Japan’s education ministry plans to develop a generative artificial intelligence program that produces medical and scientific hypotheses by learning from research papers and images of experiments, Nikkei has learned.
Relying on foreign technology could lead to technology leaks. By developing homegrown technology, the Ministry of Education, Culture, Sports, Science and Technology hopes to ensure data safety and raise national competitiveness.
The generative AI will be first used for medical and material research with other areas added in the future. Developing generative AI for one research area is estimated to cost roughly JPY 30 billion (USD 212 million). The ministry will seek funds for initial development in the fiscal 2024 budget.
The Riken research institute will lead the effort. The plan is to open up the technology for use in outside laboratories and companies on a trial basis starting from fiscal 2025.
The project is expected to last for eight years, with the technology made available for researchers nationwide from fiscal 2031.
Additional research data will be fed into the developed generative AI to make it capable of identifying substances that cause illnesses or designing materials for use in the medical or industrial fields.
Riken has a trove of accumulated research data that can enhance generative AI. Research in both fields uses massive amounts of data relating to candidate compounds or diagnostic images, making AI a good match.
The program will also assist in the preparation of papers by researching past literature. In the future, researchers will interact with generative AI to propose and test new hypotheses.
Some estimates suggest that even with existing AI, the time from the conception of an idea to the publication of a paper can be reduced to less than one-tenth in some research areas. Early adoption of generative AI could determine Japan’s future international competitiveness.
In the US, the Department of Energy and the federally funded Argonne National Laboratory announced plans in May to develop generative AI for scientific research.
US companies OpenAI and Google are leading the development of generative AI. In Japan, NTT, SoftBank and others are working on models compatible with the Japanese language. The technology is expected to improve productivity in areas related to writing, such as email, documents, and minute keeping, but Japan has not made much progress in areas applicable to scientific research.
The government will also improve the infrastructure for such research. The Ministry of Economy, Trade and Industry will introduce a new supercomputer for research in 2024. The education ministry will increase the computing power of Riken’s Fugaku supercomputer to make it easier to use in generative AI research.
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.