生成式人工智能责任机制的技术与法律建构

Legal and Technical Construction of Generative AI Accountability Mechanisms

以ChatGPT为代表的生成式人工智能的技术突破带来了重大的社会变革,也带来了新问题和挑战。生成式人工智能存在语言失衡、算法黑箱、算法偏见的内在缺陷,同时又有算法舒适圈、数据污染、算法侵权、不准确输出等负外部性。这些问题导致生成式人工智能难以满足正义、公平、透明的要求。对此,应充分了解生成式人工智能的显著特征和能力,在人机和谐共存的关系中通过技术方案和法律规制两个维度,围绕技术优化、伦理对齐、算法透明、算法解释与算法问责等要素建构生成式人工智能的责任机制。

The advent of ChatGPT is a major breakthrough and leap in artificial intelligence technology in which algorithms are embedded in society.Technological breakthroughs in generative artificial intelligence and large language models,represented by ChatGPT,have brought revolutionary impacts,but also brought new problems and challenges.Due to its own data mining and screening defects,algorithm black box,algorithm discrimination bias and lack of legal value judgment and empathy,as well as the algorithm comfort zone in its application,data pollution,algorithm infringement,and inaccurate output,it does not meet the responsibility or accountability characteristics of justice,fairness,transparency and legality.In this regard,it is necessary to fully understand Generative AI and its salient characteristics and capabilities,and try to build a responsibility and accountability mechanism for Generative AI around data processing and technology optimization,algorithm transparency and openness,algorithm interpretation and algorithm accountability in the harmonious coexistence of man and machine.