Revolutionary OpenAI o1 Achieves Perfect Score in ITA Exam and 82% in USP Medical Residency: A New Era in AI Problem Solving

OpenAI’s new artificial intelligence (AI), so-called Open AI o1you can now get a “10” on the test Aeronautics Institute of Technology (ITA) and be approved for various specialties in the medical residency exam at the University of São Paulo (USP). The o1 was launched last week with the promise of logical and mathematical reasoning abilities and complex problem solving. Now, AI is being tested all over the world by experts and researchers.

In Brazil, entrepreneur Vinícius Soares installed the Open AI o1 good test IT 2024, where the model performed the test, showing its ability to solve complex mathematical problems. Unlike previous models, which tried to answer as quickly as possible by identifying connections between words, the Open AI o1 invests more processing time to evaluate the existing data and look for different ways to arrive at an answer, which happened in the experiment tested by Soares.

How did OpenAI o1 stand out at the Institute of Aeronautical Technology (ITA)?

The idea of ​​confirming ChatGPT o1 in the IT appeared after the entrepreneur read the publication of the new model and its improvements in the ability to solve problems related to sets, functions, geometry, trigonometry and statistics. After taking the ITA entrance exam, considered one of the most difficult in the country, Soares was curious to see how AI would fare against this challenge.

“I took the questions and entered them one by one into the new ChatGPT module” explained Soares. “Not only did he provide the correct answers to all 10 questions, he also described the reasoning used to arrive at each solution.” Comparing the alternatives it correctly identified with the answer key published by ITA, the AI ​​got 100% of the questions correct.

How did OpenAI o1 perform at the USP medical residency?

Another example to prove the potential of this new technology in the medical field made by the Medical Manager of Medical Education and Digital Health, Matheus Ferreira, who conducted a bold experiment: submitting o1 to the USP-SP medical residency exam in 2024. The AI ​​achieved an accuracy rate of 82% , even without the ability to analyze images, outperforms GPT-4 models, also from OpenAI, and Claude Sonnet 3.5, from Anthropic, which scored 76%.

The test, which consisted of 120 questions, was divided into six blocks of 20 questions each, respecting the original order. All blocks were sent to the Open AI o1along with prompt asking the AI ​​to answer the questions like a doctor and indicate the answer for each alternative.

What are the benefits of the new OpenAI approach?

oh Open AI o1 represents a paradigm shift in the AI ​​response process. Instead of focusing only on the volume of data, OpenAI invested in an architecture that allows the system to devote more processing time to evaluating existing data, looking for different ways to reach an answer. This approach, called “chain of thought”, allows AI to imitate human reasoning, evaluating each step of constructing answers and correcting errors along the way.

AI combined with human experience

The results obtained by ChatGPT Open AI o1 in the USP-SP medical residency exam open up a range of opportunities for future medical education. Matheus argues that AI should be seen as a friend in the teaching and learning process, and not as a threat. “I believe we should see AI as a partner in the educational process. Instead of trying to prevent or prohibit its use, we should encourage correct and ethical use, at appropriate times.”

The Potential of AI in Medical Education and Research

• Personal trainer: AI can be a private “teacher”, adapting to each student’s needs and learning pace. Research conducted at Harvard showed that students, when using an AI tutor, had a 22% higher grade result than those who only used the “traditional” study method.
• Regular update: The device can also help healthcare professionals stay up-to-date in a situation where medical knowledge doubles every 73 days, something humanly impossible to keep track of without the help of technology.

Ferreira sees a promising future for artificial intelligence in medicine. He believes that AI will become an essential “copilot” for doctors, helping with diagnosis, treatment and research, allowing professionals to dedicate themselves more to the doctor-patient relationship and understand aspects that cannot to capture the device.

OpenAI’s new approach with Open AI o1combining chain thinking with reinforcement learning, has revolutionized the field of artificial intelligence. By achieving impressive results in several areas, the Open AI o1 stand as a milestone in evolution technologicalopening promising avenues for the future.