摘要
为推动生成式人工智能的健康发展,防范其潜在风险,文章构建了一个基于全生命周期的生成式人工智能治理框架,旨在通过对技术要素与主体要素的治理协同,实现技术进步与社会价值的和谐统一。首先,通过文献分析将生成式人工智能的风险分为技术安全风险、社会伦理与法律风险、信息传播风险三类,并理清这些风险在全生命周期各阶段的具体表现;其次,从数据资源、算力资源、算法模型三个方面分析治理的技术要素,并从政府、企业、社会三个层面分析治理的主体要素;再次,基于各阶段的风险表现以及技术与主体要素的分析,构建全生命周期视角下的生成式人工智能治理框架;最后,依托治理框架对生成式人工智能在虚假信息传播案例和国家监管方面的典型案例进行深入讨论,展示了该框架的实际可行性和有效性,以期为生成式人工智能的综合治理提供理论指导和实践参考。
In order to promote the healthy development of generative artificial intelligence and prevent its potential risks,this paper constructs a generative artificial intelligence governance framework based on the full life cycle,aiming to achieve the harmonious unity of technological progress and social value through the governance coordination of technical elements and subject elements.First,through literature analysis,the paper divides the risks of generative artificial intelligence into three categories:technical security risks,social ethics and legal risks,and information dissemination risks,and clarifies the specific manifestations of these risks at various stages of the full life cycle;secondly,the technical elements of governance are analyzed from three aspects:data resources,computing resources,and algorithm models,and the subject elements of governance are analyzed from three levels:government,enterprise,and society;thirdly,based on the risk manifestations at each stage and the analysis of technical and subject elements,a generative artificial intelligence governance framework is constructed from the perspective of the full life cycle;finally,based on the governance framework,the paper conducts an in-depth discussion of typical cases of generative artificial intelligence in false information dissemination cases and national supervision,demonstrating the practical feasibility and effectiveness of the framework,in order to provide theoretical guidance and practical reference for the comprehensive governance of generative artificial intelligence.
作者
黄炜
张玉滢
刘勇
张瑞
Huang Wei;Zhang Yuying;Liu Yong
出处
《图书与情报》
CSSCI
北大核心
2024年第6期73-85,共13页
Library & Information
基金
国家自然科学基金资助项目“面向大学生价值观引导的智能算法分发信息服务方法与机制研究”(项目编号:72304090)
国家自然科学青年基金资助项目“科技竞争态势下技术需求弱信号感知研究”(项目编号:72404081)研究成果之一。
关键词
全生命周期
生成式人工智能
技术治理
主体治理
治理框架
full life cycle
generative artificial intelligence
technology governance
subject governance
governance framework
