SYDNEY: Humans have triumphed over advanced AI models from Google and OpenAI at the prestigious International Mathematical Olympiad (IMO), despite the AI systems reaching gold-level scores for the first time.
Google confirmed that its upgraded Gemini chatbot solved five out of six problems at this year’s competition, held in Queensland, Australia.
“We can confirm that Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points — a gold medal score,“ the company stated, quoting IMO president Gregor Dolinar.
OpenAI also announced its experimental reasoning model achieved a gold-level 35 points.
Researcher Alexander Wei called it a “longstanding grand challenge in AI” at “the world’s most prestigious math competition.”
However, neither AI matched the flawless performance of five human participants who scored full marks.
Around 10 percent of contestants earned gold medals, highlighting the continued edge of human problem-solving in complex mathematics.
Google improved significantly from last year’s silver-medal performance, solving problems within the 4.5-hour time limit compared to days of computation in 2024.
The IMO noted that tech firms privately tested closed-source AI models on the same problems faced by 641 students from 112 countries.
IMO president Dolinar praised the AI advancements but cautioned that organisers could not verify computing power used or potential human involvement in the AI submissions. – AFP
