生成式模型下算法风险的伦理解读与出路探析——以ChatGPT为核心  

Ethical interpretation and exploration of algorithm risks under generative models——Centered around ChatGPT

作　　者：唐俊 陈杉 TANG Jun;CHEN Shan(School of Marxism,Changsha University of Science&Technology,Changsha 410114,China;Innovation Research Center for Philosophy of Science and Technology and Ethical Governance of Science and Technology,Changsha University of Science&Technology,Changsha 410114,China)

机构地区：[1]长沙理工大学马克思主义学院,湖南长沙410114 [2]长沙理工大学科技哲学与科技伦理治理创新研究中心,湖南长沙410114

出　　处：《佛山科学技术学院学报（社会科学版）》2024年第6期46-54,共9页Journal of Foshan University(Social Science Edition)

基　　金：湖南省社会科学基金一般项目(19YBA002);湖南省教育厅重点项目(23A0236);湖南省研究生科研创新一般项目(CX20230936)。

摘　　要：以ChatGPT为代表的生成式模型的广泛应用,给人工智能领域带来了惊人的技术进步,但复杂的伦理挑战也伴随而来。生成式模型下的算法伦理问题,如算法偏见或歧视问题、数据隐私挑战以及对人类自主性的撼动,已深度触及社会的诸多方面,其负面效应凸显。究其原因主要在于算法依赖与数据局限性的矛盾、海量数据需要与隐私保护需求的冲突、算法智能化对人类自主性的消弭等三个核心方面。因此,在推进人工智能技术创新的同时,必须平衡其不利影响,着力构建负责任的AI,做到敏捷治理与创新共进,以人本主义为契机捍卫人类自主性,形成一个多主体的综合治理体系,共同构建一个负责任、可信赖、符合公共伦理标准的生成式人工智能发展生态体系。The widespread application of generative models represented by ChatGPT has brought amazing technological advancements to the field of artificial intelligence,but complex ethical challenges have also emerged.The ethical issues of algorithms under generative models,such as algorithmic bias or discrimination,data privacy challenges,and the impact on human autonomy,have deeply touched many aspects of society,and their negative effects are highlighted.The main reasons lie in three core aspects:the contradiction between algorithm dependence and data limitation,the conflict between massive data needs and privacy protection needs,and the challenge of algorithm intelligence to human autonomy.Therefore,while promoting innovation in artificial intelligence technology,it is necessary to balance its adverse effects,focus on building responsible AI,achieve agile governance and innovation,take humanism as an opportunity to defend human autonomy,form a comprehensive governance system of multiple subjects,and jointly build a responsible,trustworthy,and public ethical standard generative artificial intelligence development ecosystem.

关 键 词：生成式人工智能 算法伦理 大语言模型 ChatGPT 

分 类 号：B82-057[哲学宗教—伦理学] TP18[自动化与计算机技术—控制理论与控制工程]