Google DeepMind achieved its first gold medal at the International Mathematical Olympiad (IMO) using its Gemini model in Deep Think mode, solving five out of six challenging problems across algebra, combinatorics, geometry, and number theory. This achievement, earning 35 out of 42 points, highlights the capabilities of the model, which functioned entirely in natural language without external aids. Unlike previous approaches that used formal languages, this version showcased a significant advancement in AI reasoning and theorem-proving, trained with specialized methods to promote multi-step problem-solving. Notably, OpenAI also claimed an IMO gold, further indicating a transformative shift in AI capabilities. Both breakthroughs demonstrate that advanced AI can tackle complex mathematical challenges consistently, marking a significant milestone in the development of reasoning-capable AI. This emerging phase raises questions about scalability and architecture but reinforces that machines are approaching human-level performance in mathematics.
Source link
Share
Read more