You must not have any artists in your life, a lot of us are hollering that all of these things are massive copyright infringement machines that are trained on our hard work without any permission, never mind giving us a single penny of the money the companies building these things are making off of them.
Would your opinion change if these models were trained exclusively on works in the public domain? Do you consider it to be infringement if new and novel art is created simply in the style of other artists? If so, how should we treat art students when they try to recreate a previous work?
It's also worth noting that producing new art is only a subset of possible use cases. So how should we treat new and novel photo-realistic images? Is that infringement too?
For a model trained exclusively on the public domain, I would still be unhappy about the prospect of a world where a lot of low-level gigs that keep artists alive have been replaced by spending a half hour fucking with prompts. It would however be free of the "this corporation abused fair use to create a training model which it is selling to people" problem.
And no, it's not infringement if you work in someone's style - but it's also not professional to do this for too long, especially with a living artist. One piece in someone else's style is a homage, several years of pieces in their style starts to get you labeled a "clone" or a "style thief". Unless you are their assistant, in which case "working in their style" is exactly what you are being paid to do.
If you would like an example of this happening, read the 'controversy' section of Kieth Giffen's Wikipedia page, which sketches out the way fans and fellow artists took Kieth Giffen to task for doing a complete style swipe from José Muñoz' shadow-drenched crime comics for several years: https://en.wikipedia.org/wiki/Keith_Giffen#Controversy
Silence and ableism are what we get from artists. We're still in the kneejerk reaction phase. "Having to do it yourself" is an awful precondition for art and expression.
The concept of AI art is fine. The implementation by scraping copyrighted work and claiming that their magic algorithm makes they immune to the consequences is where the complaints are.
With the current length of copyright, there is no possible way to build it otherwise. Saying that AI needs to only train on public-domain works, is equivalent to saying that it shouldn't exist.
Of course, if it didn't exist, I wouldn't be buying art from artists instead. I'd simply go back to my stories not being illustrated at all. So from my perspective, this story is all about artists wanting to tear the ability out of my hands.
One could procure the necessary rights from the copyright owners / artists for all the training data. Sure, it would be quite costly and take a long time - but it is possible.
No, it isn't possible. Tracing copyrights for many things is literally impossible. Might be owned by defunct company X that died out several decades ago and had records destroyed in a fire.
One does not need every single copyrighted object in the world, just a large, diverse and representative set. One would have to skip the cases where rights cannot be secured. This is possible, see for example Spotify.
Nonsense, the claim made by those that say the end of knowledge work is nigh is that an AGI is coming that is superior to human intelligence. Humans do not need to train and read / consume gigabytes and terabytes of past solutions to produce new work and they do it at a trivial fraction of the amount of watts.
From my perspective, our magic algorithm does indeed make us immune to the “consequences” of scraping copyrighted work. Legally, diffusion models learn from work rather than copying any specific piece, and copyright does not protect a style or give artists the right to retroactively restrict learning from open content. Ethically, our society accepts legally automating people out of jobs, and there’s no good reason that art is different - simply social prestige and classism. Current models, even fine-tuned ones on specific artists, are not wrong, however sad or harmful for the artist they may be. There are winners and losers, that’s capitalism, for better or worse.
I am sorry for artists and think we should probably have a subsidy for human creativity. But for me, Twitter’s bad-faith arguments about “stealing” or “copyright violation”, and maximalist demands for bans on AI art, make compromise undesirable and probably impossible.
> Legally, diffusion models learn from work rather than copying any specific piece, and copyright does not protect a style or give artists the right to retroactively restrict learning from open content.
Unless you are a lawyer you are just pulling reasoning out of thin air. All aspects of Copyright with human's have been decided by decades of court cases. Assuming you understand the underlying theory enough to guess how a court case would go is presumptuous.
> Ethically, our society accepts legally automating people out of jobs
And society accepts paying people for their work.
> There are winners and losers, that’s capitalism, for better or worse.
Capitalism is all about avoiding paying for what you did in order to artificially increase your profits, I agree there.
> I am sorry for artists and think we should probably have a subsidy for human creativity
I mean if the AI models just... Trained on purchased or free materials there wouldn't be a problem.
Other media figured out how to handle when a "purchase" isn't a purchase in the traditional fashion: see streaming services.
AI art is just the latest version of piracy. "It is just bytes" yet again.
You keep thinking ML models are copying your work but I think this comes from ignorance about how they work.
They're not copying, they're doing the equivalent of looking at thousands of paintings and from that figuring out what paintings are supposed to look like. If you were asked to draw something in the style of Picasso and did so, would your new art work be a copyright violation?
Is every painting you've ever seen public domain? Have you ever got an idea or influence from a non public domain artwork?
Is lossly encoding generative or copyright violation? Let’s say you encode a bunch of copyright images as lossy jpegs, specifically? Most people would say “copyright infringement“.
Now, if I have an encrypted zip file, random binary blob… but with a pass phrase, it decodes to a bunch of copyright images. Still copyright infringement right?
Now, if I have an AI compression algorithm where the model that rebuilds the content of the archive is a local binary blob, and all you transmit is the pass phrase, it’s still copyright infringement right?
Even though, you’re actually not sharing the content of the archive, you’re just sharing the pass phrase and the AI is rebuilding it from that pass phrase.
So, 1) when you rebuild a binary distinct output that sufficiently closely resembles the input, 2) regardless of the manner in which you are doing it, technically, it’s copyright infringement.
Now, this is where your argument fails.
…because your argument is that the technical means by which the content is “generated” inherently makes the content distinct and not copyright infringement.
However, “generation” and “decompression” are the same thing; the technical means by which the output is generated is irrelevant.
The difference is that in this case, the output is sufficiently distinct that you could reasonably argue that it is different from the training data. That’s a fair argument!
…buuuut, here’s the thing: if your model can generate outputs that are copyright infringement because they are similar but not binary identical to existing specific copyrighted works… then what you have “effectively” is a giant compressed archive that has both copyright and non-copyright work in it.
That’s problematic.
I am not misunderstanding how the models work; I’ve built models like this, and yes, there’s no “copy of the image” in the checkpoint file… just like an encrypted zip file doesn’t have a “copy of the image” in it until you apply the correct decompression and decryption algorithm, it’s just a blob with high entropy.
> Have you ever got an idea or influence from a non public domain artwork?
That is fundamentally not the issue here.
The issue is that the model checkpoint file contains a combination of data and algorithm that can rebuild both novel and copyrighted work.
If a zip file with both copyright and non copyright work in it is infringement, so is this.
Ie. tldr: yes, it’s complicated, but this “this comes from ignorance about how they work” BS that is floating around the SD community is both dishonest, and lacks itself an understanding of how the models work. It’s a pretty ironic thing to say really. The right prompt can generate copyrighted content. That how the models were trained. Not binary identical, but as we’ve already established, that’s not necessary condition for copyright infringement, or like every lossy image format would be free game.
…and yes, I get it, if someone invented a purely algorithmic decompression algorithm that could take “source code for MS Word” and generate exactly that as the output, then I’m arguing it would be infringing. Yup. …because the only way you could create that algorithm would be to encode the original source code into it.
But you can't rebuild the original image unless you're so incredibly specific that those same instructions given to a random artist would make the original image. And even then...
So your argument rests on a premise that, as far as I can tell, isn't true. What prompts did you give to what ML model that gave you something copyrighted?
Because I just tried a whole bunch of different prompts to make the bloody Mona Lisa with no luck.
Try image-to-image on the source image with a low strength. Can it regenerate the image? If the answer is yes, then there is categorically some latent that maps to the output. i.e. It is technically possible to generate the image from the model.
The question is, how would you generate that latent from a prompt?
> a painting of a woman with blue eyes, by Aleksi Briclot, trending on Artstation, ghostly necromancer, red haired young woman, downward gaze, the blacksmits’ daughter, her gaze is downcast, dressed in a medieval lacy, gothic influence, screenshot from the game, from netflix's arcane, high priestess
(Yes, it is: in this case, the top hit is a match score of 0.3968).
4) Crack open stable diffusion.
Put the cfg scale up (do not allow variation) and pick a some step value that generates reasonably accurate images. Maybe like k_euler, 50 steps, cfg scale 30, 512x768.
You're now generating images that are 'nearby' in the latent to the target; now its just pissing around in the seed and with variations to narrow the gap.
> So your argument rests on a premise that, as far as I can tell, isn't true.
...but the point I'm making is that it is a) possible, and b) plausible, if you can be bothered doing a seed search. Can you be bothered? I can't be bothered.
Like... I mean, dude, the model was trained to be able to do this. That's literally what it's supposed to do. The VAE can map a latent to real existing images trivially (that's what it literally does when you use image-to-image). The latent space is a 64x64x4 vector that you're moving through with your prompt in 'latent space'.
Look, I get it, the chances are you picking the exact seed that generates this exact image are pretty slim right? 64x64x4 is a massive fucking number, and the chances of stumbling on exactly the right seed is like winning the lottery right?
...but that would only be true if you were RANDOMLY moving through the latent space, and you're not. You're specifically homing in on 'good' latent space values around real images using your prompt. That means that the chance of the latent you pick being a real picture is not 1/64x64x4... it's actually plausibly higher.
There is a non-zero chance that any generated output is actually a real image.
So, back so our original quesiton:
Is sharing a magic seed (123123123 or whatever) the same as sharing a password to an encrypted zip file?
...because that's what it comes down to.
You have:
- (algorithm) that takes (pass phrase) and generates (output).
- If you can provide (pass phrase) and generate (output) that is copyrighted content, then is (algorithm) infringing?
- The answer applies the same way to encrypted zip files and to AI models.
So... you gotta pick which way you wanna roll with it, but you can't have one or the other. You get both, with the same rules.
I'm with you. "Oh no look at that ableist piece of shit Andres Segovia with his ten fingers" give me a break. Chuck Close painted his giant portraits with what, one finger?
sorry if anyone is disabled and i hope they're able to get the support they need, but equating DIY with ableism only makes sense if your disability is "lazy and kind of a moron." and to be fair, i know plenty of people that suffer from this chronic condition, their art is pretty unnecessary, and they're often trying to ally themselves with whichever new protected class in order to garner kudos in ways other than being remarkable on their own merits.
i love this dumb premise, though. i'm going to continue thinking about it. damn, "thinking," how ableist of me... there are people in comas who can't think!
not to mention the blatant disregard for the many disabled people who rely on art as our only means to make a living. i've seen far too many singularity nutcases saying artists are regressive luddites who should be crushed under the noble wheel of progress from once side of their mouth while crying about gatekeeping and '''ableism''' from the other.
I was divided on this issue, but now, I agree. The noble wheel of progress will continue to turn. They only have the choice to get out of the way or be crushed under it
"Artists shouldn't care keep their industry" - so how do you feel about the SFWA's campaign to make Disney uphold the contracts for tie-in novels they bought along with Fox? https://www.sfwa.org/2020/11/18/disney-must-pay/
Yet another triumph for the bosses of the world; in the absence of unions, automation only profits the owners of the machine, and none of the people who are now performing much less skilled labor gets a better life.
In this case, it's a labor that's already a way to barely scrape by for a lot of its practitioners.
This is the difference. Is this what you are cheering for? Is this the world you want?
You’re relying on multiple meanings of the word “train”. Humans produce new knowledge, art, solutions, etc, without consuming the full body of human created media. It’s hardly even reminiscent of the same things. Humans produce solutions, at times, without ever being trained in a fraction of the vast body of human knowledge. This is not the same as a generative AI.
