生成式人工智能的责任能力研究  

On the Capacity for Responsibility of Generative Artificial Intelligence

作　　者：袁曾 Yuan Zeng

机构地区：[1]上海大学法学院 [2]最高人民检察院法治前海研究基地、南方海洋科学与工程广东省实验室(珠海)

出　　处：《东方法学》2023年第3期18-33,共16页Oriental Law

基　　金：国家社科基金后期资助项目“人工智能的法律人格与未来发展研究”(项目批准号:22FFXB026)的阶段性研究成果。

摘　　要：以ChatGPT为代表的生成式人工智能投入规模化应用后,至少带来创造性成果归属、意思表示能力确认、刑事犯罪规制、侵权损害救济、数据滥用等多领域的现实法律困境。从传统稳定的社会结构与数字社会新生风险两个维度形成了治理困境,需要从责任的角度确定何种主体应为生成式人工智能技术的决策负责。生成式人工智能已经具备类人化意识与行为能力的基本形态,在拟制主体对人类经济发挥巨大作用的现实借鉴下,可考虑由生成式人工智能自身承担部分责任,但由于其责任能力的限制以及以人为本的伦理基础,对该责任能力应当进行明确限定。通过“穿透人工智能面纱”的归责原则以及相应的配套机制构建,促使发展“负责任的人工智能”成为符合人类共同利益的选择。The large-scale application of generative artificial intelligence represented by ChatGPT will at least bring about realistic legal dilemmas in various fields,such as attribution of creative achievements,confirmation of ability of declaration of intention,punishments of crimes,remedy of tort damage,data abuse and the like.It caused governance problems for the stability of traditional social structure and new risks of a digital society.From the perspective of accountability,it is necessary to determine what entities should be responsible for making decisions by generative AI technology.Since generative AI has already developed a basic form of human-like consciousness and behavioral capacity,it can be proposed that generative AI itself bear part of the responsibility considering the fictious entity's huge impact on human economic development.However,due to the limits of its ability and human ethics,its determination of responsibility should be expressly made.By"piercing through the veil of artificial intelligence"and building corresponding supportive mechanisms,the development of"responsible Al"has become a choice in line with the common interests of mankind.

关 键 词：生成式人工智能 ChatGPT 预训练生成式聊天模型 责任能力 法律人格 大型语言模型 

分 类 号：DF51[政治法律—民商法学]