In the last few months, we went from AI music being a curiosity you needed a good Linux box to play with, to something pretty crappy but looking somewhat interesting, to... this.
@stirlock@Nawer_Rapter At worst, this is going to be an Open AI situation. An insane amount of restrictions for a while (does anybody still remember pre-Chat-GPT Open AI draconian usage policies?), and then somebody is going to release an open-source model and all the restrictions are magically going to vanish. This is a consistent pattern in AI by now.
@miki Just in case, do you know about the chronology and strategies used?
The thing is that a non-profit organization made those libraries, with dubious moral overall involved and a lot of copyright issues that, in the case of image, pass as a gray area.
In the world of music copyright is so powerful that you can't play a copyrighted song for a charity, not to talk about non-profit.
The thing that was killing music may save it from being on the shitty situation illustrators are now.
@Nawer_Rapter You may be right, the music industry seems a lot more litigious and trigger-happy than anybody else. All we really need here is a single country approving fair use for AI training though, and they have far less lobbying power outside the US.
@miki That's a funny thing cuz law doesn't work like that on libraries that can be used worldwide on music copyright issues.
This isn't the casino island dream, it's an issue that makes it impossible to put music from a country on another's country library, due to how the law on music copyright is applied globally.
That issue happened in Finland-Spain with online shopping, being Finland's classical music CDs just a bit cheaper and thus being bought by people.
@Nawer_Rapter None of this is going to matter once an open model gets out. Sure, it's going to be illegal in the US, just like Bittorrent and Piratebay and Napster and Kazaa and Gnutella and all other platforms like this, it probably won't be used commercially, at least for a while, but users are going to have a field day with it regardless, and then somebody will have to budge, just like what happened with Spotify.
@Nawer_Rapter You might be right that this won't treaten artists directly, commercial usage is a lot bigger problem than people fooling around at home. And from what we've seen with Stable Diffusion, once an open model is out, we're absolutely off to the races, no matter what the law actually says.
@miki Yeah I know. I love the fade out and everything. I mean the artifacts are still there and all. But that's a small price to pay for what I'm thinking is someone typing some lyrics and stuff into the AI and having it be generated. You have to wonder how much all this costs to do per user. I'm thinking $20 per song or so.
@zersiax Yeah, it even gets the southern accent just right, and that wasn't even in the prompt I think. There are other genres too, and if not for the crappy quality and artifacts, this would definitely be listenable.
Add comment