Openai has recely unveiled a significa progress in the field of artificial ielligence that can change the equations in the field of logical reasoning. An unpublished artificial ielligence model of the company has managed to perform at the World Mathematics Olympiad (IMO). The news was first announced by Alexander Wei, a research scieist in Openai, who works on large language models and their ability to reason, on the social network X. He described this as a response to one of the “great and old challenges in artificial ielligence”.
According to him, the experimeal model has been able to answer 5 questions out of 6 questions in this valid competition and score 35 out of 42. To better understand this success, we must know that the World Mathematical Olympiad is one of the oldest and most difficult scieific competitions in the world. In this eve, couries will send teams of 6 elite studes to solve extremely complex issues in the field of algebra and accouing. These issues, though seemingly simple, need creativity and innovation to get the highest score. In this year’s competition, out of 630 participas, only 67, about 10 %, won the gold medal, which shows the high level of performance of the Openai model.
Artificial ielligence usually performs well in the processing of massive datasets and repetitive tasks, but often challenges to solve issues that require complex creativity or decisions. However, Openai’s rece success shows that its new model can handle complex mathematical issues with a human -like reasoning. “We have achieved a model that can provide complicated and accurate argumes at the level of human mathematicians,” Alexander, he writes. Both he and Sam Altman, CEO of Openai, emphasized that the model would not be publicly released for several mohs. This means that although the company’s next model, the GPT-5 will improve compared to its predecessor, it will not have such a significa ability to compete in the World Mathematical Olympiad at the time of launch.
Source: Engadget
