数据圈地运动背景下人工智能训练他人作品的合法性分析——以技术过程、创新激励和数据公平为视角
Legality Analysis of Artificial Intelligence Training Other People's Works under the Background of Data Enclosure Movement——From the Perspective of Technological Process,Innovation Incentive and Data Equity
陈栋1
作者信息
摘要
人工智能和互联网时代,机器训练日益成为知识传播和交流的重要方式.语料是影响人工智能能力和水平的关键因素,作品是不可替代、高质量训练语料,但面临大量侵权指控和纠纷,缺乏针对性和明确的处理规则.机器训练过程中,复制行为是必要行为和争议焦点问题.当前TDM例外规则、合理使用规则、法定许可规则等难以妥善应对训练行为、对象和目的及后续行为的复杂性.本文源于训练成本高、纠纷多的现实问题,尝试以技术过程、创新激励和数据公平为视角,在比较相关规则适用性基础上,聚焦训练中的复制行为,主张按照被诉行为过程、作品类型、训练方式并结合后续行为进行分步骤、综合性评价.
Abstract
In the era of artificial intelligence and the Internet,machine training has increasingly become an important way of knowledge dissemination and exchange.Corpus is a key factor affecting the ability and level of artificial intelligence.Works are irreplaceable and high-quality training corpus,but they are faced with a large number of infringement allegations and disputes,and lack of targeted and clear handling rules.In the process of machine training,copying behavior is a necessary behavior and the focus of controversy.The current TDM exception rules,fair use rules and statutory licensing rules are difficult to properly cope with the complexity of training behavior,object and purpose and subsequent behavior.Due to the practical problems of high training cost and many disputes,this paper tries to focus on the replication behavior in training from the perspective of technical process,innovation incentive and data fairness,and on the basis of comparing the applicability of relevant rules,advocating a step-by-step and comprehensive evaluation according to the process of the accused behavior,the type of work and the training method combined with subsequent behaviors.
关键词
人工智能/机器训练/著作权/合理使用/作品性使用/转换性使用Key words
Artificial Intelligence/Machine Training/Copyright/Fair Use/Creative Use/Transitional Use引用本文复制引用
出版年
2024
NETL
NSTL
万方数据