摘要
ChatGPT“火爆出圈”后,其快速性、同步性、多端性的人工智能生成内容受到广泛关注,过往结果取向的知识产权治理,越来越难以契合形势发展,过程型和结果型并重的风险治理正成为一种迫切需要。在对已有成果述评和风险源研判的基础上,本文基于人工智能生成内容的生态构成,经过对数据来源、算法嵌入、平台集成、用户反馈四个维度风险派生路径的分析,发现存在着披露未许可、虚假不真实、偏见放大、偏差隐藏、被恶意使用、不道德取利、理性致瘾、小概率惊恐八类风险;进一步对AI嵌入内容生成的产生原理及作用机理进行解构与深化,提取出风险治理的隐私保护、说明责任、公平无歧视性、透明可解释性、防第三方侵害、政府责任、人类控制、增进人类价值八项原则;并推演出了开展AI数据的镜像审查治理、促进算法和内容的同步治理、推动生成内容的对象治理转向智能平台的生态治理、实施内容生成和渠道推广的风险共担治理四方面的建议,建构起人工智能生成内容的系统伦理框架。
Along with the popularity of ChatGPT, its characters of speed, synchronization, and multiend have attracted widespread attention to the content generated by AI. The previous result oriented intellectual property governance has become increasingly difficult to adapt to the development of the situation, and risk governance that emphasizes both process and result oriented approaches is becoming an urgent need. On the basis of existing achievements and risk source analysis, based on the ecological composition of AIGC, through risk derivation path analysis of four dimensions: data source,algorithm embedding, platform integration, and user feedback, it is found that there are eight types of risks: unauthorized disclosure, false and untrue information, bias amplification, bias concealment,malicious use, unethical profit taking, rational addiction, and small probability of panic. By further deconstructing and deepening the principles and mechanisms of AI embedded content generation,eight principles of risk governance are extracted: privacy protection, accountability, fairness and non discrimination, transparency and interpret ability, prevention of third-party infringement,government responsibility, human control, and enhancement of human value. We propose four aspects of governance: conducting mirror review of AI data governance, promoting synchronous governance of algorithms and content, promoting ecological governance of content generation objects towards intelligent platforms, and implementing risk sharing governance for content generation and channel promotion. Then a systematic ethical framework for AIGC is constructed. Finally, the governance destination of ChatGPT leading generative artificial intelligence is envisioned.
作者
谢梅
王世龙
Xie Mei;Wang Shilong(School of Public Administration,University of Electronic Science and Technology of China;Party School of Sichuan Provincial Subordinate Agency of CPC)
出处
《新闻界》
北大核心
2023年第8期51-60,共10页
Journalism and Mass Communication
基金
四川省社会科学重点研究基地项目“人工智能在文化遗产资源传承保护利用中的应用”(19CDCM08)。
关键词
人工智能生成内容
派生风险
治理策略
AIGC
生成式人工智能
artificial intelligence generated content
derived risks
governance strategy
AIGC
generative artificial intelligence
