Fair Use of Generated Artificial Intelligence in Machine Learning
Compared to previous analytical artificial intelligence(AI),generative AI technology possesses a more potent capability to imitate and generate works.It can learn from existing works while producing similar ones.Consequently,certain AI-generated works may present challenges to the established copyright market for human authors.Given that generative AI necessitates machine learning before generating works,it inevitably relies on existing copyrighted materials.However,whether such utilization can be considered copyright fair use requires further research.This article studies whether the machine learning process of generative AI can be categorized as fair use.In contrast to existing literature,this article examines the fair use issue through the lens of the varied purposes of generative AI.For generative AI lacking a specific intent to imitate,owing to the substantial transaction costs associated with obtaining ex-ante permission and the transformative nature of its purpose of use,it can generally qualify as fair use.However,generative AI designed to target specific objects of imitation should not be considered fair use.In analyzing generative AI without explicit objects of imitation,this article primarily considers the transaction costs required in obtaining permission and the transformative nature of its purpose of use.In most cases,AI encounters considerable transaction costs when utilizing works for machine learning purposes.Therefore,it becomes necessary to consider exempting AI developers from copyright infringement under the fair use doctrine.From a transformative perspective,utilizing others'works for machine learning purposes is inherently transformative and should thus be deemed fair use.However,when generative AI is specifically intended to imitate the works of a particular author,applying the fair use exception is inappropriate,for such imitation may significantly substitute the market demand for the original author's works.In such instances,developers of AI should seek permission from copyright holders.The contribution of this article's study of the fair use doctrine within AI's machine learning lies in its advocacy for a balanced regulatory approach toward AI.By emphasizing the need to strike a balance between technological advancement and copyright protection,it underscores the value of inclusive and moderate regulatory policies in fostering the innovative development of new technologies.Consequently,within the machine learning process,AI developers are generally not required to seek ex-ante permission from copyright owners for all training data.Moderate regulatory policies serve to prevent overly stringent copyright enforcement from impeding the advancement of new technologies.By adopting such policies,regulators can encourage the responsible use of AI while safeguarding the interests of copyright holders.
generative AIfair use of copyrightsmachine learningtransformative use
万方数据
维普
