Japan unveils generative AI that passes national medical exam

By Berk Kutay Gokmen

ISTANBUL (AA) – Japanese researchers developed a generative artificial intelligence system that successfully passed the national medical examination, Kyodo News reported Tuesday.

The Large Language Model, created by a team at the National Institute of Informatics as part of a government-backed project, can suggest potential diagnoses based on medical interview results.

The system was trained using an extensive dataset that includes more than 700,000 licensed medical papers, 16 million documents from reputable medical websites, as well as medical textbooks and other verified sources.

The model exceeded the passing threshold of the national medical licensing exam and achieved one of the highest scores among advanced Large Language Models developed globally.

The government plans to release the model for use by medical institutions to support clinical treatment, improve operational efficiency and encourage companies and research institutions to develop AI-powered medical tools.