Yangzi Li (The Chinese University of Hong Kong (CUHK) – Faculty of Law) has posted Does black-box machine learning shift the US fair use doctrine? (Journal of Intellectual Property Law & Practice, Volume 16, Issue 11, Oxford University Press, 2021) on SSRN. Here is the abstract:
Machine learning (ML) has become one of the most eye-catching AI technologies in generating creative output. However, it is unable to know why and how the machines make such creative decisions. In other words, there is a black box inside the ML algorithms. Even though the mysterious black box problem engenders opacity in the algorithms, such algorithms have remarkable performance in creative industries because of robust training data sets. In the context of US copyright law, these data sets, which assembling the existing copyrighted works, would implicate copyright infringement without licenses or permission under a general limitation, namely, fair use. Will the creative ML with black box problem shift the current fair use doctrine? This article argues that the black box problem shifts the conventional fair use doctrine by breaking the balance between the rights of copyright holders and public interest and augmenting the uncertainty of fair use determination.To make the point, it first focuses on exploring and discussing three creative uses of black-box ML in creative industries and then discerning their nexus with the profound and well-established fair use doctrine. Overall, this article aims to fill the gap in the current literature on the relationship between fair use doctrine and black-box ML and make a contribution by providing thoughts for further copyright reform in the digital age.