ABSTRACT
The wondrous and beguiling technology of generative artificial intelligence (‘generative AI’) leaves us at a loss for words as to how to understand or even describe it. But in an effort to fill the gap we rush to assign metaphors and analogies to the complex technological operations of deep learning, transformers, neural networks, and large language models. The gravamen of my thesis here is that the analysis and resolution of copyright legal issues relating to the training and operation of generative AI have become enslaved to inapt and fundamentally inaccurate metaphors and analogies about how the complex and advanced technology of visual generative AI works, and these mistaken metaphors drive the determination of manifestly important issues of copyright law. At one end of the spectrum, the United States Copyright Office has nearly completely anthropomorphized generative AI tools and declared them to be the authors and creators of works designed and created by human end-users of these tools. Although AI technology is wondrous and powerful, it most certainly is not human, it is not alive, and, at present, it does not design and create works on its own. At the other end of the copyright spectrum, artists and creators who feel threatened by the undeniable democratizing power of generative AI that enables the proliferation of artistic and literary skills to a huge number of new practitioners of the arts strive to find metaphors and analogies that appear to match the legal understanding of terms such as copying and infringement while avoiding uncomfortable analogies that would lead to findings of non-copying, non-infringement, and fair use.
This article seeks to clarify and explain what generative artificial intelligence is and how it actually works so that its role in copyright law can properly be analyzed and assessed. These inapt and incorrect metaphors and analogies of ‘creation’, ‘authorship’, ‘copying’, ‘collaging’, and ‘theft’ have complicated and at times confused the analysis of key copyright law issues regarding whether the elements of copyright infringement have been met, whether established copyright fair uses protect the alleged uses of material in the training of generative AI platforms, and whether, in a different inquiry, generative AI platforms have usurped the original and creative role of human users of these platforms as to the authorship and rightful ownership of creations made by the humans with the assistance of generative AI.
The purpose of this article is not to end the use of metaphor in the law of artificial intelligence. In fact, I will be the first to admit that it is impossible to discuss this technology of algorithms, mathematics, and data science without using metaphors to try to communicate the abstract concepts represented in the technology. My purpose is to point out and hopefully convince others to work to correct or at least clarify the most misleading metaphors associated with generative artificial intelligence that confound the important copyright law decisions of our day as to whether the training and operation of generative AI models involves copyright infringement and whether the outputs produced by humans using generative AI models are copyrightable and owned by the human who produced them using the generative AI system as a tool.
Murray, Michael D, AI Pirated my Art and Birthed Infringing Works, and Other Metaphors that Confound Copyright Law (January 24, 2025).
Leave a Reply