A neural network model developed by Russia’s Sberbank has succeeded in passing the medical doctor’s exam given to students completing six years of medical school, Sberbank’s press service told Izvestiya newspaper on Tuesday.
GigaChat scored 82% on the 100-question written test and 82% on the oral exam, which included 3 scenario-based questions from various specialties ranging from surgery to gynecology. In one of the scenarios, the program was asked to make a diagnosis, create a treatment plan and prescribe additional tests.
The developers of the model stated that in order for the algorithm to pass the test, they used a 42 GB dataset of educational and methodological materials, articles, guides and anonymized health data for medical students in its training for 6 months.
Experts emphasize that artificial intelligence still cannot replace doctors for the time being. It is thought that GigaChat can be used as a doctor and patient assistant in the future, creating new possibilities in healthcare and easing the workload of healthcare professionals.
Sergey Zhdanov, Director of the Sberbank Healthcare Industry Center, said that the GigaChat model could be used in the future as a doctor and patient assistant, creating new possibilities in healthcare and easing the workload of healthcare professionals.
However, while AI can examine a given data set at superhuman speed, it cannot respond to situations outside that data set. This is seen as a weakness of all current AI systems.
Ruslan Permyakov, an expert at the National Technology Initiative Competence Center, said: “Let’s say we have a system that has learned to classify apples and pears… Such a system can determine what an apple is, what a pear is and the stage of ripeness of the fruit much faster than a human. But if we ask it to analyze an eggplant, we can 100% guarantee that the system will not say ‘I don’t know’, because it has not learned to say that.”
Sber is Russia’s largest banking and financial services company. The company launched GigaChat in April 2023 as an alternative to ChatGPT. As of January 2024, the chatbot has more than 2.5 million users.
AI programs such as ChatGPT are excelling in tests for humans in fields such as law, business and medicine. For example, the Russian YandexGPT 2 neural network passed the country’s Unified State Examination in October, which includes a literature test that is considered to be very difficult for algorithmic networks to solve.
GigaChat’s success once again demonstrates the potential of AI in the medical field. In the future, AI is expected to help improve doctors’ ability to diagnose and treat, accelerate medical research, and provide patients with more personalized healthcare.
Some highlights from the news report:
GigaChat’s 82% score in the medical doctor exam shows the potential of artificial intelligence in the medical field.
It is thought that GigaChat could be used as a doctor and patient assistant in the future, creating new possibilities in healthcare and easing the workload of healthcare professionals.
For now, AI still cannot replace doctors, but it can help improve their ability to diagnose and treat patients.
source: Izvestia.
