摘要
从全球范围来看,利益相关体已经提出了诸多人工智能治理原则以试图通过"软体系"的方式应对人工智能发展和应用所引发的伦理问题与安全风险。本文指出,人工智能的主体性挑战、代码作为规则重要性的提升以及"黑箱性"的存在构成了以"软体系"应对人工智能伦理问题与安全风险的特殊原因,而推进理念共识、聚焦局部议题、创新治理机制则构成了其具体内涵。虽然现有方案有利于推动人工智能伦理问题与安全风险的治理进程,但其在关键概念上的模糊、具体内容上的缺失或争议以及决策过程的不足,也引导人们产生更多反思。中国近年来在三个方向积极推进并于2021年初形成了《网络安全标准实践指南——人工智能伦理安全风险防范指引》的阶段性成果,该文件界定了人工智能治理的相关主体范围,总结了人工智能伦理问题与安全风险的具体类型,并明确了相关主体所应履行的治理职责。结合全球比较和中国实践的分析,文章进一步对未来人工智能伦理问题与安全风险治理的发展可能提出了三点展望。
Global stakeholders have proposed many artificial intelligence(AI)guidelines in recent years,aiming to confront the ethical challenges and safety risks of AI that have been caused by its special characteristics through a"soft"system.First,AI might induce principles without the guidance of human capacity for expression,thus forming the characteristics of subjectivity.Second,AI might empower the code to work as the laws governing human behavior,complicating the reasons causing the risks.Third,AI may obfuscate the inner mechanism of technology,leading to the problem of unexplainability.The global guidelines mainly emphasize three points.First,stakeholders are trying to realize a consensus on values guiding the development and application of AI.Second,stakeholders are focusing on specific risk issues of AI and are proposing corresponding governance solutions.Third,stakeholders may make innovations in the structure and principles of the governance framework to confront the new challenges.Furthermore,the existing guidelines are criticized because of their fuzziness about key concepts and their lack of specific content and their undemocratic process.Following the three approaches for building AI governance systems in China,〈The Guidelines for Practicing Network Safety Standards:Guidelines for the Prevention of AI Ethical Challenges and Safety Risks〉were issued in early 2021.The main content of the guidelines can be summarized by the following three points.First,the guidelines define the related stakeholders,including the researchers,producers,and deployers as well as the users.Second,the guidelines summarize the five categories of challenges and risks,including uncontrollable risks,social risks,illegal risks,discrimination risks,and accountability risks.Third,the guidelines propose general principles and specify the governance responsibilities of each stakeholder.Based on a comparison of global proposals and an analysis of efforts in China,this article points out three future trends for the development of AI ethics and security risk governance.
作者
贾开
薛澜
JIA Kai;XUE Lan(School of Public Affairs and Administration,University of Electronic Science and Technology of China;School of Public Policy and Management,Schwarzman College,Tsinghua University)
出处
《公共管理评论》
CSSCI
2021年第1期122-134,共13页
China Public Administration Review
关键词
人工智能
伦理问题
安全风险
治理反思
artificial intelligence
ethical challenges
safety risks
reflection on governance
