EN, essay

AI Governance in 2020: A Year in Review in Japan

From Human-Centric to Planetary-Scale Problem Solving: Challenges and Prospects for AI Utilization in Japan

Society 5.0, a human-centric society that simultaneously achieves economic development and resolution of social issues through a system that integrates cyberspace and physical space. The Cabinet Office released the “Social Principles of Human-Centric AI” in March 2019. Since then, relevant ministries and agencies have formulated AI-related guidelines, such as guidelines for the development of medical diagnostic imaging support systems (Ministry of Health, Labor and Welfare), guidelines for contracts in the agricultural sector (Ministry of Agriculture, Forestry and Fisheries), a handbook for AI utilization (Consumer Affairs Agency), and certification systems for AI education (Ministry of Education, Culture, Sports, Science and Technology and Ministry of Economy, Trade and Industry and CSTI).

On the other hand, the response to COVID-19 exposed the fact that we were completely unprepared for the transition to a Society 5.0 society. The fact that government and private companies signed an agreement on COVID-19 to promote the use of data can be regarded as a step forward. However, data sharing between the central and local governments and medical institutions was found to be ineffective. In addition, we found out that the Tokyo Metropolitan Government was using fax machines to report the number of infected people [1].

The problem of data sharing is not limited to COVID-19, but has been pointed out previously. Especially in Japan, there are more business-to-business companies than business-to-consumer companies. In other words, there are many cases where data acquisition, AI model development, and service providers are different. Therefore, issues such as AI safety and fairness must be addressed not only by one company, but by all parties involved in a multi-layered dialog. Moreover, start-up companies lack resources, making it difficult for them to deal with risks [2]. From this perspective, the Japan Deep Learning Association, whose members are mainly start-ups, is considering how to assess and respond to AI systems in the external environment, including such as insurance, auditing, accident investigation, consumer protection, whistle-blowing systems, and standardization, in addition to AI governance within companies since the summer of 2020 [3].

Last but not the least, AI governance must be discussed internationally. The 2nd French-German-Japanese AI Symposium was organized in November 2020 [4]. The joint statement of the first conference held two years ago emphasized a human-centric approach. Therefore, in line with this direction, the theme of the second conference was human-centric AI [5]. However, we are now facing planetary-scale challenges, such as the COVID-19 pandemic, climate change, and community fragmentation. Therefore, the joint statement in 2020 proposed that planetary-scale problems, from anthropocentric to environmental, should be addressed. It is important to work on AI governance for planetary-scale issues in collaboration with various stakeholders and organizations [4]. 

[1] Arisa Ema, Challenges of AI and Data Utilization and Governance in Japan Emerging from the COVID-19 Response, September 2020, https://ifi.u-tokyo.ac.jp/en/project-news/5138/
[2] Takashi Matsumoto, Arisa Ema, RCModel, a Risk Chain Model for Risk Reduction in AI Services, July 2020, https://ifi.u-tokyo.ac.jp/en/news/4815/, http://hdl.handle.net/2261/00079367, https://arxiv.org/submit/3262407
[3] JDLA, AI Governance Study Group, https://www.jdla.org/about/studygroup/sg01/ (In Japanese)
[4] https://www.ai-symposium-france-germany-japan.com/
[5] https://www.dwih-tokyo.org/en/activities/event-reports/1st-japanese-german-french-dwih-symposium-on-artificial-intelligence/joint-statement-of-the-german-japanese-and-french-participants-on-intensified-collaboration-in-ai/