摘要
以生成式人工智能为代表的新一轮人工智能技术带来了一系列规制挑战,而传统的“基于权利的方法”难以有效因应。“基于风险的方法”主张根据人工智能所带来的不同类型和等级的风险,采取相应的治理措施和方式,平衡技术创新与有效治理的关系。这是一种符合人工智能风险特性和发展规律的治理方法,以风险社会、适应性治理和回应性规制等理论为指引,以风险识别与分析、风险评估与分类、风险控制与监督、风险沟通与参与作为制度构成。在基于风险的人工智能治理实践中,欧盟、美国和中国分别采取“自上而下模式”“自下而上模式”“混合模式”,并取得了初步成效。然而,人工智能治理中“基于风险的方法”也面临批判和质疑,应当构建多元共治的人工智能风险治理体系,并建立动态适应的人工智能风险治理机制,同时搭建开放协作的人工智能风险治理平台。
The latest batch of AI technologies,such as generative AI,presents regulatory hurdles that are hard to tackle using the conventional“rights-based approach”.The“risk-based approach”proposes adopting suitable governance measures and techniques that correspond to the various risks posed by AI,at different levels,to balance the interplay between technological innovation and effective governance.This is a method of governance that adheres to the norms of AI hazards and development laws.It is guided by theories of risk society,adaptive governance,and responsive regulation.The system is composed of risk identification and analysis,risk assessment and classification,risk control and supervision,and risk communication and participation.In the practice of risk-based AI governance,the EU,US,and China use the“top-down model”,“bottom-up model”,and“hybrid model”to manage risk.Preliminary results have been achieved.However,experts are questioning the“risk-based approach”to governing AI,and suggest creating a versatile and collaborative system for managing AI risks.This should include a dynamic mechanism for managing AI risks and an open platform for shared governance.
作者
张涛
ZHANG Tao(China University of Political Science and Law)
出处
《华中科技大学学报（社会科学版）》
CSSCI
北大核心
2024年第2期66-77,共12页
Journal of Huazhong University of Science and Technology(Social Science Edition)
基金
教育部人文社会科学研究青年基金项目“数字政府时代行政机关处理个人信息的规范建构研究”(23YJC820054)
安徽省哲学社会科学规划青年项目“人脸识别技术在地方政府治理中的应用风险及其法律规制研究”(AHSKQ2022D118)。
关键词
人工智能
基于风险的方法
回应性规制
风险社会
artificial intelligence
risk-based approach
responsive regulation
risk society
