There's a huge difference between these AI programs and BiaB, sample libraries, and MIDI tracks: the AI companies don't have the rights to use the voices and instruments that are in their songs.
First, you know I respect you BIG TIME, I hope. I think you are making fantastic points all the way around. I may not agree with some of them now, but that doesn't mean I disagree either. I more of the mind of "I don't know" enough about some of it to really say with confidence on some of it.
As an example, and there are many I believe, BIAB has a "style" for Dire Straights that sounds pretty spot on to me. I think if you generated it, most people who know the band would identify it as being that. I'm fairly certain PG Music didn't pay Dire Straights for that signature sound. I don't know though. IF they didn't, I'm still ok with it. It's a sound. A signature sound though. I see this as being a similar situation. Just my opinion my friend. Still see your points.
They make of point of not telling where they get their materials from, but here's a project in 2020 that described how it gathered its source material (emphasis added):
Quote
To train this model, we crawled the web to curate a new dataset of 1.2 million songs (600,000 of which are in English), paired with the corresponding lyrics and metadata from LyricWiki.
See:
https://openai.com/research/jukeboxThat is, these programs trawl through millions of copyrighted songs to get their training material. No artist has granted them rights to use their materials, and no artist is compensated.
The companies creating the AI programs do a number of things to make it difficult to determine where the source materials come from.
I agree, and I know this point has been made. To me, it's similar to what what we do as humans in many ways. We take reference materials to learn what we need to do then spend the rest of our lives spitting our those things in new combinations. I think the sticking point many times I've seen from people is when WE do it, it's creativity. When a program does it, it's WRONG.
Pablo Picasso is often credited assaying “good artists borrow, great artists steal.” Some people say he wasn't the first to say it. To me, this would be ironic if he took that from someone else.
One of the most obvious is omitting artist names in tags. So although the AI is capable of rendering a song with the voice/instrument/arrangement of a particular artist, there's no way to request the AI do so.
And since a specific voice can't be requested, you'll end up getting a voice that's a combines the attributes of similar voices - enough so that the original singer can't be identified.
Isn't that kind of saying that it's a new voice then. If the voice sounds similar but isn't, it simply isn't that voice, in my opinion.
I am SO curious how they did this. Specifically what prompts they used. I find it EXTEREMLY difficult to get it to sound similar to some of the voices that I was hoping I could get it to sound like for my own purposes of fun. I've got a bit of a voice crush on Halsey and have not come even REMOTELY close to that sound. Like, at all. I must be missing something.
Udio didn't pay the rights holders of The Beatles songs to use their songs, and Paul McCartney didn't authorize Udio to use his voice.
In my mind, this is theft (well, technically massive copyright infringement), disguised the same way that money laundering hides the source of illegal profits.
I think you said this in a really effective way. You always have such good points. I sort of feel like they shouldn't have to pay them, in my opinion because we aren't generating any song that sounds exactly like any Beatles song that exists. Yes, it sure sounds like them...but that song doesn't exist anywhere else. I feel the same could be said for the voice issue.
Your points are truly fantastic and I can't tell you how much I appreciate them because they make me think, and re-evaluate my thoughts on the subject. Thanks for that! We may not agree but I'm not here to argue. I'm HearToLearn.