Well that yikes, when i made the console and gamepads models i use artbreeder in order to get the textures and reference for modelling, and i though there's not any trouble, because in the terms of artbreeder i't says

"Image Ownershipand Intellectual Property
•EachArtbreederimage is owned by the user who createdthatimage.
•You agree to license any images you create on Artbreederunder theCreative
Commons CC0license. This effectively releases any image you create on Artbreeder
into the public domain. Such images may be usedby anyonefor any purpose, with
or without attribution.If you use another user’s image, please consider providing
attribution in the form of a link to the image’s lineage page."

link: https://www.artbreeder.com/terms.pdf

So should i remove the console and gamepads models i upload just in case?

I'm going to try to be brief here. My original was about fifty times as long, with excessive detail, because I am passionate about it.

We're currently at the intersection between people not understanding copyright law, people not understanding what companies can state in their terms of service, and people not understanding how AI art diffusion models work.

Lets go with the terms of service question first. Companies cannot have a contract that says, "By working here you give up your rights to collective bargaining and overtime and OSHA requirements" If it were that easy to get around the law, all companies would do it and employment laws would have no effect. Similarly, a software company can't say, "By using this site you agree that everyrthing you produce is not covered by copyright." I don't care if the terms of service say it, that's insane. Copyright law still exists. It's even protected internationally by the Berne Convention. You can't just wish it away.

Similarly, artists are spreading the fiction that if someone ever looked at their artwork in an attempt to learn how to make new images that *they* made the resulting images and should be paid. That's blatant overreach and copyfraud. And they better be careful, because what would stop another artist from claiming that their work sort of looks like their own work and demand money from them? There are already systems in place to determine what to do about inspiration and copying. and those rules, which they should be aware of if they are in that profession, do not at all support what they are claiming.

For example. Andy Warhol republished photographs of Prince without permission after he redrew the images and made a stylistic modification. Current copyright allows that, because the art was considered transformative. It's obvious, just looking at the photos and the artwork that he copied them. We even know he waas given access to photos the general public didn't have, which included those photos. He didn't take the photo of Prince and turn it into a grizzly bear with seven fingers holding a toaster. He took specific photos of Prince and traced them and made drawings of Prince that he sold to accompany an article about Prince. And it was perfvectly legal, according to actual court rulings. I bring this case up since it is in the news, so people who pay attention to the real world should already know this. It might get overturned in the Supreme Court. In fact, I hope they do overturn it. But this is an example of where the line currently is. That line is nowhere near what AI art does. AI art diffusion models set rules for how certain images are made. They then use prompting for specific kind of images and overlay it on random collection of static in various shapes, The different aspects means that the resulting images generally bear no resemblance to the initial images at all, except by trivial degrees. A cat has fur, let's make fur now. It's shaped this way, so, based upon the noise model we have, let's put a cat shape here. And so on. At times people who program AI muse that they might create a near copy of original source image by accident, but that's always been mere speculation. Pure random chance can do a lot of outrageous things. It has about the same chance of actually happening as a human artist has of inadvertantly making art that looks eerily similar to another artist's work. But AI art is less likely to do that. though, as they have taken extra steps after the fact that try to ensure that that never happens.

On the data collection end, Google was affirmed by a court to be able to scan copyrighted books and collect the data. They were even allowed to make the data be searchable and to produce pages of the books for free and available to the public. Check Google Books out sometime. They don't own those books, they are just there. On their side they have full digital copies of everything. If you search repeatedly and meticulously, you can get it to show you sample pages over time that would be the whole book. For free. Courts don't care. Now people collected data about images for making new images with AI. Why do they suddenly want us to believe that this is unfair or illegal?

Or, in other words, locking AI art so it can't be downloaded on this site seems really pointless.  In the absense of new court cases on AI art copyrights, we should default to like, you know, the actual copyright laws and not what some vocal subset of highly misinformed people on social media say. If they started a Twitter campaign tomorrow that said no art is covered by copyright, you'd ignore them. What they are saying now is equally baseless, just focused on AI art because it's easy to try to paint it as villain of the week because it is so new.

i make use of ai art alot in my games, mostly for title screens, but never in my oga submissions for exactly this reason.

it ultimately falls on the dataset, as MedicineStorm stated. it is new territory.

take this as an example. nightcafe studio is a site i use often. nightcafe studio's TOS says that creating art the IP rights are transferred to the creator. but look at these images based on a simple prompt. seems pretty obvious to me if the algorithm knows what "super mario" means, then the data set is not all open content. how can nightcafe grant me the IP rights on these images? i would say that they can't, even though their TOS says those images i generated are "mine."

The villainizing that dannorder is referring to is happening elsewhere, not here. And it is happening - in the last 12 months or so we've crossed a threshold where a large number of people suddenly become extremely spooked by a new technological development that doesn't fit into the old rules.

A somewhat similar thing happened with the media giants and file-sharing networks back in the 2000s... companies with real clout suddenly noticed their products being shared on a huge scale, and they went berserk. From what I understand, in the end it was clear that the public were just not going to give in to their bullying, so they took a completely different strategy, which worked great - they changed their business model to adopt the new tech, and since then nobody even talks about file-sharing networks any more.

The big difference this time is that it seems to be much less the giant media companies who are up in arms. It's small creators-for-hire and aspiring artists who are fearing for their livelihood and career prospects. All of the noise that I've seen echoes of, is really coming from the little people this time. Whereas with the file-sharing networks it was overwhelmingly the public versus the greedy corporations, with AI art we seem to have the public splitting into two camps.

You'd think that Nintendo and Disney (etc) would definitely care about protecting their rights to Mario and Elsa (etc), but perhaps they simply aren't threatened by fair use. They would probably never go after somebody drawing their characters for a personal portfolio, whereas they might go after someone who was selling figurines of their characters. So GenML users churning out endless images of their characters may not bother them. Users churning out images of their characters which then become CC0, and might lead to their rights to the character being Xeroxed, almost certainly will bother them, so I expect we will see them move heaven and earth to prevent that. But they may not care at all if people churn out characters that are simply influenced by Mario - whether they drew the character by hand, or got an AI tool to do it - because how does that threaten them? My guess is that we'll only see them try that "your character is influenced by ours, so we own your character" thuggery on the day where some character generated by little people using AI tools becomes huge enough to be really worth something; I believe there's some high profile precedent from the music industry for that kind of ambushing.

This is similar to file sharing in the sense that a lot of people are talking about it and it came about through technological advances, but the two cases are not really similar beyond that. In file sharing, people were making exact copies of things and distributing them electronically. If the files had all been copyright free, there wouldn't have been anything to complain about. The whole reason people were actually using file sharing was to reproduce exact copies and not pay for them. In AI art, the new art being produced is completely different from any previous image. The people complaining do not own those new images. There's nothing about those images that suggest they have any sort of legal claim whatsover, other than they just sort of wave their hand and say it is stealing. Copying discrete items under copyright is stealing. Making brand new things based upon what you learn by looking at other things is *not* stealing. I don't know if they are scared of technoilogy and shaking their arms like a muppet in its general direction or if they really think AI art must be like a patchwork quilt just stitching copyrighted images together because they can't conceive of any other possibility. It doesn't work like how they imagine it works.

The idea that using the art is dependent on an interpretation of fair use is the wrong way at looking at this. The only way that would make sense was if AI art were copying and pasting discrete elements of the original art. That, again, is not how the technology works. You can't line up any part of AI art to any art in original dataset. It uses the original art to devise complicated instructions on how to make different types of art. And, copyright expressly does not cover procedures / recipes / instructions.

If there were anyrthing, anything at all to the idea that there were copyright problems, the person making the claim should produce the original image that they think was infringed upon and then the AI image, and point out the similarities. The original database is public. People can search it. They are even using AI to try to find anything remotely similar to the AI art they are searching on. I've done it myself as a test. It's pretty eye opening. I thought you might get something thats sort of kind of close but not rising to the level of copyright infringement. What I found instead are a bunch of images that look *nothing similar* to any image actually in the database. I know people have a hard time grasping that, because they think it must be like a Xerox machine on ten different images and blurring them together or something. No, it doesn't do it that way.

People who still try to complain are only getting anywhere by twisting the meaning of words. It's "copying," and all copying is bad, especially digital copying, so AI art is bad. But the copying is not copying any discrete element. It's not copying anything covered by copyright.  It's more akin to McDonald's having a successful burger with red sauce and suddenly other fast food places introducing items with red sauce. That's copying, or derivative, in the broadest possible sense, but it's not copying or derivative by intellectual property law.

Fair Use is an attempt to justify why what otherwise would be infringing should not be considered infringing. This isn't fair use, because the part they are using - the instructions or rules that can be used to make the example images they are looking at - isn't even covered by copyright. When someone goes to art school and learns to be an artist, it's ridiculous to then claim that they are somehow tainted by that knowledge and everything they do after that must be infringing the copyrights of someone. But that's exactly what these people are claiming. I don't know why anyone is taking what they say seriously, other than the age-old issue of people being afraid of new things.

I'm going to pitch in with my thoughts and my current stand about AI, and i will keep it short.

I do believe it is great tech and the end results are absouletly brilliant, i can't deny that, and will help many in many different industry fields in various ways, and i would like to support it, but i am one of those that can't support it due to ethical reasons. The ends doesn't justify the means for me.

The AI algorythims used to create 'new works' are 'trained' using datasets. These datasets are 'images'. The images that comprise the dataset have been taken from creators and used without permission, and in some cases go against the license the creator has allowed. One of the technique used to do this is 'Scraping'.

The founders of the applications are aware of what they have done, and instead of working with creators, went along and did it anyway, ignoring creators wishes and licensing terms and generally copyright in all forms to progress thier software to what it is today. The applications are incredible, but it wouldn't be what it is today without the contributions of creators uploading their work to the online world, for which have had no say on whever they wanted to be a part of it or not, given no choice or even any sort of reconignition for the contribution, most importantly, against the license they chose for thier works.

I can't support this unethical approach and ignorant attitudes being/been used in current AI systems. Give me an AI that is using a dataset with works that have permissions and or license then im all for it, will support it and may even use it. :)

it would be a cool project to train your own dataset using only cc0 assets.

it'd be very time consuming though, as you would have to know python well, and it would require human identification of images. atleast that's how it would work with vqgan+clip and similar algorithms, atleast as i understand the tech. i have not studied how stable diffusion and how other algorithms work.

Unrelated to the court cases, elsewhere I wondered whether Stable Diffusion's model licence affects the terms that people could release images under, due to the "You are accountable for the Output you generate and its subsequent uses" clause ( https://opengameart.org/comment/99142#comment-99142 ). I noticed that someone has submitted an issue with Stable Diffusion making similar points at https://github.com/CompVis/stable-diffusion/issues/131 , so I thought I'd post the link here.

Unfortunately there's been no response though.

Hi, i have a question, i managed to run stablediffusion in my own PC (offline) mostly for make textures for 3d models, if i want to uploads those models with the textures in this site, I'm able to do that or i have to meet certaint requirements? do or avoid something specific?

Thanks in advance, i'm kinda lost with this situation.

Maybe you can add this to the list:
https://opengameart.org/content/pixel-art-creatures-from-clipdrop-stable...

I have a question, I'm making a videogame, and i'm using the this texture for a building: https://opengameart.org/content/golgotha-textures-widopeakwindowswalljpg (i'm also using others but i'm putting this one to explain my situation)

But i wanted to make the texture hight res, so i used this: https://www.iloveimg.com/upscale-image   it's an AI powered image upscaler.

My question is, in case i wanted to upload the building here with the upscaled texture, not the original, i'm able to?

Thanks, IDK if i'm clear, my english is not good.