类型化视角下机器学习的合理使用制度研究
Fair Use in Machine Learning:A Typological Analysis
江璐迪1
作者信息
摘要
机器学习的版权合法性是人工智能时代亟待解决的重要议题.在规范价值层面,机器学习对社会福利的增进效益是多维度的,且具有市场失灵情形下免于承担著作权侵权责任的经济理性.在事实认定层面,技术类合理使用案件为机器学习的合理使用定性提供了重要启示:应关注不同类型的机器学习在输出结果上的差异,从系列结果的整体效果认定合理使用.在对机器学习进行分类讨论时,应采用"现实可能性"标准和"理性可感知"标准,为技术创新保留"可喘息的空间".具体来说,一般推定"非表达型"和"通用表达型"机器学习构成合理使用,但在后者的情形下应为著作权人设定"退出-选择"机制和利益分享机制;特定作者表达型机器学习所输出的结果使用了特定作者的个性化表达,不构成合理使用;非营利性机器学习因内在的公益价值构成合理使用.
Abstract
Addressing the copyright legitimacy of machine learning is crucial in the era of artificial intelligence.This paper explores the multi-dimensional benefits of machine learning to social welfare,suggesting that exempting it from copyright infringement liability in cases of market failure is economically rational.Examining fair use cases in technical contexts reveals insights into characterizing fair use in machine learning,emphasizing the importance of considering differences in output results and the cumulative effect of these results.In discussing the categorization of machine learning,we propose adopting criteria such as"realistic possibility"and"rationally perceivable"to allow for innovation.We argue that"non-expressive"and"generally expressive"machine learning may constitute fair use,with mechanisms for copyright holders to opt out or benefit from such use.However,author-specific expressive machine learning,which relies on personalized expression,should not qualify as fair use.Conversely,non-profit machine learning,with its inherent public welfare value,should be considered fair use.
关键词
人工智能/机器学习/合理使用/技术创新/分类讨论Key words
Artificial Intelligence/Machine Learning/Fair Use/Technological Innovation/Classification Discussion引用本文复制引用
基金项目
2023年度国家社科基金重大项目(23&ZD161)
出版年
2024
NETL
NSTL
万方数据