Latest news with #AlphaProof
NDTV
6 hours ago
- Science
- NDTV
World's First AI Model Wins Gold At International Math Olympiad. Check Details
Google's artificial intelligence (AI) research arm DeepMind has won a gold medal at the International Mathematical Olympiad (IMO), the world's most prestigious competition for young mathematicians. It is the first time a machine has solved five of the six problems in algebra, combinatorics, geometry, and number theory -- signalling a breakthrough in math capabilities of AI systems that can rival human intelligence. IMO problems are known for their difficulty, and solving them requires a deep understanding of mathematical concepts -- something which the AI models had not been able to achieve up until now. However, an advanced version of Gemini Deep Think managed to ace the competition where 67 contestants, or about 11 per cent, achieved gold-medal scores. "We can confirm that Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points, a gold medal score. Their solutions were astonishing in many respects. IMO graders found them to be clear, precise and most of them easy to follow," said IMO President Dr Gregor Dolinar. Last year, DeepMind's combined AlphaProof and AlphaGeometry 2 systems achieved the silver-medal standard in the competition, but it took two to three days of computation. This year, the advanced Gemini model operated end-to-end in natural language and managed to produce the results within the 4.5-hour competition time limit. The DeepMind team trained the advanced Gemini model on novel reinforcement learning techniques that can leverage more multi-step reasoning, problem-solving and theorem-proving data. "We'll be making a version of this Deep Think model available to a set of trusted testers, including mathematicians, before rolling it out to Google AI Ultra subscribers," DeepMind CEO Demis Hassabis wrote on X (formerly Twitter). Prior to Google, an OpenAI researcher also claimed that the startup had built technology that achieved a similar score on this year's questions, though it did not officially enter the competition. 1/N I'm excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world's most prestigious math competition—the International Math Olympiad (IMO). — Alexander Wei (@alexwei_) July 19, 2025 The advancements shown by the AI systems suggest that the technology was less than a year away from being used by mathematicians to crack unsolved research problems at the frontier of the field. "I think the moment we can solve hard reasoning problems in natural language will enable the potential for collaboration between AI and mathematicians," Junehyuk Jung, a math professor at Brown University and visiting researcher in DeepMind AI unit, was quoted as saying by Reuters.
Business Upturn
a day ago
- Business
- Business Upturn
Google DeepMind's AI earns gold medal at International Math Olympiad
Google DeepMind has achieved a major breakthrough in AI performance by reaching gold medal status at the prestigious 2025 International Mathematical Olympiad (IMO), one of the world's toughest math competitions for high school students. The AI system, a new and improved version of DeepMind's Gemini Deep Think, scored 35 out of 42 points, solving five out of six complex math problems with perfect accuracy. This marks a significant jump from last year, when DeepMind's earlier AI models (AlphaProof and AlphaGeometry 2) earned a silver medal with 28 points. The IMO's President, Prof. Dr. Gregor Dolinar, confirmed the achievement, praising the AI's solutions as 'clear, precise,' and often 'easy to follow.' That's a major compliment, given that the Olympiad problems are designed to challenge some of the brightest young minds on the planet. What makes this feat even more impressive is how the AI approached the problems. Unlike earlier systems, which needed human experts to rewrite the questions into computer-friendly formats, Gemini Deep Think worked entirely in natural language, reading and understanding the exact same problems given to human contestants, and solving them within the official 4.5-hour time limit. The model achieved this by using advanced reasoning tools, including something DeepMind calls 'parallel thinking,' which lets the AI explore multiple problem-solving strategies at the same time, similar to how skilled mathematicians think through different angles before committing to a solution. The system was also trained using new reinforcement learning techniques and a curated set of high-quality math solutions to boost its problem-solving abilities. Google plans to make this version of Gemini available first to trusted testers, including professional mathematicians, before rolling it out to Google AI Ultra subscribers. The development comes amid growing competition in the AI space. OpenAI also announced that one of its experimental models hit the same score, 35 out of 42, in the same IMO challenge. With both companies now demonstrating gold-level performance on a human math competition, this signals a major leap forward in AI's ability to reason logically, handle abstract problems, and produce rigorous, formal proofs. Google CEO Sundar Pichai described the progress as 'just astounding,' while DeepMind CEO Demis Hassabis called it 'incredible progress' in AI's ability to handle high-level intellectual tasks. While the AI isn't replacing mathematicians anytime soon, its success opens up new possibilities for AI-assisted research, education, and mathematical discovery, and represents a milestone in machines tackling one of the most human of challenges: creative problem-solving through logic.
