I want to train a model for a musician based on piano covers they have recorded of popular songs and therefore own, but because of copyright infringement they still pay royalties to the writers of these songs. Curious if this is legal or not – I would presume not since the recordings are not being monetized in this scenario but rather fed into a model, and the recordings are still owned 100% by the musician licensing them to me.
We need to back all of this up- What is the purpose of you training this model? Are you going to make money off of it? Because from a practical standpoint, if you are not going to make money off of it, you are completely safe to do whatever you want. This coming from a 20 year professional artist in film and tv where we deal with copyright daily when sourcing and using found assets, both at work and at home. The only way you actually get in trouble for stealing art, is when you make money off of said stolen art. For this to happen, the person you stole music from would have to find out you did in the first place, which is easy to avoid. Then if they were to sue you, they’d have to PROVE DAMAGES to take your money. They’d have to show a judge that using your model, containing their work, you have made money, and that money opportunity was lost to them. Would someone be able to do those things with how you are planning on training and using this model? I don’t know your plans, but I’m guessing absolutely not. So in short, assuming this is a hobbyist project, you’re fine.
Okay yes, technically it would very difficult to track and no one would probably even know the musician is using such a model. But as an academic, I’d prefer to do legal things only :P especially if I want to use results in a publication down the line.
And yes, I’d make money off it. would charge $x/full piano cover request.
This is a bit of gray zone then. To be honest, I’d probably just do it because every major ai company in the world is already doing way worse than what you’re doing, and making billions. The reality is their actions are creating legal precedent in your favor. Not LITERAL legal precedent, but abstract in the sense that every day they do business and DON’T get sued, we trend towards their actions being considered legal when it comes to blows later. It’s an open secret that landmark laws and cases are often dictated simply by what aligns with the interests of the current government administration, not necessarily what is most fair to everyone, or even constitutional. If the US government is doing millions/billions of dollars worth of business with ai companies, and these companies are doing ethically ‘gray’ training, it’s a pretty strong bet the judgements will be rigged in favor of Ai. I’m not telling you to betray your morals, but I’m just saying the wind is blowing that way, and it’s hard to sail against it.