Back in the early days of the internet (say, late 90’s), it was becoming fairly obvious that copyright law needed a serious update. Digital changed the nature of “copy”, and concepts around sampling and remix were happening in earnest. Not to mention Napster. And in fact the entire search engine industry depended on a rather broad definition of “fair use.” Google certainly became the giant that it currently is due to being able to scrape vast amounts of information without worrying about copyright issues.
Fast forward to today and we have generative AI that has used a similar approach for training the models – scrape anything and everything – under the guise of fair use. The problem is that search engines don’t necessarily create derivative work. Rather they index (and rank, which is another discussion) what they find. gAI however does create derivative work, and as the capabilities continue to mature around text generation, image generation, video generation, and music generation, it’s going to be like 1995 all over again. What about copyright and those that create the content getting scraped?
The Verge has an interesting piece discussing this problem, with the flash point being a song created by “AI Drake” which went viral and then was taken down – but not in a straightforward manner. Some have been talking about issues around deep fakes for some time, and this song crystallizes a number of the issues. The music industry (and especially musicians – you know, humans) has seen a lot of ups and downs driven by technology. Recording, MIDI, drum machines, Pro Tools, Napster, sampling – the list goes on. But this is a somewhat different kettle of fish. You are essentially sampling everything, then creating a work that is both derivative and new.
The main question is, “who gets the royalties?” Like most things in modern life, this will end up in the courts, possibly until there are no courts (or once we have AI judges and juries)