For over 300 years, copyright law has been central to innovation and design, protecting artists, innovators, and creators. Rapid technological development over the past few decades has necessitated new interpretations and applications of the law in order to maintain this function. Although complex in application, the copyright framework has managed to integrate these new technologies. Fast-developing Artificial Intelligence (AI), however, poses a challenge to the protective dimension of copyright law.
The digitally-driven shift of copyright from protective to perplexing was the focus of discussion in Latham & Watkins Forum, “How Copyright Law Crafts the Contours of Technological Innovation and Design,” held at NYU School of Law on April 11. Samantha Fink Hendrick, a current law student with a cyberlaw specialty; Fred von Lohmann, former director of copyright at Google; and Michael Weinberg, intellectual property and general counsel at Shapeways, provided an overview of the development of copyright law and current challenges in the field.
Computers have long been used as an instrument to produce creative expression, but AI presents a key shift in that it relegates the role of the programmer in the creative process. Previously, heavy input from the programmer was required in order to produce the creative work. The machine was just a tool. However, with AI, and machine learning software specifically, programmers set parameters, but the artistic work that is produced is actually generated by the computer. The machine learns from data and makes future decisions. This dynamic raises questions as to who has rights to the work that is produced.
Additionally, in the AI sphere, the restraint of copyright law poses a challenge to the protective dimension of the law because it limits access to datasets. For example, if one needed to source 20 million photographs for their machine learning system, they would need to determine how to obtain clearance for those 20 million photographs. Courts have not yet determined whether training an AI qualifies as infringement. This obscurity may not pose a concern to powerhouse tech companies that can afford costly legal battles, but will certainly have a negative effect on start-ups and other small businesses.
Copyright restrictions on dataset access forces researchers to use biased databases in their algorithm training. This distorts the outcomes of machine learning programs because the data used will have been sourced and pre-cleared. To use an example provided by the panelists—imagine a speech-learning software that only had access to old English texts on which there were no copyright restrictions. The usefulness of this software would be severely limited; it would not adequately serve language-learning tools or educational tools for those with learning disabilities. These dynamics show how copyright law, initially a tool to promote innovation, must face its transition to a positon that could also have harmful effects on innovation.
The ambiguity of the law concerning AI training has already had negative effects in that major AI companies such as DeepMind, Facebook, Google, Apple, IBM, and Microsoft have shrouded their data sources in secrecy. This has not only restricted competition, but has also made it difficult for journalists and academics to uncover biases in algorithms that are being used in contexts with major societal impacts, such as in determining sentencing and bail for individuals in criminal proceedings. Amanda Levendowski, a clinical teaching fellow at NYU School of Law, has argued that copyright law can ameliorate this problem. If training an AI were classified as fair use, researchers would not only have access to more and less-biased data, but would also not fear disclosing their sources to academics and journalists. The core goal of innovation would also be promoted, resulting in better algorithms and outcomes.
Throughout its history, copyright law has had to continuously transform to meet technological developments. Although the current challenges are particularly complex, maintaining a dual focus of promoting innovation and transparency should guide decision makers in the right direction. AI and copyright law will be able to reclaim their stardom together.
Ana Namaki is a J.D. candidate, 2019, at NYU School of Law.