On July 21, Google DeepMind announced its software scored at the gold-medal level in the International Mathematical Olympiad, showing significant progress. Previously, the AI system required human experts to translate problems and solutions, but this year it utilized a large language model called DeepThink, designed for natural language processing. DeepThink scored 35 out of 42 points, with its solutions graded by the same judges as the human competitors. This represents a significant shift in AI capabilities for mathematics as it processes arguments end-to-end without human intervention.
DeepMind's AI system achieved a gold-medal score in the International Mathematical Olympiad, signifying significant progress in solving complex mathematics problems with natural language.
This year's approach utilized DeepThink, a large language model with enhancements that improved its ability to produce mathematical arguments and process multiple chains of thought.
DeepThink scored 35 out of 42 points on the Olympiad problems, showcasing a 'big paradigm shift' where solutions were produced end-to-end in natural language.
The previous year's performance required human intervention for problem translations, illustrating the advancement in AI capabilities for handling maths problems directly.
Collection
[
|
...
]