🤖 ai art 🎨 debate

[Isaac]

I haven't done a poll in a while so let's see if it actually works. I'm curious to see what you pick. I know I can add more options because each choice is pretty well defined and I could have even more reasoning/scenarios in new choices, but if you don't agree with one you selected, pick the best one then comment on your own personal clarification.

[Jeff250]

The conspiracy theorist in me thinks that these types of questions are seeded by "AI" companies to exaggerate the capabilities of their machine learning models so that we think that their models are so powerful that they introduce all of these terrible ethical and legal quandaries.

[Top Gun]

Unless the artists whose material is being scraped have given permission for their art to be used in this way, then yes, absolutely. It's using the artists' own works in an attempt to make them obsolete. Completely scummy.

Also I just...fundamentally don't get the purpose. A major reason why I consume any creative work is getting to share in the creative process of actual people using their actual talents. Why would I have any interest in something an algorithm crapped out?

[Isaac]

Are you saying something that's transformative use can also be theft?

[Jeff250]

I'm not a lawyer, and this ultimately seems like a legal question. But to use an analogy from something that I'm familiar with, if you copy even a single line of GPL'd code into your project, then your project is derived from that project, and now you are obligated to GPL your project as well. How does this transfer to what LLMs are doing? I don't exactly know. Maybe all outputted code should be GPL licensed? (In reality, it would likely still be nondistributable even if you only used open source code due to code licensed under incompatible licenses being combined to generate the output.)

[Jeff250]

But really, who is going to claim copyright infringement over a generated person with three arms and missing a thumb?

[Isaac]

... if you copy even a single line of GPL'd code into your project, then your project is derived from that project, and now you are obligated to GPL your project as well.

I love this argument. I didn't even think of that. Something similar could apply to training images.

But really, who is going to claim copyright infringement over a generated person with three arms and missing a thumb?

So @Samdoesart was overwhelmed with the number of people that able to replicate this style with the the first wave of generative ai models. If you get any model you almost always get his style of art using "in the style of Samdoesart" in the prompt. There are also many models called "Samdoesart" where the entire model can replicate this.

Facebook's Meta ai has a generative art component that has a safety measure that prevents you form using Samdoesart in a prompt. It will shut down and say something to the effect of "sorry I can't do that dave". However, this still doesn't mean Samdoesart keyword doesn't exist somewhere in the Meta weights list. This also probably applies to thousands of other artists and photographers that the model is built on. Are they owed something or does that safety net that's triggered protect facebook?

[Jeff250]

Is style of art something that can be copyrighted? Is there anything stopping me from making art in Samdoesart's style (other than an obvious lack of skill and ability)?

[Vander]

[Isaac]

Is style of art something that can be copyrighted? Is there anything stopping me from making art in Samdoesart's style (other than an obvious lack of skill and ability)?

Top Gun would say theft has occurred and thus a violation of copyright.

IMG_0021.jpeg

[Krom]

Is style of art something that can be copyrighted? Is there anything stopping me from making art in Samdoesart's style (other than an obvious lack of skill and ability)?

Yeah, you could copy someone else's style if you have the skill and ability to do so, but that skill and ability takes a certain amount of your time to accomplish and that time is valuable even if the work ends up being "derivative". An AI can also copy someone's style if you train it off everything they have put out, but unlike you, the AI can do it quickly and only using a relatively cheap amount of electricity.

[Top Gun]

Are you saying something that's transformative use can also be theft?

It already is in some countries that don't recognize "fair use" as a legally-protected status. In fact, I was just reading about such a case earlier today. And I'm not saying that I agree with this prosecution, quite the opposite in fact, but it's something that commonly happens. The image and text-scraping processes that generative AI use are exponentially worse in terms of violating such copyright protections.

You've dodged the question though: again, tell me why I should have even the slightest interest in the output of a soulless algorithm that runs an end-around the basic human creative process, and indeed seems explicitly designed to replace said process?

[Isaac]

Are you saying something that's transformative use can also be theft?

It already is in some countries that don't recognize "fair use" as a legally-protected status. In fact, I was just reading about such a case earlier today. And I'm not saying that I agree with this prosecution, quite the opposite in fact, but it's something that commonly happens. The image and text-scraping processes that generative AI use are exponentially worse in terms of violating such copyright protections.

I don't think photoshopping roaches onto art would significantly change the original work enough to qualify as Transformative Use even under US law. So I don't think it's relevant to the original question of this post.
The question is if I took an artists entire gallery to train a Stable Diffusion model, would my output, which wouldn't match any photo in the artist's gallery, meet the requirements to be Transformative Use? It's possible I'm wording this clearly enough. Let me know if you need an example of what I'm saying.

You've dodged the question though: again, tell me why I should have even the slightest interest in the output of a soulless algorithm that runs an end-around the basic human creative process, and indeed seems explicitly designed to replace said process?

I didn't want to stray off topic, but I think your link on photoshopping bugs might mean we're already off topic anyway. So I'll answer it now:

Why should you be "interested"? Such a weird question to ask. I wonder if "interested" was the right word... maybe you meant "...why should I enjoy..."? You should be interested in identifying the shift that's about to occur, because it's going to be everywhere at the commercial entertainment level. I'm not talking about artists on DeviantArt losing rank, but hundreds of thousands of outsourced animators who were paid to draw in-between frames now have competing ai companies trying to replace them by developing workflows that only need the two keyframes and the ai solves the inbetween frames.

Another area is bad or rushed CGI in movies not needing experts anymore as you can just run the footage through a Stable Diffusion Lora to add more detail. CGI studios are already going out of business because they don't get paid enough and whatever jobs they don't have are rushed. Now these studios, which relied on jobs being outsourced to them by the studios, are likely going to compete with software packages that can either fix bad renders or go even farther.
On the bright side, every anime is going to get more episode drops per year... small clap?

[Krom]

Anime studios are famously abusive sweatshop work places with horrible pay and usually there are at least a couple stories a year about some animator actually dying from overwork; so that is one area I wouldn't actually be all that bothered by if AI replaced the majority of human labor on the in-between frames.

[Top Gun]

Why should you be "interested"? Such a weird question to ask. I wonder if "interested" was the right word... maybe you meant "...why should I enjoy..."? You should be interested in identifying the shift that's about to occur, because it's going to be everywhere at the commercial entertainment level. I'm not talking about artists on DeviantArt losing rank, but hundreds of thousands of outsourced animators who were paid to draw in-between frames now have competing ai companies trying to replace them by developing workflows that only need the two keyframes and the ai solves the inbetween frames.

Another area is bad or rushed CGI in movies not needing experts anymore as you can just run the footage through a Stable Diffusion Lora to add more detail. CGI studios are already going out of business because they don't get paid enough and whatever jobs they don't have are rushed. Now these studios, which relied on jobs being outsourced to them by the studios, are likely going to compete with software packages that can either fix bad renders or go even farther.
On the bright side, every anime is going to get more episode drops per year... small clap?

I'm not sure if "enjoy" is the right word, either. Maybe something like "acknowledge as an actual creative process," which I am absolutely not willing to do. As for being "interested," I will always stand against any use of technology designed to take the creative efforts of actual people with the unstated intent of actively replacing them, just so numbers can keep going up for shareholders. Count me out of that dystopian hellscape.

Anime studios are famously abusive sweatshop work places with horrible pay and usually there are at least a couple stories a year about some animator actually dying from overwork; so that is one area I wouldn't actually be all that bothered by if AI replaced the majority of human labor on the in-between frames.

...except starting with in-between work is how most animators gain the skillset to eventually become key animators. The Japanese animation industry is already experiencing a pretty substantial brain drain, probably largely due to the horrific working conditions you mentioned, so if you remove that pathway, then where is new talent going to come from?

[Krom]

...except starting with in-between work is how most animators gain the skillset to eventually become key animators. The Japanese animation industry is already experiencing a pretty substantial brain drain, probably largely due to the horrific working conditions you mentioned, so if you remove that pathway, then where is new talent going to come from?

People who have the talent will likely use the AI in-between software to fill in their own key frame hobby work and get noticed that way. Very few people actually advance to director/key frame/etc in the industry because the overwhelming majority burnout long before that point from the horrific/deadly working conditions. The current method of just grinding actual humans to dust is not sustainable and the industry has been showing cracks for a long time because of it. I think this is one of the areas where AI has the potential to actually do mostly good because it could make animation much more accessible for people who otherwise wouldn't survive the industry meat grinder.

[Top Gun]

People who have the talent will likely use the AI in-between software to fill in their own key frame hobby work and get noticed that way. Very few people actually advance to director/key frame/etc in the industry because the overwhelming majority burnout long before that point from the horrific/deadly working conditions. The current method of just grinding actual humans to dust is not sustainable and the industry has been showing cracks for a long time because of it. I think this is one of the areas where AI has the potential to actually do mostly good because it could make animation much more accessible for people who otherwise wouldn't survive the industry meat grinder.

Yeah, I do agree that there are obvious benefits to this software usage, but I think it has to be coupled with a far more substantial training pipeline. I've seen multiple industry veterans bemoan the lack of opportunities to train up new talent, and there have been a few small initiatives started up (things like the Japan Animator Expo), but there's a lot more to be done there. Drawing and animation are definitely fields where you learn best by doing, so there still needs to be a way to gain that experience, but in a sane and humane way. As it stands the industry has been forced more and more to rely on freelancers posting their own work on social media to get noticed, which is by no means sustainable.

[Isaac]

Top fun,

I don't think you can be counted out of the dystopian hellscape. Also, I don't think you've answered my question in my previous post:

"The question is if I took an artists entire gallery to train a Stable Diffusion model, would my output, which wouldn't match any photo in the artist's gallery, meet the requirements to be Transformative Use? It's possible I'm wording this clearly enough. Let me know if you need an example of what I'm saying."

[vision]

The current method of just grinding actual humans to dust is not sustainable and the industry has been showing cracks for a long time because of it.

I feel like it needs to be pointed out that the grinding down of people is a choice not a technical problem in need of a solution. Also, tweening artists are not just machines, it's just that the work feels that way (I've done big animation projects, I know the feeling). Those artists still contribute to the creative process by catching errors and making suggestions, something a AI is not likely to do. They are part of the team and that human experience counts. Also, it's more likely that a tweening artist will develop novel technical solutions rather than an art director since the tweening artist is doing the actual work. It's in their wheelhouse so to speak. But yes, I also agree that animation because of it's grueling mechanical labor could really use technical solutions provided those solutions don't have negative effects.

[Top Gun]

Top fun,

I don't think you can be counted out of the dystopian hellscape. Also, I don't think you've answered my question in my previous post:

"The question is if I took an artists entire gallery to train a Stable Diffusion model, would my output, which wouldn't match any photo in the artist's gallery, meet the requirements to be Transformative Use? It's possible I'm wording this clearly enough. Let me know if you need an example of what I'm saying."

I would say that if you took all of an artist's work without their explicit permission, trained a model on it, and then used it to spit out images in that artist's style, then no, that would not not count as transformative use and would be infringing on that artist's copyright. I'm of the opinion that "transformative work" implies some sort of basic creative process, i.e. an actual human being deliberately modifying the original work in some way for artistic purposes. This would not cover an algorithm that is essentially slicing said images into very tiny bits and stitching them together based on some external prompt (and yes I know that's a gross oversimplification of what's going on but you get my point).

[Isaac]

First let me say, thanks for giving me your full answer and I really do appreciate you taking the time to answer.

That being said, I do think you're wrong.

In United States copyright law, transformative use or transformation is a type of fair use that builds on a copyrighted work in a different manner or for a different purpose from the original, and thus does not infringe its holder's copyright.

Lets say I trained a model, specifically a LORA but that's not important, and then I generated cakes that had the shape and look of a generic looking princess that clearly draws inspiration from the LORA, but you couldn't identify a specific IP. According to your argument, Disney has the right to sue and Transformative Use does not apply because the process included training on copyright protected material, meaning the end result doesn't matter: the process/medium is the key factor.

Please correct me if my interpretation of your reply is wrong.

[Ferno]

We now have a way to fight back against AI theft.

[Tunnelcat]

Fascinating. I wonder how long it will take for A.I. to get around Glaze and Nightshade?

[TheWhat]

This is one of the AI songs making the rounds with audio people. Pretty freaking weird - I was expecting Euro-throb-Sprokets music... Not this kind. I guess some people can hear artifacts that give away the artificial intelligence

[Isaac]

We now have a way to fight back against AI theft.

So if it's theft, Tranformative Use can't apply? If it does can it be theft?

[Tunnelcat]

Hell, the words are a giveaway it's A.I. created and a creepy Dolly Parton??? imitation I'm guessing as well. That's the one thing that separates A.I. from humans. Emotions. The irrational logic set behind human emotions is so convoluted that an A.I. can never hope to copy it. At least I hope not, or we're all screwed if it ever gets pissed off or falls in love with something and becomes a stalker.

[Vander]

This is really good:
https://ludic.mataroa.blog/blog/i-will- ... -ai-again/

[vision]

This is really good:
https://ludic.mataroa.blog/blog/i-will- ... -ai-again/

LOL I read that this morning before coming to this site.

[Isaac]

This is really good:
https://ludic.mataroa.blog/blog/i-will- ... -ai-again/

It's absolutely spot-on for 99% of these businesses hyping up their products with ai, though actual use-cases are coming out as time progresses. Scammers are already taking advantage of new ai applications, from fake voices to official looking emails and letters. There are a growing amount of Instagram and Onlyfans models that are clearly ai, but too many people don't have a eye for spotting ai women. It doesn't help that the models keep getting better. AI artists with no actual artistic skill can now make new designs on their products to sell on etsy or deviant art. I'd hardly call these examples legitimate work, but it's one of the biggest legitimate uses.