This is the trainee level.
The popularity of ChatGPT and other similar neural networks is growing every day. The chatbot already writes diplomas, articles, program code, albeit at a basic level, and answers questions. The bot was reportedly able to take the next step – it passed a medical knowledge test that was used to test the skill levels of medical trainees and medical students in the United States.
The chatbot was able to correctly answer 89% of the test questions. At the same time, the examiners noted that the neural network was able to put forward a “non-obvious, but clinically valid hypothesis” in the test diagnosis.
Note that a similar PubMedGPT language model, trained exclusively on medical literature, scored 50.8% of correct answers. Although experts doubt that a chatbot has the same completeness of knowledge as a person.
This is not the first such result. The system has previously passed exams at the Wharton School of Business at the University of Pennsylvania and at the University of Minnesota.
And earlier, the chatbot successfully passed an interview at Google for a junior developer position with a salary of $183,000 per year.