He said that AI can dilute the market, “making original creations harder to find and violating artists’ legal rights to compensation from their work”.
If someone’s listening to an AI thingie then they’re not listening to your work and you deserve no compensation. See how this works? Just the same as if I listen to Chris Rea (say) then Beyonce doesn’t get a cut.
(Actually, there could be some fun with this sort of thing. Train the AI on Beggar’s Banquet, say, then get it to play Sgt Pepper’s in that style. It’d rack up the views as well)
Actually, there could be some fun with this sort of thing. Train the AI on Beggar’s Banquet, say, then get it to play Sgt Pepper’s in that style. It’d rack up the views as well
Until the Beatles’ feral attack lawyers get involved, anyway.
There is still a “creator” at the source of this work. Someone has programmed this and created a new object.
As long as it isn’t really passing itself off as someone else or using their original material, then there is no problem.
Making “singing a song in the manner of someone else” illegal will destry the whole tribute band industry. This sort of thing started 30 odd years ago with hip hop and techno, where kids could program a synth ( and later a pc ) to produce “music”, it is a logical next step.
Otto, no-one programmed this to create something new. Someone programmed some rules (“some” doing heavy lifting there) to determine the most likely follow up from content from input.
“As long as it isn’t really passing itself off as someone else or using their original material, then there is no problem.”
That’s exactly the problem here… The program ( I refuse to call it “AI”) uses voice-snippets of the real artists, and probably other elements, to pass a song as “real”.
Same region of problems as Deepfakes. Just as illegal. Just as hard to nail down/prosecute if the creator is careful enough with the aliases.
I have but a tiny violin to play for the Big Music Industry, but this stuff is rapidly maturing, and some of the less savoury elements on TikTok are already using it in clips to make people they don’t like say stuff they never did. And do it so well it’s hard to spot for the average punter..
It does need to be dealt with, but not because it “threatens music”. It needs to be tackled because it gives people an automated way to rather easily commit identity fraud.
Which is a tad more serious than faking a pop song by overhyped “artists”.
At the moment “AI” is hoovering the internet for content to process (lots of “Getty Images” logos sneaking through, etc) but what happens if this business model disincentivises original content creation so much that the pool of new dries up? AI feeding on its own output could be quite amusing for a while I suppose, but I’m not sure cultural enrichment will improve for us humans long term.
We’re going to find out, and soon. The cliché is out of the metaphor and there’s no putting it back.
@PJF
If everything is derivative then one of two things happens:
(1) Everyone is happy with derivative. Derivative becomes much cheaper. Everyone is happier.
(2) Some people are not happy with derivative and will pay for “human-created” things.
Of course, there’s (3) grifters will get money from taxpayers to produce “art” that nobody wants, but that happens already.
“The program ( I refuse to call it “AI”) uses voice-snippets”
“hoovering the internet for content to process”
It’s important to remember that the models are orders of magnitude smaller than the training data. The training data *informs* the model, but the model does not include the training data. The models absolutely do produce original work.
“what happens if this business model disincentivises original content creation so much that the pool of new dries up?”
I think it’s unlikely. I think people will enjoy art created by other people for purely emotional reasons; including that they like to have something in common with the artist.
As for training data, sure the computer models can generate their own, or look into the real world with cameras and microphones, or humans can be employed to provide training data.
There’s an idea: develop a new art style, then train a computer model to make new pictures in that style. This makes computer models just a more efficient way for humans to make art.
Another reason it’s unlikely there will be no need for old fashioned humans in the face of all this comes from John C Wright: “It is true that the Sophotechs can perform any of these operations more swiftly and more efficiently than can we. But it is also true that they cannot do everything at once, at every place at once, as cheaply as everyone wishes. There is always someone somewhere who wants some further things done, some further work accomplished. There is always someone willing to pay much less for work moderately less well done.”