News for hackers. Stuff that matters.
Re:This Artwork Does Not Exist by jonplackett on Wednesday March 04, 2020 @06:54PM (#22489490)
Who owns the copyright to something created like this?
Re:This Artwork Does Not Exist by gwern on Wednesday March 04, 2020 @08:57PM (#22490229)
No one. All the 'X Does Not Exist' websites use random generation, without any human curation; the US Copyright Office is clear, non-humans (and machines in particular) cannot create a copyright, and since there is no human author expressing any creativity in random generation via NNs, there is no copyright, and they are automatically public domain. I got this question a lot on my anime faces as well and provide a few references on why they're public domain: https://www.gwern.net/Faces#faq
Re:This Artwork Does Not Exist by gridlockd on Thursday March 05, 2020 @01:27AM (#22491403)
What is that interpretation based on? I'm not aware of any case law on this.

The input of these networks are all copyrighted. The closest analogy in terms of copyright is probably "collage".

In any case, the idea that you can just take some artist's life work, train a CNN on it, and then do what you please with the results doesn't seem "right".

Re:This Artwork Does Not Exist by gwern on Thursday March 05, 2020 @10:53AM (#22494589)
> What is that interpretation based on? I'm not aware of any case law on this.

There's plenty of case law that non-humans are not creators and cannot get a copyright. How could you have not heard of the Naruto case, for example? It was famous! (And funny.)

Re:This Artwork Does Not Exist by gridlockd on Monday March 09, 2020 @05:51AM (#22524168)
Late reply, but here we go:

You're ignoring the fact that the inputs are copyrighted and that therefore the "non-human" is not the (sole) creator and the "work" is derived. The CNN isn't creating anything new except for random numbers.

Taken to the extreme, your position would be that any transformation performed by a computer program and involving random numbers would effectively remove copyright. That can't possibly be true, so there would have to be some "threshold of originality". Again, no case law on this, as far as I'm aware.

In effect, I believe you are spreading legal misinformation based on a flawed understanding of copyright law, which may put others in harms way.

> How could you have not heard of the Naruto case, for example?

Not really a good comparison. The "inputs" here are the monkey and the jungle, neither of which are themselves protected by copyright.

A better comparison would be a monkey creating a "director's cut" of Disney's Jungle Book, interspersed with footage from National Geographic, both of which are copyrighted. I doubt the "a monkey did it" kind of defense would stand here.

Re:This Artwork Does Not Exist by gwern on Monday March 09, 2020 @11:37AM (#22526243)
The 'inputs' are the random numbers. Not the training data. They don't produce a copy of an existing work, any more than a human trained on looking at faces is producing a copy and 'stealing other people's lifework' (without even paying them a cent! how dare they), unless it's so similar to serve as a substitute (among other considerations).

> A better comparison would be a monkey creating a "director's cut" of Disney's Jungle Book, interspersed with footage from National Geographic, both of which are copyrighted.

That wouldn't be a good comparison at all, because a human could very easily identify exactly which frames are literally copied from which copyrighted work. That's not 'transformative' at all. In contrast, if you look at the faces, most of them don't have even a remotely similar exemplar in the dataset (you can do nearest-neighbor lookups using features or pixels, and GAN papers often do to demonstrate that they are not just 'memorizing' - I discuss this in my FAQ section).

Re:This Artwork Does Not Exist by gridlockd on Tuesday March 10, 2020 @04:00AM (#22533582)
> The 'inputs' are the random numbers. Not the training data.

You're dodging. The inputs to the neural network are clearly the training images, they don't just disappear. The outputs are derived works, at least.

> They don't produce a copy of an existing work, any more than a human trained on looking at faces is producing a copy and 'stealing other people's lifework' (without even paying them a cent! how dare they), unless it's so similar to serve as a substitute (among other considerations).

I'm not talking about the faces, I'm talking about CNNs in general. Surely these faces all look like generic anime faces, because that's what you trained on. On the other hand, if you trained on the works of Van Gogh with his very distinctive style, that would easily be discernible.

Also, the very fact that it is not a human doing this task is very relevant for copyright, as you point out yourself.

Note that I'm not saying this is illegal or that this violates copyright, I'm saying that there is no case law on this, so you can't just claim that there is no issue here. It's undecided. It comes down to the intention behind copyright. Furthermore, even if you can dodge copyright law, you may not be able to dodge trademark law.

> In contrast, if you look at the faces, most of them don't have even a remotely similar exemplar in the dataset (you can do nearest-neighbor lookups using features or pixels, and GAN papers often do to demonstrate that they are not just 'memorizing' - I discuss this in my FAQ section).

Again, your argument fails the general case. Overfitting and memorizing is a problem when working with CNNs. Some CNNs may just literally copy - and visibly so. At which point are you not just literally copying? What's the "legal threshold" here? It would have to be decided on a case-by-case basis, just like a "fair use" defense.

Re:This Artwork Does Not Exist by amdavidson on Thursday March 05, 2020 @09:58AM (#22493998)
Is your example not transformative in the most basic sense?
Re:This Artwork Does Not Exist by gridlockd on Monday March 09, 2020 @05:56AM (#22524184)
Arguably, but transformation may or may not help you in a "fair use" defense. That is decided case-by-case.
Re:This Artwork Does Not Exist by visarga on Thursday March 05, 2020 @05:16AM (#22492371)
You're forgetting about 'who paid for the electricity, hardware and set up the model used in generating this image?'

Couldn't this allow the person hosting the model to claim copyright?

I am also thinking about hyper-parameter configuration and selection of training data as creative human input in this process.

Re:This Artwork Does Not Exist by gwern on Thursday March 05, 2020 @10:52AM (#22494570)
No, it doesn't. Again, the law is very clear about this. The US Copyright Office guidelines are extremely blunt: machines are not authors. Period. It doesn't matter if you twiddled with the hyperparameters or how much you paid for the electricity. (Similarly, you can spend billions of dollars measuring a physical constant, but as a fact, it is not copyrightable.)
Re:This Artwork Does Not Exist by valentinvieriu on Thursday March 05, 2020 @01:35AM (#22491439)
I was wondering this too. Thank you for sharing this. I've had a similar idea with https://art42.net and wondering what type of copyright to add. Is this Creative Commons?
Re:This Artwork Does Not Exist by Cpoll on Thursday March 05, 2020 @11:42AM (#22495178)
What happens if you sign it with your name, or modify a single pixel manually?
Re:This Artwork Does Not Exist by buboard on Thursday March 05, 2020 @03:42AM (#22491957)
has anyone ever sued for using their dataset? Most publishers will not care enough, but museums are very protective of their art, and i can see them suing if someone uses a large collection.
Re:This Artwork Does Not Exist by 3131s on Thursday March 05, 2020 @02:30AM (#22491640)
What about non-NN generative art?
Re:This Artwork Does Not Exist by J5892 on Wednesday March 04, 2020 @06:56PM (#22489508)