Absolutely not true. Just not how it works for most things most times with any material that is protected legally. There are small separate kinds of agreements that are made sometimes like licensing agreements and whatever (this is how a lot of sampling in hip hop is handled), niches where exceptions are made for certain uses of certain things such as limited publication of a privately owned work of art by a museum.
I’m definitely not an expert but if you don’t believe me then please test this with Nintendo or Disney IP and report back when you get successful enough for them to notice you.
You can print out something and use it privately somewhat, like you could glue a picture of Mario on a hand made card you give to a friend… but if you try to launch a line of Mario cards that’s going to be a problem for you. I think some small time artists in fandoms take this as far as they can, law be damned, but the law wouldn’t be on their side if it was tested.
Right now AI seems to be marketed as kind of a SAAS solution to all things in all places, which is definitely not private or non-commercial use.
Now, if I do NOT sample a piece of a song, but I listened to the entire copyrighted work of an artist and I write my own music in their style, am infringing copyright?
I’m not really doing something original, for sure.
Next step in that analogy: I do a statistical analysis of the œvre, and derive some model out of it which I can run. The model result is some music very much, but not entirely unlike, the artists music.
Another step in the analogy: I analyse everything on air on all radio stations for the last decade or so and come up with a large music model. Of course the material is copyrighted, but it was also publicly available. My statistic just analysed the shit out of said publicly available data without the creators consent, but the artist even sent their music to the radio stations themselves to be played. Now, my model build on their aired music produces something very much, but not entirely unlike, the a chosen artists’ music. Or mix a genre, but as it a particular musician wrote it.
Just imagine a Nick Cave bluegrass album, or a Beyoncé death metal single.
Quite likely, that would sound terrible. But it would be something people would enjoy the novelty of.
if you do it by hand, i’m sure you’re fine. if you do it using a computer as a one time art experiment, you also might be fine. when your business model is dependent on extracting the works of others and running them through your mill ( babbage, mechanical turk, or otherwise ) - it has become something entirely different
and note, certain albums, like early beastie boys stuff ( and i’m sure many others ) can no longer be made. since that time the laws on sampling have become much more fraught - and it’d be exorbitant to obtain all the rights
if the grey album hadn’t been a limited release, things might have gone differently for danger mouse
I am not using the original data. I am abstracting it. I use models of it. I could, e.g., analyse the bestselling music tracks of all time, and try to come up with a mathematical description. Then, I could do something with that model.
The same can be done for written words.
Maybe for every piece of art.
Outcomes may be soulless, the direct process is not comparable to human creativity (however, conceiving and writing the concepts and codes are a creative process), and it isn’t “intelligent”; it is a model. But it is not a copy of an artwork. It is a model, a likeness of the (unknown) rules governing a body of creative works. It is algorithms.
I am inclined to believe that this does not violate the rights of the artists who’s (published) works were used to derive the rules, and that is not become ‘something else entirely’. It’s a bjt like a summary. Like a lexicon.
For the record, I’m still not convinced of my own argument, but everything brought forth so far is not convincing me of the other arguments as well.
Making money out of this kind of encyclopedic summary is where the unease lies, I think. But the “business model” or “commodification” discussion should, in my view, not be lumped with the question if this form of model endangeres creativity itself.
as far as law is concerned, i believe it’s gong to be easier to define what is permissible input than what is permissible output.
well, see the atwood quote in the op
Judging from the results I’ve seen so far, AI can produce “art” of a kind. It sort of looks like art; it sort of sounds like art. But it’s made by a Stepford Author. And it’s dead.
so people recognize both issues. my belief is that protecting against one may help protect against the other, because flooding the world with dead art would be harder if it cost money to do so.
that said, our best defense might be reality itself. like rob says:
but it would be nice to have some back-up plans in case reality fails us.
