Many large sophisticated machine learning models, like those employed in generative AI, are trained on immense amounts of copyrighted images or text. How is that legal? In this episode we delve into the exceptions to copyright law that enable such uses to not be seen by courts as infringement. This includes expressive vs functional uses of a copyrighted work, fair use, and the possibility of a data mining safe harbor law. We also discuss whether such interpretations are to the benefit or detriment of society as a whole.
A note: as mentioned in the episode, we are not lawyers, and this episode should not be considered legal advice. It is just a discussion of the issue based on our somewhat limited understanding of the legal arguments and expanded to consider the societal implications. Also as mentioned in the episode, we based much of our understanding on the article "Does Training AI Violate Copyright Law?" by Jenny Quang which is linked below in the show notes.
Show Notes
Follow us on X @KopecExplains.
Theme “Place on Fire” Copyright 2019 Creo, CC BY 4.0
Find out more at http://kopec.live
Read transcript
