Google, Openai win gold in the Mathematical Gymnasium Competition

Alphabet’s Google and Openai said their artificial-law models won gold medals in a global mathematics competition, marking a major discovery in mathematics in the struggle to build strong systems that can compete with human intelligence.

The results marked the first time Ai Systems crossed the gold-medical rating threshold in International Mathematics Olympia for high school students. The models of the two companies were solved by five of the six problems, achieving the result using general purpose models elaborated by mathematical concepts using natural language, as opposed to the previous approaches used by AI companies.

The achievement states that AI is less than a year from being used by mathematicians to break unresolved research problems at the field border, according to Junehyuk Jung, a math professor at Brown University and visiting researchers at Deepmind’s Deepmind Unit.

“I think the moment we can solve the harsh problems of logic in natural language will allow the possibility of cooperation between AI and Mathematics,” Jung told Reuters.

The same idea can be applied to research projections in other areas such as physics, said Jung, who won a gold medal IMO as a student in 2003.

Of the 630 students participating in the 66th IMO on the Sunshine coast in Queensland, Australia, 67 contestants or about 11 %, achieved gold-medical scores.

Google’s Deepmind AI unit last year achieved a silver medal that uses AI systems specialized for mathematics. This year, Google used a general -purpose model called Gemini Deep Think, a release of which was previously presented at the Annual Developer Conference in May.

Unlike previous AI efforts based on formal languages and long calculation, Google’s approach this year has worked entirely in natural language and solving problems within the official 4.5 -hour time limit, the company said in a blog position.

Openai, which has its own set of reasoning models, also created an experimental edition for the competition, according to a position by researcher Alexander Wei on the social media platform. He noted that the company is not planning to release anything with this level of mathematics for several months.

This year it marked the first time the competition was officially coordinated with some AI developers, who for years used prominent mathematics competitions such as IMO to test the model capabilities. IMO Courtges certified the results of these companies, including Google, and asked them to publish results on July 28.

“We respected the original request by the IMO Board of Directors that all AI Labs were sharing their results only after official results were verified by independent experts and the students had rightly received the precision they deserved,” said Google Deepmind Demis Hassabis CEO.