Google, OpenAI earn top honours for major advancements in mathematical capabilities

#Google #OpenAI #earn #top #honours #major #advancements #mathematical #capabilities

The alphabet Google and the Open announced that their AI models won gold in the competition for global mathematics, which highlighted an important leap in the ongoing race to develop a human intelligence AI system.

This success is the first time that AI’s models have crossed a gold medal benchmark at the International Mathematics Olympia (IMO) for high school students.

Models of the two companies solved five of the six issues, which resulted in the general purpose using “reasoning” models that acted on the concepts of mathematics using the natural language, unlike the previous approach used by AI firms.

Although the Google Deep Mind worked with the IMO so that the committee could be classified and verified by the committee, the Open did not formally enter the competition. Startup revealed that his models received a gold medal on Saturday’s questions this year, citing the grade of three external IMO medalists.

According to the Junyuk War, a researcher at Brown University’s mathematics professor and Google’s Deep Mind AI unit, it successfully shows that AI is less than a year to break the field of research on the field border by mathematics.

“I think when we can solve the problems of strict reasoning in the natural language, it will enable the potential for mutual cooperation between AI and mathematics,” Jung told Reuters.

Openi’s progress was achieved with a new experimental model, based on a mass scaling “Test Time Computer”. According to Open researcher Nam Brown, it was done by allowing both models to “think” for a long time and allow parallel computing power to run several lines of reasoning simultaneously. Brown refused to say how much it costs in computing power, but he has called it “very expensive”.

Open researchers’ IT, this is another clear sign that AI models can order extensive reasoning capabilities that can spread from mathematics to other areas.

This hopefulness has been shared by Google researchers, who believe that the capabilities of AI models may be applied to other fields such as physics, which won the IMO Gold Medal in 2003 as a student.

Of the 630 students who participated in the 66th IMO at the Sanshine Coast in Queensland, Australia, 67 contested, or about 11 %, scored a gold medal.

Google’s Deep Mind AI unit last year received a silver medal using the AI system for mathematics. This year, Google used a general purpose model called Gemini Deep Think, a version of which was unveiled at its annual developer conference in May.

The company said in a blog post, contrary to the previous AI’s efforts, which relies on formal languages and long counts, Google was running completely in natural language this year and problems were resolved in the official boundaries of 4.5 hours, the company said in a blog post.

According to a post by researcher Alexander V on social media platform X, the open came, which has a set of models of his reasoning, as well as an experimental version for the competition.

For the first time this year, the competition formally coordinated with some AI developers, who for many years used prominent mathematics like IMO to examine the model’s skills. IMO judges confirmed the results of these companies, including Google, and asked them to publish the results on July 28.

“We respected the original request of the IMO Board that all AI labs share their results only after confirming official results by independent experts, and students were properly attributed to which they deserve.”

Open, who published its results on Saturday and claimed the first gold medal status, said in an interview that he was allowed to do so with a member of the IMO board after the closing ceremony on Saturday.

The president of the IMO’s board, Gregor Dullenar, told Reuters, on Monday allowed companies to publish results.