Google and OpenAI have announced that their AI models achieved gold medals at a global mathematics competition, marking a significant advancement in AI's mathematical capabilities. This is the first instance of AI systems surpassing the gold-medal threshold at the International Mathematical Olympiad (IMO) for high school students.
Both Google and OpenAI's models successfully solved five out of six problems, utilising general-purpose 'reasoning' models that processed mathematical concepts using natural language. OpenAI's model employed a new experimental approach centred on massively scaling up 'test-time compute', allowing the model to 'think' for longer periods and deploying parallel computing power to run numerous lines of reasoning simultaneously. Google used Gemini Deep Think, a version of which was previously unveiled at its annual developer conference in May.
The achievement suggests AI is nearing the point where it can assist mathematicians in tackling unsolved research problems. However, some experts have questioned the validity of OpenAI's claim, stating that the model was not graded based on the IMO's official guidelines. Despite this, both companies believe that AI models' capabilities can be applied to research in other fields, such as physics.