex high-dimensional structures allowing arrangements of points that diverge from previous human concepts. This unexpected outcome thrilled mathematicians, some of whom doubted Erdős’ conjecture would be disproved during his lifetime.

Less than a week later, Thomas Bloom, a professor at the University of Manchester in the UK, along with colleagues, utilized a similar approach to refute another well-known theory: the sum-product conjecture first proposed by Erdős in 1976.

“I was astonished because I had been pondering this matter extensively,” Bloom stated. He and his team examined OpenAI’s innovative use of number theory for solving geometric problems, realizing they could apply the same strategy to the sum-product conjecture. “Once you recognize that something might be achievable, it encourages you to exert extra effort to implement it,” he explained.

Erdős’s sum-product conjecture posits that when numbers from a set are added or multiplied, at least one of the resulting sets must be significantly larger than the original, with both resulting sets being unable to be minimized equivalently. For instance, multiplying numbers from 1 to 5 yields a greater set than adding them, due to overlapping results from additions. Considering another set, such as 1, 2, 4, 8, 16, etc., the added set surpasses the multiplied set in size because the multiplications merely yield different powers of 2.

Erdős originally set a limitation on how small the larger of the two resultant sets can be generated through addition and multiplication, believing this was universally applicable. However, Bloom and his colleagues used advanced high-dimensional techniques to uncover sets where both the sum and product yield values smaller than Erdős had predicted. Instead of relying on geometric progressions, such as powers of 2, they created number progressions in multiple dimensions, discovering configurations with notably fewer distinct totals.

“The simplicity of this discovery was genuinely surprising to me,” Bloom remarked. “It’s straightforward to articulate the structure now, and I truly grasp why [Erdős’s conjecture] fails. This insight will also provide clarity on numerous related issues.”

“Mathematics often resembles a competitive arena,” noted Mischa Rudnev from the University of Bristol, UK. “Any new concept inspires many researchers to work relentlessly to explore further applications. Typically, these individuals are exceptionally capable and expedient.”

Rudnev indicated that Erdős’ initial intuition suggested this conjecture predominantly focused on integers, which appears accurate since Bloom and his colleagues employed complex number systems that grew increasingly intricate with larger sets. Bloom concurs, affirming that “there’s still an enormous amount of work to be undertaken, and we still lack comprehensive understanding.”

The key insight derived from this proof is that seemingly geometric problems, like those involving powers of two, can actually be addressed using number theory techniques, according to Bloom. “This revelation opens these mathematical challenges to a broader audience, particularly those in algebraic number theory who previously had little interest in these puzzles.”