Guideline for Safe AI
The government of Japan released a draft of guidelines for taking advantage of artificial intelligence in business. Kishida administration is going to establish a new organization which deals with safety of AI this month. There is a discussion in European Union or United States, which demands AI users to disclose information with legal mandates. While Japan has been lagging behind in setting regulations on use of AI, the government is leading in making certain rule in introducing AI into the society.
The government of Japan has been believing that the development of AI would have positive impact on human society. It published a document titled Social Principles of Human-centric AI in 2019, which recognized AI as contributing to the contemporary issues such as global environment, gap between the rich and the poor, scarcity of natural resource, low birth rate, labor shortage, depopulation or increasing fiscal deficit. It raised human dignity, diversity and inclusion, and sustainability as the three principles of the society with AI.
However, the focus of the world is shifting to the risks of AI. Kishida administration established AI Strategy Conference in Cabinet Office last May. The seventh meeting of the conference in December discussed a draft of guideline from the viewpoint of risk management in the business. The government will wrap up the guideline by the end of March.
The draft of guideline raises ten principles for development, supply or use of AI. They are 1) human-centric, 2) safety, 3) equality, 4) protection of privacy, 5) security, 6) transparency, 7) accountability, 8) education and literacy, 9) fair competition and 10) innovation.
To make Ai human-centric, the draft emphasizes protecting human rights guaranteed by the Constitution of Japan or international norms. It prohibits development, supply or use of AI for the purpose of manipulating human emotion in decision making or recognition of things. It also requires necessary measures to prevent fake or biased information.
In case AI may harm life, body, property or environment of human, the draft requires the developer, supplier or user to disclose related information to everyone involved in AI. It also ensures that AI will not generate unjust discrimination based on race, gender, nationality, age, ideology or religion. It requires recording and keeping the logs of studying process, inference and the reason of decision.
Prime Minister Fumio Kishida told that Japan needs an organization for studying how to assess the safety of AI. The name of the organization newly established later this month will be AI Safety Institute. It is going to set a standard for the developers to assess the safety of generative AI. There is some discussion among the lawmakers to introduce new legislation for securing safety of AI.
Comments
Post a Comment