Moscow (Agencies): In a groundbreaking achievement, the GigaChat neural network model, developed by Russia’s Sberbank, has successfully conquered a medical doctor exam typically reserved for students who have completed six years of rigorous medical education.
The exam, administered by Sber’s Health Industry Center, consisted of two components: a written test comprising 100 challenging questions and an oral examination that demanded the examinee to navigate complex scenarios related to surgery, therapy, obstetrics, and gynecology. GigaChat demonstrated its prowess by scoring an impressive 82% on the written portion and skillfully addressing the situational tasks during the oral assessment.
During its six-month training period, GigaChat was fed a substantial 42 GB dataset, which included educational materials, methodological resources, articles, and anonymized medical data used by Russian medical students. The model’s ability to diagnose, create treatment plans, and recommend additional examinations showcased its potential as a valuable tool in the medical field.
Sergey Zhdanov, Director of Sberbank’s Health Industry Center, envisions a future where GigaChat serves as the foundation for a doctor-patient assistant, revolutionizing healthcare delivery. While AI excels at processing vast datasets at lightning speed, experts emphasize that it remains confined to the knowledge within its training scope. Ruslan Permyakov, an expert at the Competence Center of the National Technology Initiative, concurs, highlighting the need for continued collaboration between AI and human clinicians.
As the medical landscape evolves, GigaChat’s success heralds a new era of AI-assisted healthcare, promising innovative solutions and improved patient outcomes.