> The purpose of music is (usually) to touch people emotionally. If it works, it works. Doesn't matter how it was made.
The touching you emotionally part is due to the quality of the underlying creative work. I'm sure the GP's wife was touched- they put in the work to make something- but the fact is that work they did was enabled by the theft-at-scale of work others have done.
You can square this with your own ethics however you like but there's simply no getting away from the fact that all of this, the text, the music, the video, all of it only exists because of theft of creative work on an industrial scale. These models did not come from the ether- they are weighted mathematical averages based on ingesting shit tons of existing creative work, made by people, the vast majority of which was ingested against those creatives' explicit wishes.
Unfortunately most people don't give a shit where things come from as long as they get whatever they want in the end, which is why our economy is almost exclusively run by sociopaths.
"You can square this with your own ethics however you like but there's simply no getting away from the fact that all of this, the text, the music, the video, all of it only exists because of theft of creative work on an industrial scale"
I know enough of music creation to know, all music we enjoy is created by "theft". Meaning taking a riff from here, a melody from there. And tweaking it. AI just automated it. Not sure, it sucks with the whole buisness modell around it. That only some profit and not the truly creative composers. But that .. is hardly a new thing. There are many, many awesome musicians out there. Always have been. But only some become "superstars". Where a whole industry pushes them so they stay on top no matter what. That's not fair, but AI did not change this.
This argument always hinges on pretending that "humans being influenced by other humans" is the same thing as "a model ingesting millions of works without permission," and it just isn’t. Not even remotely.
Human influence is selective. It’s contextual, filtered through taste, memory, culture, and intent. A metal songwriter doesn't subconsciously absorb the entire global corpus of music; they draw from the artists who shaped them, or deliberately subvert something specific. That’s literally the creative process: choosing what to reference, twist, or subvert, reject, counter.
A model doesn't do that. It doesn't choose influences, it doesn’t have tastes, and it doesn't have intent. It just digests everything it's fed into a massive statistical set of averaged patterns that it has found, and then regurgitates them on command so as to "minimize error." Calling that the same thing as human inspiration is like saying a wood chipper is just an automated sculptor because both involve wood going in and differently shaped wood coming out.
The music industry fucks musicians raw, to be sure, but this is not guaranteeing anything for musicians in the slightest, quite the opposite: it just makes it so users of the models can also fuck musicians. How is that good at all? The exploitation of artists at scale being the status quo is not a reason to excuse even more exploitation, that's certifiably insane.
"A metal songwriter doesn't subconsciously absorb the entire global corpus of music; they draw from the artists who shaped them"
They also draw from the people around them. The music they hear when they are in public spaces. Etc.
Out of curiosity, have you ever done composing?
Anyway, one can create very shallow songs by hand. One can also give empty vague prompts.
Or one can make a complex music arrangement, manual editing of tracks, have some AI generated mixed in, very detailed prompts etc. If that ain't creative to you, that is your opinion. I think different.
If it were just an average, we'd only get gray sludge. Models learn the manifold distribution - they don't just mix existing works; they discover the hidden rules by which those works were created. This is reverse-engineering of human culture, and that's exactly why control over the latent space is so critical - otherwise, we surrender culture to an algorithm that has "learned the rules" but has no concept of meaning
The touching you emotionally part is due to the quality of the underlying creative work. I'm sure the GP's wife was touched- they put in the work to make something- but the fact is that work they did was enabled by the theft-at-scale of work others have done.
You can square this with your own ethics however you like but there's simply no getting away from the fact that all of this, the text, the music, the video, all of it only exists because of theft of creative work on an industrial scale. These models did not come from the ether- they are weighted mathematical averages based on ingesting shit tons of existing creative work, made by people, the vast majority of which was ingested against those creatives' explicit wishes.
Unfortunately most people don't give a shit where things come from as long as they get whatever they want in the end, which is why our economy is almost exclusively run by sociopaths.