基于伦理嵌入的人工智能新型风险治理体系建构及治理路径解析

Study on Avoiding New Risks of Artificial Intelligence and Analysis on the Management Systems of AI on the Basis of Ethics

人工智能在释放巨大潜能、产生经济价值和带动广泛的产业变革同时,也产生了诸多风险。而现有治理体系难以适应人工智能的高度开放性和不确定性,无法满足风险治理需求,同时人工智能技术应用引发的新兴风险治理议题超越了既有治理体系范围,传统风险治理体系急需向以伦理嵌入为关键支撑的新兴风险治理体系迁移。从伦理嵌入视角出发,构建了新型风险治理体系,摒弃了传统以政府为绝对主导的弊端,采取多主体参与的协同治理模式,强调风险治理主体之间的"垂直连通"和"水平协商",并充分发挥伦理道德能动作用,将治理触角渗透进人工智能技术创新到规模化应用的诸多阶段,治理范围更加宽泛,治理方式更加灵活多样。同时,伦理道德通过内嵌于人工智能技术研发、技术测试、技术转移和技术产业化四个阶段,形成兼具分段治理和整体协同的风险治理路径。研究旨在降低人工智能风险等级,为发展负责任的人工智能提供理论指导和有益启示。

Artificial intelligence is fulfilling its enormous potential,generating economic value,and even bringing a wide range of industrial evolution.At the same time,it also creates many risks.The existing management system cannot cope with the uncertainty and the endless possibilities of the AI,so it cannot meet the needs of minimizing the risks.What’s more,managing the new risks triggered by the application of artificial intelligence technology goes beyond the scope of the existing management system.So it is high time that the traditional risk-management system shift to a new way of management on the basis of ethics.The authors propose that the new management systems of AI application risks should be established on the basis of the ethics;the traditional system which relied too much on the government should be replaced by a collaborative management system participating by various parties;it should be emphasized that the participants should cooperate within one field as well as in trans-field ways;professional ethics should be put into practice not only in technological innovation process but also in the mass application of the AI,therefore the ethics can be used in broader range and in a more flexible way.In this way,the new management system can be formed where ethics is used in the four different stages of AI development,such as the research and development,testing,technological conversion,and technological industrialization.The paper can help reduce the risk level of AI and provide theoretical guidance and enlightenment for developing responsible artificial intelligence.