生成式人工智能时代信息生态秩序的失衡风险与治理方案OA
Imbalance risks of information ecological order in the era of generative artificial intelligence and governance scheme
生成式人工智能驱动了人工智能生成内容时代的到来,变革了信息生成和传播的机制.在变革过程中,信息生态秩序的失衡风险也随之产生.受到语料库、模型目标设定、模型微调中相关主体价值观念等内外因素影响,生成式人工智能存在输出和传播虚构信息、错误信息、误导性信息的问题,会对现有信息生态的有序性、真实性、可信性产生影响.为应对失衡风险,应在包容审慎、动态敏捷的柔性治理理念指引下,调动政府部门、运营管理主体、产品使用者等多元主体共同参与风险治理;针对模型训练、服务嵌入、内容生成、内容发布、信息传播等具体阶段和场景展开分层分类精细治理;综合采取法律规制、伦理规制、行业自律、技术规制等多种治理方案展开立体化治理.
Generative artificial intelligence drives the era of artificial intelligence generating content,and changes the mechanism and process of information generation and dissemination.In the process of change,it will lead to the risk of imbalance of informa-tion ecological order.Influenced by internal and external factors such as corpus,model goal setting and related subject values in model finetuning,generative artificial intelligence has the problems of outputting and disseminating fictitious information,false in-formation and misleading information,which will affect the order,authenticity and credibility of the existing information ecology.In order to cope with the risk of imbalance,multi-subjects such as government departments,operational management entities,and product users should be mobilized to jointly participate in risk governance under the guidance of the flexible governance concept that is inclusive,prudent,dynamic and agile.For specific stages and scenarios such as model training,service embedding,con-tent generation,content publishing and information dissemination,hierarchical and classified fine governance should be carried out.Finally,we should comprehensively adopt various governance schemes such as legal regulation,ethical regulation,industry self-discipline and technical regulation to carry out three-dimensional governance.
王文玉;代金亮
西南大学 法学院,重庆 400700西南大学 法学院,重庆 400700
计算机与自动化
生成式人工智能信息生态秩序失衡风险风险治理审慎包容
generative artificial intelligenceinformation ecological orderimbalance riskrisk managementprudent and com-prehensive
《网络安全与数据治理》 2025 (4)
40-45,6
国家社会科学基金青年项目(24CFX048)
评论