Creativity, Computers and Copyright: Promoting the Progress of Creation in the Age of Generative Artificial Intelligence

Abstract

Generative Artificial Intelligence (AI) models — such as ChatGPT, MidJourney, Stable Diffusion, Jukedeck, and Mubert — have proven remarkable in their capacity to generate new content across mediums such as text, images, and music. Although these models seem to conjure content out of the ether, they are able to create because they are trained on massive amounts of copyrighted works. This thesis addresses two principal questions at the intersection of Generative AI and copyright law: 1) Does AI training infringe on copyright protections? and 2) Do AI outputs infringe upon copyright protections?

Many scholars argue that AI’s training on copyrighted works constitutes infringement, however, I assert that it ought to be considered fair use and that we ought to instead recenter our focus toward regulating outputs. In examining these outputs, I take an approach that expands upon the current literature’s textual analysis and supports it with an analysis of copyright law’s purpose of progressing creation. Through this analysis, we find that 1) outputs that are direct copies, 2) outputs that replicate the style of an artist and 3) outputs that reflect characters pose substantial harm to the progress of creation.

We conclude the paper with policy recommendations that address how the law, economics, social norms, and technology can mitigate the generation of outputs that hinder the progress of creation. As we enter this new age of AI, it is essential that we keep progress at the forefront of our minds.