My comment at the Roundtable on AI Governance
0 likes29 views
The document discusses ethical AI and AI governance. It states that ethical AI principles must align with business incentives for companies to adopt them, and that private initiatives alone will not achieve the social optimum due to positive externalities of ethical AI. It argues that government intervention is initially needed to internalize these externalities through subsidies and transparency requirements. The ideal end state is one where companies use AI to maximize profits while meeting standards, compliance is guaranteed by an impartial third party, and users trust and benefit from AI.
My comment at the Roundtable on AI Governance
- 1. What should be kept in mind is “It‘s the economy, stupid (Carville, 1992)." • Ethical AI and AI governance must outlive the current AI boom and on-going government initiatives. • In order to get companies to behave appropriately, the AI “principles” must be compatible with their business incentives. • Ethical AI utilization will generate significant external economies; thus private initiatives do not reach the social optimum. External intervention is a must at least initially. • Once the market starts recognizing the value of ethical AI, most of the externalities will be eliminated and optimal resource allocation can be achieved without external intervention. • What the government needs to do is to help to achieve a situation where AI governance can proceed on its own. • It is necessary to temporally introduce subsidies and fines to internalize positive externalities, and to put in place a mechanism to improve the transparency of AI utilization. • Also important to foster market understanding of the benefits that can be obtained when AI is implemented in a desirable manner. T.Jitsuzumi@Roundtable on AI Governance (Tokyo, Oct. 31, 2022) Ideal state we should aim for has the following 3 features: 1. Each company is using AI in a manner consistent with profit maximization while meeting the minimum standards. 2. An impartial third party can guarantees their compliance. 3. Users trust the overall situation and enjoy the benefits of AI.