OpenAI has reached a new milestone in the development of artificial intelligence: its experimental model achieved “gold medal” results at the International Mathematical Olympiad, according to Alexander Wei, an OpenAI researcher who focuses on language modeling and logical inference. Wei said that the model completed five out of six tasks, scoring 35 out of 42 possible points. This result is in line with the performances of the top participants in one of the most challenging Olympiads in the world.
The International Mathematical Olympiad annually brings together high school students from around the world, each of whom challenges themselves to solve algebra and pre-college mathematics problems. These problems seem simple, but require unconventional thinking to achieve the highest score. This year, only 67 of the 630 participants received gold medals, about 10% of the total.
Artificial intelligence is traditionally adept at handling big data and routine operations, but it often falls short when it comes to tasks that require flexible thinking and creativity. However, this time, the OpenAI model, according to the researchers, was able to demonstrate behavior similar to human reasoning. According to Wei, the AI has learned to construct complex and flawless proofs at the level of a professional mathematician.
That said, neither Wei nor OpenAI CEO Sam Altman expect this functionality to appear in future public releases. The next version of the model, GPT-5, will likely be significantly smarter than the previous one, but it likely won’t reach this level yet.
Follow us on Google News to receive daily updates on cybersecurity. Contact us if you would like to report news, insights or content for publication.
