We may find that letting content creators choose to have their work included, or not, in AI training sets would be helpful? Or, included in training sets for a fee, or for some sort of attribution, or ?...
The apparent current status quo of "anything on the public web is fair game for an AI training set" might not be a good permanent solution?
on edit: I just realized that of course British copyright law is also very similar to the American model.
Would you not say that using an AI to do essentially the same thing, just with the benefit that you don’t really need to pay anyone for creating the art meaning you can target more niche preferences is also “cashing in”?
Generative AI is going to allow an amazing diversification and explosion of art and music, which is going to create the next big AI application once current systems for distributing content are strained to overloading - interactive recommenders. Imagine asking for a piece of content, getting recommendations, evaluating a short clip, giving feedback to the recommender and getting better recommendations as a result in a cycle until you get exactly what you want.
At least part of the reason I like technical music is because it is challenging to play for the musician. If an AI 'plays' 'technical' music, it loses its appeal for me.
Similarly, a lot of the reason I like metal music is because of the emotion and energy poured into it by the human creator.
Being able to ask a computer make noises which sounds like some musicians you like is not, in my opinion, the same as creating art.
Humans have already been able to do this of course, but the difference is scale and automation. With this software one could, under current law, setup a 100% legal 'shadow library' that effectively infringes on every single book published. Even a leaked copy of a book could be released and shared (in its legally infringing format) before the "real" copy hit the market. And again, all completely legally. The impacts of copyright infringement are regularly grossly distorted, but I think this is the sort of technology that could genuinely damage artists and creators across many endeavors.
The exact same thing will be coming to software soon enough. 'Take this assembly/IL code, and create a functionally identical but superficially rebranded program while working to ensure you sidestep all relevant patents.' 'Sure, here you go.' The issue of this being done [relatively] instantly is going to really impact things in ways I think many are not considering. Never in a million years thought I'd see myself on the side of the copyright cartel, but this is one of the extremely rare times they're right.
The barrier to shadow libraries is marketing, which is the only thing that really separates huge blockbuster music/art/books from stuff that makes literally zero money. Quality and ideas haven't been the gate for a long time.
If not a single new piece of art were created from this point on, I'd still die with a massive backlog of material to read, see, and listen to.
At this point, I'm infinitely more concerned about the wellbeing of artists than about whether or not new art is going to be created.
> is super encouraging to the artist.
This is the "nobody goes there anymore. It's too crowded!" version of making art.
Creating more, or "limitless" art is indeed the point.
But neither of those is really what's written down, with the (U.S. law) point of copyright being "to promote the Progress of Science and useful Arts".
We might could interpret this as, training AI is progress of science, and thus the point of copyright includes training AI. Although I'm not sure that we really need copyright to do that, and accordingly, I doubt that is the intent of the law.
Really, the issue here is may be that copyright law was not formed with AI systems in mind at all, neither as creators nor as consumers, and trying to apply it to AI systems, or to reason about copyright law as it pertains to what AI systems do, doesn't necessarily work very well.
Maybe we need to go back and amend all of our written laws with a phrase like "for humans", just like so many science headlines need to include the phrase "in mice"!
if only Microsoft (OpenAI) are able to exploit works for material gain at the cost of the literally 100% of the rest of society: why should society allow Microsoft to do this?
(or even to allow Microsoft to exist at all?)