To use this site, please disable the ad blocking feature and reload the page.
This website uses cookies to collect information about your visit for purposes such as showing you personalized ads and content, and analyzing our website traffic. By clicking “Accept all,” you will allow the use of these cookies.
Users accessing this site from EEA countries and UK are unable to view this site without your consent. We apologize for any inconvenience caused.
The government has launched the development of a generative AI system designed to assist doctors with their diagnoses, according to government sources.
The AI under development is expected to help improve the quality of medical services by displaying the names of possible diseases based on the results of medical consultations. The government aims to put it to practical use within the next few years.
Regarding the use of generative AI in the medical field, experts have pointed out the risk of misdiagnosis resulting from inaccurate information. Therefore, the development team will study how to address such problems.
The team, led by Ryozo Nagai, president of Jichi Medical University, started the initiative in September. About 40 research institutes and private firms, including the National Institute of Informatics, the Research Organization of Information and Systems, the University of Tokyo, Kobe University and Kyushu University, are taking part.
The team will train a large language model (LLM) on tens of billions of characters of text, including from medical papers written in Japanese for which permission has been obtained from domestic providers. The LLM will form the basis of the envisaged generative AI system.
It plans to complete the system as early as spring after additionally training it on about 520 million images, including computed tomography (CT) scans which have been anonymized.
In addition to assisting doctors with diagnoses, the team is also considering adding a function that informs doctors when there are any particularly important medical findings, for example, indications of cancer based on X-ray screening. This function could help to prevent medical malpractice as a result of oversight.
The AI is also expected to reduce doctors’ administrative workloads by helping them update electronic medical records and draft reports, letters of reference and other forms of documentation, such as notifications on infectious disease outbreaks. For patients, the use of AI is expected to increase the time they can talk with doctors.
Major overseas IT firms are also developing generative AI for medical use. However, models developed overseas are unlikely to reflect situations in Japan due to issues with the balancing of the training data. There are also concerns that personal information might leak overseas.
Generative AI is known to be susceptible to issues such as AI hallucination, in which false or misleading outputs are generated. Therefore, the team also plans to study how hallucinations are generated and how to prevent potential problems.
The about ¥22 billion in development costs was allocated in the Cabinet Office’s supplementary budget for fiscal 2023. Team members were finalized in summer last year following an open recruitment and screening process.
To prevent technology leakage, the team is using data centers within the country.
Regarding implementation, the team envisages that makers of electronic medical records will incorporate generative AI into their systems and plans to call on them to join the scheme.
