A philosophical question: Assume these companies didn't pirate the files to train their models. What's the ontological difference between this and a musician buying song on iTunes and then recording a near identical cover? Other than one's a person and one's a computer. https://www.404media.co/listen-to-the-ai-generated-ripoff-songs-that-got-udio-and-suno-sued/
@arossp The big difference is that with AI, the creative impetus is split. Yes, the AI is presumably trained on real and popular music, but it seemingly won't ever fully replicate that music, and to get it to even come close, you have to write very specific prompts and fiddle with it... So, when AI companies have procedures in place to prevent copyright infringement, but users relentlessly prod and push for ways that can fail, don't the users take on a degree of responsibility in the resulting infringement?
If I google "paradise city covers" and listen to an infringing cover made by a human, I obviously don't have any responsibility for the creation of that infringement. But if I twist an AI to do so, does that liability change?
