Alphabet’s Google introduced two new artificial intelligence systems on Thursday that have shown significant progress in tackling complex mathematical problems, a challenging area in generative AI development.
Traditional AI models, which operate by statistically predicting the next word, have faced difficulties with abstract math due to its need for advanced reasoning abilities similar to human intelligence. DeepMind, Google’s AI division, reported that its latest AI models, AlphaProof and AlphaGeometry 2, successfully solved four out of six questions at the 2024 International Math Olympiad, a prestigious competition for high school students. According to a blog post by Google, one question was solved within minutes, while others took up to three days, exceeding the competition’s time limit. Nevertheless, these results represent the highest performance by an AI system in the competition to date.
AlphaProof, a system designed for reasoning, was created by combining Gemini, the language model behind Google’s chatbot, with AlphaZero, an AI system that has previously outperformed humans in board games like chess and Go. AlphaProof solved three of the competition’s problems, including the most difficult question, which only five out of more than 600 human contestants managed to solve. AlphaGeometry 2 solved an additional math problem.
Earlier in July, Microsoft-backed OpenAI was working on a reasoning technology project named “Strawberry,” previously known as Q*. This project was considered such a significant breakthrough that several staff researchers wrote to OpenAI’s board of directors in November, warning that it could pose a threat to humanity.
