OpenAI has unveiled its latest AI technology, OpenAI o1, which enables its chatbot to reason through complex tasks in mathematics, coding, and science, The New York Times reported.
The project aims to develop systems that solve problems carefully and logically, following a series of step-by-step processes that build on each other, much like human reasoning. According to the report, this technology could be particularly beneficial for computer programmers who use AI to write code, and it could also enhance automated tutors for math and other subjects.
OpenAI also highlighted that its new technology could assist physicists in generating complex mathematical formulas and support healthcare researchers in conducting experiments.
‘Previous models like ChatGPT would respond immediately to questions,’ said Jakub Pachocki, OpenAI’s chief scientist. ‘This model can take its time, think through the problem in English, and break it down systematically to find the best answer.’
Access to this new technology became available on Thursday for subscribers of OpenAI’s ChatGPT Plus and ChatGPT Teams services, as well as for businesses and developers building their own A.I. applications.
In a demonstration for The New York Times, Dr. Pachocki and Szymon Sidor, an OpenAI technical fellow, showcased the chatbot solving an acrostic, a complex word puzzle, answering a Ph.D.-level chemistry question, and diagnosing an illness based on detailed patient information.
OpenAI reported that this new technology significantly outperformed previous models on standardized tests. On the International Mathematical Olympiad (IMO) qualifying exam, the premier math competition for high school students, OpenAI’s previous technology scored 13%, whereas OpenAI o1 achieved a score of 83%.