Alphabet has announced that a new AI system developed by Google DeepMind has been trained to solve intricate math problems featured in this year’s International Mathematical Olympiad (IMO), achieving a score comparable to that of a silver medalist for the first time.
The IMO is a challenging math competition for high school students from around the globe.
During the two-day competition, teams of six exceptionally talented young mathematicians are tasked with solving six difficult problems. Each problem is worth a maximum of seven points, making the perfect score 42 points in total. This year, the USA Maths team clinched the top spot, with five out of six participants securing a minimum of 29 points, which is required to earn a gold medal.
Google DeepMind’s most recent system, named AlphaProof, was not far behind. Even though it did not solve the six IMO problems within the time limit of the competition, it achieved 28 points – the highest score that AI has attained to date.
Companies such as Google DeepMind argue that solving complex abstract math problems is crucial for developing the logic and reasoning skills needed for artificial general intelligence, in which technology surpasses humans in most tasks.
The new system represents an advancement compared to the company’s prior AlphaGeometry model. It has successfully resolved 83% of all IMO geometry problems over the past 25 years, an increase from the previous rate of 53%, which includes the most challenging geometry problem ever presented in the competition.
AlphaGeometry was limited to handling only geometry, while AlphaProof is capable of addressing other areas of mathematics, such as number theory, algebra, and combinatorics.