This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: A few short years ago, asking AI to add two four-digit numbers together for a third-grade math assignment would have been a bad idea. Over half the time, the answer would be wrong, studies have shown.
Long division or a "leaky tank" problem, an algebra class staple? Although AI still struggles with arithmetic, in 2024, Google DeepMind's AI systems scored high enough for a silver medal in an elite high school competition that tests mathematical reasoning rather than basic calculation. A year later, OpenAI grabbed the gold.
In May 2026, AI made headway on a question that mathematician Paul Erdős posed in 1946. Given a fixed number of dots on a flat surface, how many pairs can sit exactly one unit apart? Experts had a long-standing prediction, but the bot overturned it once and for all.
In a crowning moment in September, OpenAI announced that its agents found a solution to the Navier–Stokes problem, which describes how fluids move. The answer could help scientists understand when computer predictions of air or blood flow become unreliable. The 88 hours it reportedly took might sound like a lot.
Still, it's nothing compared to the roughly 90 years people spent trying to put the question to rest. These well-defined problems make for useful test cases, Northeastern mathematics professor Ken Duffy told Northeastern Global News. So are mathematicians about to get taken out of the equation?
Experts in mathematics and computing put the breakthroughs in perspective. Here are five takeaways from the conversations. Schools and universities aim to equip future experts with the tools of the trade.
Extract — continue reading at the source.
