The best answer about surgery
Like every student who graduated from the 6-year medical school, Sber’s artificial intelligence was tested and he answered the questions on the ticket. The final score is 4.
A standard oral exam ticket includes three situational tasks in therapy, surgery, obstetrics and gynecology and 3-5 questions on them (“state the expected diagnosis”, “make a treatment plan”, “prescribe additional examinations” and so on). GigaChat also passed the 100-question test. It scored 82% against a 70% passing threshold.
“From my perspective, everything went great because I saw almost a year ago what our student could do. And to be honest, I was a little worried today. And I must say, the progress he has made with our help is absolutely amazing. Because today was an unexpected result for me,” said the professor of the department of faculty therapy with the clinic of the National Center for Medical Research. VA Almazova Olga Bolshakova.
In his opinion, four is a perfect score for a first try.
GigaChat has a similar relevance to surgical diseases, he added. And according to the professionals, the surgery response was the best, but the therapy response was weaker.
“But I will repeat that compared to where we started, of course, it has come a very long way and has made very good progress. But since therapy is probably the broadest specialization, of course there is something to work on. Thank you,” the professor said.
unusual experience
Head of the Faculty Department of Surgery with the National Medical Research Center clinic bearing his name. VA Almazova Ivan Danilov admits that the experience was unusual.
“This is unusual, because in any exam, when there is person-to-person communication, you see the emotions of the interlocutor, how he reacts to the questions. Stuttering or not stuttering when answering. There is an unusual feeling here, no hesitation, just a short delay to think. And then a completely straight answer emerges. This is unusual. It’s a very good experience for me too. “And of course it’s something we need to strive for more,” he said.
It turns out that the first problem for AI is obstetrics-surgery, while the second problem is purely surgical.
“It seemed to me that surgery was the most complete and best answer, especially the second one. When it came to therapy, it was a little weaker, but there were also questions that were more comprehensive in volume. We still have more specific questions and comments about the surgery. I liked the way I answered, because the second question about surgery was definitely a solid A and two solid Bs from the other questions,” Danilov noted.
He added that some of the answers were overly detailed and large, and some were unnecessarily unnecessary, and that they predicted analysis and examination methods that were not used in surgery due to lack of time.
beyond expectations
“Probably to the surprise of the entire committee, and I share the view that it has exceeded even the expectations of my colleagues. But of course, we have come this far from the very beginning, we have naturally watched it develop and have always been pleased with the result. And what we saw today, of course, in many respects exceeded even our wildest expectations,” said Deputy Director General of the National Medical Research Center VA Almazova in information technology and project management Dmitry Kurapeev.
The admissions committee member believes that the neural network fully deserves the “good” rating.
“There is something to work on. This is due, first of all, to specialization and diversity of fields. Therefore, I hope just like our student. Our language model is broad, he will continue the specialization, perhaps graduate from school and become a truly versatile, first-class specialist,” concluded Dmitry Kurapeev.