There used to be pitchforks for cheating with your art, now it's standard practice. Oh well. At the very least art will still be valuable. We can throw peanut butter on it to demand climate change policy or burn it to sell NFTs. I've been suspecting distaste for and disrespect for art from the general public for some time. I think what they want is dada; tears on the pretty girl.
Near future - “Hey Netflix” give me Batman Vs Robocop in a Pixar style directed by Steven Spielberg, the film is a Viking time period set on the Highland boarders and they are the clan leaders.....play...... nah “Hey Netflix”, try Neill Blomkamp…….“Hey Netflix” add my AR glasses.....“Hey Netflix Pause” I have a call on my Neuralink brain chip.
“Hey Netflix” Add a third clan leader who is a hand with strong legs and a nice round derrière “Hey Netflix” put some clothes on the hand! 🙈
Neuralink brain chip, “Thinking in the head head -psst Netflix- bookmark the hand design” 🙊
Well ... not sure if the audience here "will like these" considering that these images are only possible because the model spitting them out has been trained on CG renders without any consent from the original artists.
Anyways - not even a few months in and this kind of images is already incredibly boring, somehow ...
Fair point, have you got any examples of where you can point this plagiarism out? I would like to see what this looks like and how different it is from the original, particularly with concept art.
I am like a moth to a flame with this tech, while I don't like the lack of consent and acknowledgement to the original artist my jaw drops with its improvement.
Its a total ‘Wild West’ of Generative AI at the moment. For example this lot, what do you think, Craig Mullins style for example? who knows but it’s amazing & scary how quickly AI is improving and is making valid acceptable designs.
"Fair point, have you got any examples of where you can point this plagiarism out?"
Well no, because that's not how the tech works... I personally don't care if the pictures can be qualified as plagiarism in the traditional sense of the word or not, because that's not the point. The point is that the model was trained on artwork without the consent of the artists, and this tells everything one needs to know about the people working on these commercial applications. They didn't have to do it that way, but they willingly chose to do so.
So in a way, similarly to how "the cat is out of the bag" in regards to the way these models were trained, it is also out of the bag when it comes to the perception of the people behind these tools. And from there, the perception of the pictures themselves.
At the end of the day the one thing that truly saddens me is the way this will potentially make the intense pleasure of designing something (with the huge amount of lessons one can learn from sticking to it, learning from mistakes, improving ones design vocabulary, and so on) even more out of reach to even more people. But I guess whoever goes straight to AI prompts probably doesn't have much of a desire to design anything to begin with. Or, they perhaps never had the chance to chat with artists, learn about their influences, hear about why someone choses to paint this or that way, and so on.
The point is that the model was trained on artwork without the consent of the artists,
But that's no crime since you don't need any consent to look at images that are public available. Since they are public available. Looking at other artists work is how an artist learns since humans exists. It was very common in the 19th century to work in the style of the old nederland masters for example. Why should a real artist be allowed to look at other people's images and art, and a AI not? Do you have the consent of Dali and of HR Giger to reuse their elements and style? That's one of the most common names when it comes to science fiction art, and this long before AI.
You dislike AI, no question. And even i am a bit biased here. Since it will cost jobs. But im am also highly excited since it saves me so much time. And there is simply no crime.
Interestingly enough, Stable Diffusion did react to the concerns. And has removed quite a few datatsets in version 2.0. You cannot use various artists names anymore. Greg Rutkowsky in the prompt will not longer work. But this was a gesture of good will. Not because they had to. And this has a massive downside The AI model is weaker than the first release now. And there is lots of umhappy users now. And this is just Stable Diffusion. DallE and Midjourney still uses all the data. And most probably will continue to do so.
At the end of the day the one thing that truly saddens me is the way this will potentially make the intense pleasure of designing something (with the huge amount of lessons one can learn from sticking to it, learning from mistakes, improving ones design vocabulary, and so on) even more out of reach to even more people.
I said it at another point already that i don't think this will happen. I don't fear for traditional art. AI is just another tool in the belt. And it is gambling. When you need an exact result like you really need it, then you still need to do it the traditional way. AI makes imho the world richter, not poorer.
"Looking at other artists work is how an artist learns since humans exists."
Yes, absolutely. But, there's a meaningful difference between how humans intake images and learn, vs feeding data into an algorithm explicitly made to regurgitate from the sources.
"Fair point, have you got any examples of where you can point this plagiarism out?"
yes indeed thare are plenty examples by now. the biggest issue right now is not that the tech exists but how its being used right now. the databases that are fed into them are simply just mass crawled from art sites such as deviantArt or Artstation without any consent. And they admit it in their copyright info. LAION 5B being the one behind stable diffusion and other tools.
License
We distribute the metadata dataset (the parquet files) under the Creative Common CC-BY 4.0 license, which poses no particular restriction. The images are under their copyright.
how can this be the baseline? "text is free, images are likely not, you go deal with it!"
then we had the shit with deviant art who now offer an opt-in model versus a forced per image opt-out option, which only came to be because the community was very loud and clear about their thoughts. and even then, the opt-in only "protects" from third party databases, dA now has their own AI database, trained on their target audience's work.
Clipstudio backpaddeling after they intergrated AI generators, basically preying on their own audience, also telling them "guys handle with care, this stuff is likely copyrighted, but its on you to make sure its safe to use", how could they ever have thought this is a good idea?
@Tiles : you are missing my point about this tool making the design process out of reach to some people. My point is not about the value or lack of value of the output - as a matter of fact, the assumption made by some people that "there will always be a need for (insert human job title)" is in my opinion quite short-sighted, especially when this very thread tracks the evolution of the tech quite well (from the earlier mangled monstrosities to the current results indistinguishable from high end fine art).
Rather, what I am trying to get at is that some people who may have otherwise become designers with something to contribute to this world might just end up giving up on the pursuit altogether, or, be easily distracted by the shiny tech. Of course everyone is responsible for their own actions, and if someone decides to give up, then so be it, it's their problem. But I still find it saddening to an extent purely from an altruistic perspective.
And I would also push back on the statement that the proliferation of such images brings more richness to the world, for two reasons :
1 - More is not always best. This is of course near impossible to quantify, but having an infinite amount of instant-pictures to work from is not necessarily helpful through a design process - similarly to how a moodboard with hundreds of pictures is near useless.
2 - Going beyond the surface quality of the images (which is now 100% indistinguishable from human made-art), the proliferation of such pictures just means ... lots of pictures with zero history behind them. Again, that doesn't mean that the pictures are not objectively pretty ; I just mean that, for instance, no one would gain anything from asking questions to the "prompt person" behind these Street Fighter AI renders about the choices made to place the light sources, the reasoning behind this or that design decision, and so on - because they wouldn't know. Similarly to how the prompt person behind these AI paintings would have no clue about brush stroke economy, value range, or composition.
Of course these points do not matter at all for many of the people on the receiving end of these pictures. But discussing the choices and influences behind a design or an illustration remains one of the most if not the most interesting thing about art and design. This to me has way more value than the pretty-ness of the pictures themselves. And even if an AI image came with its own AI-generated pseudo-history and list of influences, then I wouldn't be interested in that either because human exchange is just more fun. Just like playing an online game against humans is 100x more thrilling than playing against bots.
This is why I personally find these pictures to be uninteresting at the end of the day. Not because of the way they look, but because there is nothing to question or investigate behind them (outside of perhaps tracking back the artists who created the original pictures that the model was trained on !).
Well, the AI is in the wild for months now. And the graphics world is still not collapsed ;)
I still think you overestimate the impact. It has and will change the graphics world for sure. But i don't see it short sighted when i say that humans will still be required. There was always revolutionary stuff going on. And every change has made the world richer. Not poorer.
Well interestingly enough, it could probably be argued that the "it's just a tool" line of reasoning had more weight earlier on, when the tool sucked at anatomy. But now that the results are indistinguishable from photos of real people, the reconstruction of the Afghan girl (no doubt thanks to huge weight of the original picture in the training model, as well as the ability of the tools to refine anatomy thanks to the other millions of photographs used during training) crosses the threshold of straight up infringement. So to follow the Photoshop analogy, this would be similar to Adobe including unlicensed photos with their installs :D
Put differently : the issue isn't so much the fact that the tool can output the picture ; the issue is that it can output such an unlicensed picture without making the end-user aware of it. Meaning that any software licensing the tech is putting themselves and their users at risk, as shown with the case of CSP.
As said earlier I am personally not too interested in this side of things (the legal aspect) because of the absurdities it leads to - the moral side of things is imho more interesting and relevant, and then there's the whole topic of mental health too. But still the Afghan girl example definitely shows that at least in some circumstances, the training model still includes data that could be considered as straight up unlicensed use of an image, regardless of the way it was reconstructed by the "black box" process.
None of us necessarily possess the qualities and skills necessary to imagine and then build an automobile. Or a plane. Or to domesticate animals, or farm, or make clothing.
None of us really even possess the character and skills necessary to survive in the real world, unaltered by 10,000 years of technology which has rendered us into pretty pathetic slobs, at least compared to all the other animals which live by their wits and natural skill.
So yeah, I agree, people being able to make beautiful art by pushing a button might mean the end of something. Jobs? Maybe, maybe not. A particular set of skills and a type of character that is forged from the process of developing those skills? Most likely.
But that's just the onward march of technology. The only goal of technology is for people to have things without doing things. It's generally regarded as a good thing, but I disagree. I think we were better off as filthy animals, most having died at childbirth. But it's just words for words sake, nothing will stop peoples quest for convenience. Convenience will always win, even if it kills us all.
the moral side of things is imho more interesting and relevant
I am always very careful when somebody raises the word moral. Just look what religion did to people in the last two thousand years in the name of moral. Call me petty, but i am no fan of crusades, witch burning and telling people what kind of sex and even what kind of clothing is allowed.
I think what you mean is a philosophical and ethical problem. What if the artist becomes obsolete? But i can simply not share this fear. We have long solved this question in other areas. Robots and computers are long part of our society. And nobody moans about all the poor workers or mathematicians who lost their job. Since they did not. They have now even more work to do than ever before. It's just different work. And same will happen with art. Prompt generation has already become its own art form. And once more the world becomes richer.
the issue is that it can output such an unlicensed picture without making the end-user aware of it.
Same can be said about every artist. Do you remember the times when companies started to outsource artist jobs? Now it is outsourced to AI :)
no obviously not, because photoshop doesnt do these, but if an artist decides to plagiarize copyrighted images, he will get shunned at the lightest and sued at the hardest. but when using AI you can not know what and how much it stole and might break copyrights. if an artist does it when on hire you can sue him, can you sue the ai or ai maker when shit like this happens? no you can not.
Of course you can. The one who posts it is still in charge. And Photoshop has clone brush and intelligent background filling and all these gizmos that makes your life easier since a pretty while too.
but thats the issue, not knowing what is copyrighted makes it near im impossible to use besides randos doing shit on the internet.
in a professional environment that risk is just way too high
also photoshop doesnt sample stuff out of "thin air" it clones from what you give by sampling an area you pick and filling the background with stuff that is in your image, it is your decision.
Wn AI it is not, it takes "whatever" based on text inputs. PS doesnt crawl an near endless database of content of uncertain status to pick from and fill. at least so far.
Interesting point for sure. But this still claims that the AI uses existing images, which it does not. It is based at all the material that it has learned. You can with heavy tweaking also produce copyrighted material. Like you can do with Photoshop.
And still if this would be the case, then you simply have to do proper research. And a bit of common sense. A prompt that contains marvel characters or a famous existing image under copyright will most probably break copyright.
Well, this is a clear license break then. But what is your point with it? We have alreaydy arrived at that you can break license with every art tool. Also with AI.
In this case i would still have to sue you, not the AI :)
EDIT; this is by the way an interesting border case already. Here in europe you can reuse the artwork of another artist when you distort it in a way that it does not longer look like the original. So you might even get through with it :D
To be fair, regarding existing tools like content-aware fill : I personally have no idea if the tool is purely an algorithm, or if it does use a model trained on images. But if it is, then one can assume that Adobe was likely clever enough to train it on images that they had the right to use. But perhaps this is irrelevant anyways since it cannot produce full images on its own.
As for the moral aspect (or ethical, whatever we call it) : I am not so much talking about my perception of end-users and their use of the tools, but rather, talking about my personal perception of companies monetizing such models - similarly to how one is free to consider Ponzi schemes and faith healers to be immoral.
nobody says it always will produce something that is copyrighted, or meaningfull whatsoever. But it might, because it only knows what it has been fed.
inside say a studio environment where the studio feeds it with their own content, it becomes a safe tool to use, but filled with data is has zero clue about like midjourney or the public stable diffusion does, it is just problematic to use.
I understand your concerns. But as told, i don't share it :)
But if it is, then one can assume that Adobe was likely clever enough to train it on images that they had the right to use.
This is still the point where i heavily disagree. AI does not use Images. And so it does not need a right here. There is not a single copyright touched. No image parts are reused in the process. The AI has learned the style. And works now in style of artist x y. And you don't need the right to look at a public available image neither. So legally there is nothing wrong.
It is the artist who can steer the prompt to create copyright breaking images. Like you can break copyright with Photoshop created images too. I could add an distorted Image from Eric to my coffee mug using photoshop or Blender or any other art tool. And then we are back at the point that the artist is in charge to produce art that does not break copyright, not the tool.
Well, that's precisely the point that other posters have tried to explain just earlier : if a diffusion system is so strongly biased towards real life car logos for instance, then it can biased up to a point were it is very likely to slap a litteral BMW logo (or that of a lesser known brand) on the hood on some of the designs it generates. That's just an example of course, but this illustrates the very real risk taken by anyone using such a "black box" tool during the design process, and IMO that's a gamble that any sane studio doing creative work will not be willing to take. It being legal or not or the fine points of how it generates the designs do not really matter - the tool is just too risky to use in production, at least in the current state of things.
Even the mere background noise such a topic generates (endless watercooler discussions, heated exchanges during meetings between the "pros" and the "againsts") could be seen as a counter-productive nuisance really.
I disagree with people calling it "tool". This is not something you spend years of training to learn all the intricates like you do with oil paints and brush, with maya/blender/zbrush, with illustrator/affiniy designer etc. etc.
Or do you call website with free models (or paid models) to download and use as "tool" as well? That's what is the closest to compare "ai art". Instead of spending time with keywords on sites with stock models, you spend time writing more precise prompts.
But in the end, you don't do "crafting" yourself. You don't visually draw like with 2d or 3d, and don't even program graphics with code to get a circle and then array it 16 time via sinusoid path.
Results that you get from ai art generation is "MACRO" instead of "MICRO". Micro is having ellipses which you then construct yourself into something else. Macro is getting completed scene at once.
-
And if you still call it "just another tool", then for whom? For art directors or producers, maybe.
I disagree with people calling it "tool". This is not something you spend years of training to learn all the intricates like you do with oil paints and brush, with maya/blender/zbrush, with illustrator/affiniy designer etc. etc.
See my example image above with the blackbird and the earthworm. Yes you do. The skill that you need is to find the right keywords :)
And sometimes you still simply loose. I spent the whole day to put the earthworm into the mouth of the bird. No luck ...
Or do you call website with free models (or paid models) to download and use as "tool" as well?
Stock material is very common these days. But asides that, that's premade content then. With AI you create.
But in the end, you don't do "crafting" yourself. You don't visually draw like with 2d or 3d, and don't even program graphics with code to get a circle and then array it 16 time via sinusoid path.
Can't the same be said for rendering a textured wireframe? You did not paint it by yourself :)
AI image generation made a huge progress recently. The artworks it produces can be truly amazing, and it's absolutely fascinating how far the technology went.
But as someone who did arts in the past, I just can't stand people calling themselves "artists" by just typing text prompts and calling the AI generator "just a tool like photoshop". By that logic, anyone who ever commissioned art (you know, by using text descriptions) is an artist , and the artist they commissioned from is nothing more than just a "tool". I know a lot of people don't realize how much effort it takes to learn how to draw and I probably shouldn't be mad at their ignorance, but it truly pisses me off...
The AI is the artist here. You are just commisioning art from it. That's it.
I know we can't escape the technological advancement, that's not the point here. Just don't disrespect artists.
"AI does not use images" and yet above we have near-exact (down to the individual loose hairs hanging down over her forehead) recreations of that famous National Geographic photo down to most details. So how does that happen, if it "doesn't use images"? — images are data; if data is used, the image is used.
And beyond that, what if I've never seen the original, and I assume it's just amazing unique AI art? There's not always an easy way to check something like this. Tineye and reverse image searching can help when images are very close, but in many cases, I imagine it would be useless for catching copyright infractions.
And now we need an AI tool to find what copyrights our AI images violate....
"AI does not use images" and yet above we have near-exact (down to the individual loose hairs hanging down over her forehead) recreations of that famous National Geographic photo down to most details. So how does that happen, if it "doesn't use images"? — images are data; if data is used, the image is used.
That's a false claim, sorry. You can paint a equal image with photoshop too. Does Photoshop reuse the image then?
SD comes with a 4 gb file of AI data. This wouldn't even be enough to just put all the artstation images into the file.
But as someone who did arts in the past, I just can't stand people calling themselves "artists" by just typing text prompts and calling the AI generator "just a tool like photoshop"
I will nevertheless continue to call it a tool ;)
Is light installations art? Is video installations art? Is action painting art? Artist is a very broad term.
Anyways. Most of the users of these tools that i know are real artists with years of experience in other tools. Not some people from the street. We don't talk about apes or elephants here, pressing a button. You still need an artists eye to judge the result. And artists experience and knowledge to figure out the best prompts. But yeah, i can understand that it bothers you when you need a week for an image and somebody else just types in a few words into a prompt.
For me allowed is what works. And for me it simply matters if i need a week or a few hours to finish an artwork. I will use the most efficient way to reach my goal. Even better when it delivers a level of quality that i can't reach with my limited skills. And AI can create movies that are near to impossible to create with traditional art. Search for Deforum Stable Diffusion. A really nice tool ...
And beyond that, what if I've never seen the original, and I assume it's just amazing unique AI art? There's not always an easy way to check something like this. Tineye and reverse image searching can help when images are very close, but in many cases, I imagine it would be useless for catching copyright infractions.
The same as if you draw a plumber with a red cap and a black mustache. Or when you write a song that sounds similar to another song. It does not matter if you know the other art. When it is too similar then it is a break of copyright and design.
Yes, i see the danger too. Popular prompts will create similar images. And so the more people uses this tool, the more likely you will create something that looks similar to an existing one because of the sheer amount of existing similar images*. You can even create equal images with the same prompts and same seed. But again, it is the artist who is responsible just to release material where he has the rights for. And here we are at the point that even console artists are also artists. To create unique art you first and foremost need imagination. To repeat the popular prompts will not do it.
*it will be interesting what the lawyers will say here in the future. But for now we have the existing copyright laws. Made before AI was invented. First one to release is taker ...
And now we need an AI tool to find what copyrights our AI images violate....
No bad idea. Good that somebody else already thought about it. You even name it :)
Oh the topic of AI tool to find infringements, I have no doubt we'll start to see AI tools that do that, look for existing IPs or matching images and define how heavy it's influencing the result. We'll see AI that will give you what prompts to use to create a similar image, effectively making "being good at making prompts" become meaningless.
I'm a little surprised when creating the model they didn't filter out matching images (Afghan Girl, Starry Night, etc), I'm really curious how many times they had to be used for the model to have a heavy enough weighting to create them. You also notice this with celebrities and actors. If there's one really famous photo, poster, or scene, and you generate a prompt with them, it'll often heavily reference or reuse that face.
Some concepts everyone needs to learn if they keep discussing AI art especially Tiles.
Contributory copyright infringement is a way of imposing secondary liability for infringement of a copyright. It is a means by which a person may be held liable for copyright infringement even though he or she did not directly engage in the infringing activity.[1] In the United States, the Copyright Act does not itself impose liability for contributory infringement expressly. It is one of the two forms of secondary liability apart from vicarious liability. Contributory infringement is understood to be a form of infringement in which a person is not directly violating a copyright but, induces or authorises another person to directly infringe the copyright.
This doctrine is a development of general tort law and is an extension of the principle in tort law that in addition to the tortfeasor, anyone who contributed to the tort should also be held liable.
Vicarious liability is a form of a strict, secondary liability that arises under the common law doctrine of agency, respondeat superior, the responsibility of the superior for the acts of their subordinate or, in a broader sense, the responsibility of any third party that had the "right, ability or duty to control" the activities of a violator. It can be distinguished from contributory liability, another form of secondary liability, which is rooted in the tort theory of enterprise liability because, unlike contributory infringement, knowledge is not an element of vicarious liability.[1] The law has developed the view that some relationships by their nature require the person who engages others to accept responsibility for the wrongdoing of those others. The most important such relationship for practical purposes is that of employer and employee
Contributory trademark infringement
To be liable for contributory trademark infringement, the defendants would have had to have intentionally induced the primary infringer to infringe, or continued to supply an infringing product to an infringer with knowledge that the infringer is mislabeling the particular product supplied." Inwood Labs., Inc. v. Ives Labs., Inc., 456 U.S. 844, 855 (1982).
Perfect 10 alleged that the defendants played a critical role within the stolen-content websites' cycle of business that allowed the direct infringers to be profitable. They also claimed that the websites were using the PERFECT 10 mark in a manner likely to cause the public to believe that these stolen-content websites were authorized by Perfect 10. Perfect 10 continued by saying that Visa can choose to stop processing payments to these websites, which might have the practical effect of stopping or reducing the infringing activity.
The defendant's payment processing system was not an infringing product in itself, nor was it designed to intentionally induce trademark infringement, so contributory trademark infringement was not established.
Napster's facilitation of the transfer of copyrighted material raised the ire of the Recording Industry Association of America (RIAA), which almost immediately—on December 6, 1999—filed a lawsuit against the popular service.[20] The service would only get bigger as the trial, meant to shut down Napster, also gave it a great deal of publicity. Soon millions of users, many of whom were college students, flocked to it. After a failed appeal to the Ninth Circuit Court, an injunction was issued on March 5, 2001 ordering Napster to prevent the trading of copyrighted music on its network.[21]
Lawrence Lessig[22] claimed, however, that this decision made little sense from the perspective of copyright protection: "When Napster told the district court that it had developed a technology to block the transfer of 99.4 percent of identified infringing material, the district court told counsel for Napster 99.4 percent was not good enough. Napster had to push the infringements 'down to zero.' If 99.4 percent is not good enough," Lessig concluded, "then this is a war on file-sharing technologies, not a war on copyright infringement."
What is copyright infringement?
As a general matter, copyright infringement occurs when a copyrighted work is reproduced, distributed, performed, publicly displayed, or made into a derivative work without the permission of the copyright owner.
I am not sure what your point is. Since copyright is especially the part that is not touched by the AI solutions. Yes you can produce copyrighted material with it, like with Photoshop. No, it does not use or produce copyrighted material by design, like Photoshop. It is a tool to create graphics. And that's not my opinion but a simple matter of fact. Just ask yourself why we haven't seen a single lawsuit against DallE, Midjourney and Co. so far. And Napster has for sure absolutely nothing to do with the matter.
Proko makes this interesting point at around 12 minutes -
"Ai art is just a very small piece of research that comes under the umbrella of Artificial General Intelligence (AGI) and all the small companies making AI art picked up on this tiny piece of tech almost by an accidental open source side effect of the tools being there. Most companies are doing the “Art Ai thing” to get to the utopia which is to make AGI - Artificial General Intelligence. If you quantify AGI by a trillion times you get a trillion brain solving problems being solved. "
This thread has really helped me better frame my thoughts around AI generated art.
The biggest surprise for me is how it's happening faster than I had expected and the most valuable part of the tool (the image dataset) is the least expensive (monetarily).
While it's super awesome seeing such advancements technologically speaking, I wish it also translates into fields that would have much more real immediate impact on quality of life for humans (health/medicine).
None of these tools resolve the issue of copyright infringement, even with a "clean" dataset.
Randomly generating pixels still runs the risk of randomly generating copyrighted material.
Scanning each image for probable copyright infringement is way way too costly. It's expensive enough to generate 1 image, it'll be crazy costly to check that 1 image against all human recorded records for copyright infringement.
I think a lot of the reason I dislike a lot of anti-ai artist arguments is because I disagree with their perception of art, talent, tools, and "cheating". I'm not a fan of AI art in terms of nearly finished pieces coming out of midjourny, but I still see it as a valid tool.
I don't like viewing random online collections of AI art in the same way I don't like viewing random online collections of human art. There's a lot of noise, and a lot of it becomes meaningless without context or intent. I see AI filtering you the visual noise you ask for.
Talent doesn't equal fame and pay. There's been tens of thousands of talented artists, plenty even more technically skilled than the well known figures of art history. Sometimes their fame is luck, sometimes it's from being prolific, sometimes they were just doing the trend and were ahead of their time. The craft and the skills have always changed with time, demand for realism in paintings dropped when photography became a thing. People have to adapt.
Plenty of things have been called cheating, AI, digital art, photography, various forms of tracing and projectors, etc. Work faster and smarter, not harder. I don't believe using AI to generate background details is any more cheating than photobashing. Artists aren't going to be able to stop AI art just by calling it cheating, mocking how it does hands, claiming it can't currently do certain things (a lot of times they're wrong), etc.
I do believe there's a lot of potential copyright issues, but no more than what already existed within various spaces of art and expression.
a Pandora's box of sorts, there is no undoing it, consumers, employers, will be unable to tell the difference between hours of strokes or a single click. the art techniques, processes, and philosophies and results are all publicly available online. exponentially soon.
I hate to necro this thread but I just have to, to point out that, well, NOW we have seen lawsuits.
Also, I have to point out that the dataset used to train the AI models was intended for academic and research purposes only, not to be used for commercial profit, in case you didn't know. That's why the dataset has scraped images regardless of copyright.
But both Dall-E and Midjourney are charging money now. The models themselves don't contain copyrighted images, but the dataset used to train them DOES. It's a fact. There's a website where you can check if an image is in the dataset, and I found my paintings in there which I never gave permission to use that way. And since it was intended for research purposes, the models trained on it can't be used for profit.
If we want a fully legal for-profit generative AI, it MUST be trained on public domain images, or they have to purchase rights to all proprietary images they want to use to train it. And it's a question, how well such AI would work.
On Tuesday, the US Copyright Office declared that images created using the AI-powered Midjourney image generator for the comic book Zarya of the Dawn should not have been granted copyright protection, and the images' copyright protection will be revoked.
"(...) authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship."
Ha. The general counsel for Midjourney trying to claim it's a victory, when it's very clearly not what was written...
Copyright Office: ""The fact that Midjourney's specific output cannot be predicted by users makes Midjourney different for copyright purposes than other tools used by artists."
Midjourney's counsel: "[they're] clearly saying that if an artist exerts creative control over an image generating tool like Midjourney ...the output is protectable."
First, I think it's a cool proof of concept and a good demo of where the tech is right now. And the response has been very divisive. People saying they should be sued into bankruptcy, they have no ethics, that it's basically rotoscoping, that it's unwatchable garbage in quality, that they are talentless hacks, etc.
Personally I feel like someone had to do it, and that this was the direction things were headed anyway. Rather a small studio do something like this than NFT bros or some giant corporation lead the way. It is still a little unstable, but looks better than most straight up or automatic rotoscoping I've seen. To me, it was more watchable than even some low budget anime or filler episodes, wouldn't mind watching a show using an approach like this in places as a shortcut if it was more stable or cleaned up by hand. For the action and emotions it often feels like it's lacking expression and intensity, but not always.
Replies
There used to be pitchforks for cheating with your art, now it's standard practice. Oh well. At the very least art will still be valuable. We can throw peanut butter on it to demand climate change policy or burn it to sell NFTs. I've been suspecting distaste for and disrespect for art from the general public for some time. I think what they want is dada; tears on the pretty girl.
Aww, pitchforks! I remember the Poser drama ... :D
I should add, and it might just be the morning fog or the coffee and old fashioned donut, uh... what was I saying? nevermind.
Near future - “Hey Netflix” give me Batman Vs Robocop in a Pixar style directed by Steven Spielberg, the film is a Viking time period set on the Highland boarders and they are the clan leaders.....play...... nah “Hey Netflix”, try Neill Blomkamp…….“Hey Netflix” add my AR glasses.....“Hey Netflix Pause” I have a call on my Neuralink brain chip.
“Hey Netflix” Add a third clan leader who is a hand with strong legs and a nice round derrière “Hey Netflix” put some clothes on the hand! 🙈
Neuralink brain chip, “Thinking in the head head -psst Netflix- bookmark the hand design” 🙊
Thought you might like these (or maybe not)
Well ... not sure if the audience here "will like these" considering that these images are only possible because the model spitting them out has been trained on CG renders without any consent from the original artists.
Anyways - not even a few months in and this kind of images is already incredibly boring, somehow ...
Fair point, have you got any examples of where you can point this plagiarism out? I would like to see what this looks like and how different it is from the original, particularly with concept art.
I am like a moth to a flame with this tech, while I don't like the lack of consent and acknowledgement to the original artist my jaw drops with its improvement.
Its a total ‘Wild West’ of Generative AI at the moment. For example this lot, what do you think, Craig Mullins style for example? who knows but it’s amazing & scary how quickly AI is improving and is making valid acceptable designs.
What would your advice be to students regarding all this AI tech?
"Fair point, have you got any examples of where you can point this plagiarism out?"
Well no, because that's not how the tech works... I personally don't care if the pictures can be qualified as plagiarism in the traditional sense of the word or not, because that's not the point. The point is that the model was trained on artwork without the consent of the artists, and this tells everything one needs to know about the people working on these commercial applications. They didn't have to do it that way, but they willingly chose to do so.
So in a way, similarly to how "the cat is out of the bag" in regards to the way these models were trained, it is also out of the bag when it comes to the perception of the people behind these tools. And from there, the perception of the pictures themselves.
At the end of the day the one thing that truly saddens me is the way this will potentially make the intense pleasure of designing something (with the huge amount of lessons one can learn from sticking to it, learning from mistakes, improving ones design vocabulary, and so on) even more out of reach to even more people. But I guess whoever goes straight to AI prompts probably doesn't have much of a desire to design anything to begin with. Or, they perhaps never had the chance to chat with artists, learn about their influences, hear about why someone choses to paint this or that way, and so on.
The point is that the model was trained on artwork without the consent of the artists,
But that's no crime since you don't need any consent to look at images that are public available. Since they are public available. Looking at other artists work is how an artist learns since humans exists. It was very common in the 19th century to work in the style of the old nederland masters for example. Why should a real artist be allowed to look at other people's images and art, and a AI not? Do you have the consent of Dali and of HR Giger to reuse their elements and style? That's one of the most common names when it comes to science fiction art, and this long before AI.
You dislike AI, no question. And even i am a bit biased here. Since it will cost jobs. But im am also highly excited since it saves me so much time. And there is simply no crime.
Interestingly enough, Stable Diffusion did react to the concerns. And has removed quite a few datatsets in version 2.0. You cannot use various artists names anymore. Greg Rutkowsky in the prompt will not longer work. But this was a gesture of good will. Not because they had to. And this has a massive downside The AI model is weaker than the first release now. And there is lots of umhappy users now. And this is just Stable Diffusion. DallE and Midjourney still uses all the data. And most probably will continue to do so.
At the end of the day the one thing that truly saddens me is the way this will potentially make the intense pleasure of designing something (with the huge amount of lessons one can learn from sticking to it, learning from mistakes, improving ones design vocabulary, and so on) even more out of reach to even more people.
I said it at another point already that i don't think this will happen. I don't fear for traditional art. AI is just another tool in the belt. And it is gambling. When you need an exact result like you really need it, then you still need to do it the traditional way. AI makes imho the world richter, not poorer.
"Looking at other artists work is how an artist learns since humans exists."
Yes, absolutely. But, there's a meaningful difference between how humans intake images and learn, vs feeding data into an algorithm explicitly made to regurgitate from the sources.
If a human artist makes a new piece by using a Clone brush to lift bits of various other artists' work, would that be plagiarism?
And could artists recognize bits of their work in that piece?
These algorithms are basically doing the same thing.
"Fair point, have you got any examples of where you can point this plagiarism out?"
yes indeed thare are plenty examples by now. the biggest issue right now is not that the tech exists but how its being used right now. the databases that are fed into them are simply just mass crawled from art sites such as deviantArt or Artstation without any consent. And they admit it in their copyright info. LAION 5B being the one behind stable diffusion and other tools.
License
We distribute the metadata dataset (the parquet files) under the Creative Common CC-BY 4.0 license, which poses no particular restriction. The images are under their copyright.
how can this be the baseline? "text is free, images are likely not, you go deal with it!"
then we had the shit with deviant art who now offer an opt-in model versus a forced per image opt-out option, which only came to be because the community was very loud and clear about their thoughts. and even then, the opt-in only "protects" from third party databases, dA now has their own AI database, trained on their target audience's work.
Clipstudio backpaddeling after they intergrated AI generators, basically preying on their own audience, also telling them "guys handle with care, this stuff is likely copyrighted, but its on you to make sure its safe to use", how could they ever have thought this is a good idea?
@Tiles : you are missing my point about this tool making the design process out of reach to some people. My point is not about the value or lack of value of the output - as a matter of fact, the assumption made by some people that "there will always be a need for (insert human job title)" is in my opinion quite short-sighted, especially when this very thread tracks the evolution of the tech quite well (from the earlier mangled monstrosities to the current results indistinguishable from high end fine art).
Rather, what I am trying to get at is that some people who may have otherwise become designers with something to contribute to this world might just end up giving up on the pursuit altogether, or, be easily distracted by the shiny tech. Of course everyone is responsible for their own actions, and if someone decides to give up, then so be it, it's their problem. But I still find it saddening to an extent purely from an altruistic perspective.
And I would also push back on the statement that the proliferation of such images brings more richness to the world, for two reasons :
1 - More is not always best. This is of course near impossible to quantify, but having an infinite amount of instant-pictures to work from is not necessarily helpful through a design process - similarly to how a moodboard with hundreds of pictures is near useless.
2 - Going beyond the surface quality of the images (which is now 100% indistinguishable from human made-art), the proliferation of such pictures just means ... lots of pictures with zero history behind them. Again, that doesn't mean that the pictures are not objectively pretty ; I just mean that, for instance, no one would gain anything from asking questions to the "prompt person" behind these Street Fighter AI renders about the choices made to place the light sources, the reasoning behind this or that design decision, and so on - because they wouldn't know. Similarly to how the prompt person behind these AI paintings would have no clue about brush stroke economy, value range, or composition.
Of course these points do not matter at all for many of the people on the receiving end of these pictures. But discussing the choices and influences behind a design or an illustration remains one of the most if not the most interesting thing about art and design. This to me has way more value than the pretty-ness of the pictures themselves. And even if an AI image came with its own AI-generated pseudo-history and list of influences, then I wouldn't be interested in that either because human exchange is just more fun. Just like playing an online game against humans is 100x more thrilling than playing against bots.
This is why I personally find these pictures to be uninteresting at the end of the day. Not because of the way they look, but because there is nothing to question or investigate behind them (outside of perhaps tracking back the artists who created the original pictures that the model was trained on !).
Well, the AI is in the wild for months now. And the graphics world is still not collapsed ;)
I still think you overestimate the impact. It has and will change the graphics world for sure. But i don't see it short sighted when i say that humans will still be required. There was always revolutionary stuff going on. And every change has made the world richer. Not poorer.
Well, you can do all of these images also in Photoshop. Shall we now forbid Photoshop because you can create plagiat images with it?
Also with AI, the artist is responsible not to break copypright, not the tool.
Well interestingly enough, it could probably be argued that the "it's just a tool" line of reasoning had more weight earlier on, when the tool sucked at anatomy. But now that the results are indistinguishable from photos of real people, the reconstruction of the Afghan girl (no doubt thanks to huge weight of the original picture in the training model, as well as the ability of the tools to refine anatomy thanks to the other millions of photographs used during training) crosses the threshold of straight up infringement. So to follow the Photoshop analogy, this would be similar to Adobe including unlicensed photos with their installs :D
Put differently : the issue isn't so much the fact that the tool can output the picture ; the issue is that it can output such an unlicensed picture without making the end-user aware of it. Meaning that any software licensing the tech is putting themselves and their users at risk, as shown with the case of CSP.
As said earlier I am personally not too interested in this side of things (the legal aspect) because of the absurdities it leads to - the moral side of things is imho more interesting and relevant, and then there's the whole topic of mental health too. But still the Afghan girl example definitely shows that at least in some circumstances, the training model still includes data that could be considered as straight up unlicensed use of an image, regardless of the way it was reconstructed by the "black box" process.
None of us necessarily possess the qualities and skills necessary to imagine and then build an automobile. Or a plane. Or to domesticate animals, or farm, or make clothing.
None of us really even possess the character and skills necessary to survive in the real world, unaltered by 10,000 years of technology which has rendered us into pretty pathetic slobs, at least compared to all the other animals which live by their wits and natural skill.
So yeah, I agree, people being able to make beautiful art by pushing a button might mean the end of something. Jobs? Maybe, maybe not. A particular set of skills and a type of character that is forged from the process of developing those skills? Most likely.
But that's just the onward march of technology. The only goal of technology is for people to have things without doing things. It's generally regarded as a good thing, but I disagree. I think we were better off as filthy animals, most having died at childbirth. But it's just words for words sake, nothing will stop peoples quest for convenience. Convenience will always win, even if it kills us all.
the moral side of things is imho more interesting and relevant
I am always very careful when somebody raises the word moral. Just look what religion did to people in the last two thousand years in the name of moral. Call me petty, but i am no fan of crusades, witch burning and telling people what kind of sex and even what kind of clothing is allowed.
I think what you mean is a philosophical and ethical problem. What if the artist becomes obsolete? But i can simply not share this fear. We have long solved this question in other areas. Robots and computers are long part of our society. And nobody moans about all the poor workers or mathematicians who lost their job. Since they did not. They have now even more work to do than ever before. It's just different work. And same will happen with art. Prompt generation has already become its own art form. And once more the world becomes richer.
the issue is that it can output such an unlicensed picture without making the end-user aware of it.
Same can be said about every artist. Do you remember the times when companies started to outsource artist jobs? Now it is outsourced to AI :)
no obviously not, because photoshop doesnt do these, but if an artist decides to plagiarize copyrighted images, he will get shunned at the lightest and sued at the hardest. but when using AI you can not know what and how much it stole and might break copyrights. if an artist does it when on hire you can sue him, can you sue the ai or ai maker when shit like this happens? no you can not.
Of course you can. The one who posts it is still in charge. And Photoshop has clone brush and intelligent background filling and all these gizmos that makes your life easier since a pretty while too.
but thats the issue, not knowing what is copyrighted makes it near im impossible to use besides randos doing shit on the internet.
in a professional environment that risk is just way too high
also photoshop doesnt sample stuff out of "thin air" it clones from what you give by sampling an area you pick and filling the background with stuff that is in your image, it is your decision.
Wn AI it is not, it takes "whatever" based on text inputs. PS doesnt crawl an near endless database of content of uncertain status to pick from and fill. at least so far.
Tiles, I did some new AI art inspired by your work. I'm gonna sell coffee mugs now. No offense, it's just progress man, you can't stop the future!
Interesting point for sure. But this still claims that the AI uses existing images, which it does not. It is based at all the material that it has learned. You can with heavy tweaking also produce copyrighted material. Like you can do with Photoshop.
And still if this would be the case, then you simply have to do proper research. And a bit of common sense. A prompt that contains marvel characters or a famous existing image under copyright will most probably break copyright.
In charge is the artist, not the brush :)
I love examples ^^
Well, this is a clear license break then. But what is your point with it? We have alreaydy arrived at that you can break license with every art tool. Also with AI.
In this case i would still have to sue you, not the AI :)
EDIT; this is by the way an interesting border case already. Here in europe you can reuse the artwork of another artist when you distort it in a way that it does not longer look like the original. So you might even get through with it :D
To be fair, regarding existing tools like content-aware fill : I personally have no idea if the tool is purely an algorithm, or if it does use a model trained on images. But if it is, then one can assume that Adobe was likely clever enough to train it on images that they had the right to use. But perhaps this is irrelevant anyways since it cannot produce full images on its own.
As for the moral aspect (or ethical, whatever we call it) : I am not so much talking about my perception of end-users and their use of the tools, but rather, talking about my personal perception of companies monetizing such models - similarly to how one is free to consider Ponzi schemes and faith healers to be immoral.
This is by the way a blackbird with an earthworm in his beak ...
nobody says it always will produce something that is copyrighted, or meaningfull whatsoever. But it might, because it only knows what it has been fed.
inside say a studio environment where the studio feeds it with their own content, it becomes a safe tool to use, but filled with data is has zero clue about like midjourney or the public stable diffusion does, it is just problematic to use.
I understand your concerns. But as told, i don't share it :)
But if it is, then one can assume that Adobe was likely clever enough to train it on images that they had the right to use.
This is still the point where i heavily disagree. AI does not use Images. And so it does not need a right here. There is not a single copyright touched. No image parts are reused in the process. The AI has learned the style. And works now in style of artist x y. And you don't need the right to look at a public available image neither. So legally there is nothing wrong.
It is the artist who can steer the prompt to create copyright breaking images. Like you can break copyright with Photoshop created images too. I could add an distorted Image from Eric to my coffee mug using photoshop or Blender or any other art tool. And then we are back at the point that the artist is in charge to produce art that does not break copyright, not the tool.
Well, that's precisely the point that other posters have tried to explain just earlier : if a diffusion system is so strongly biased towards real life car logos for instance, then it can biased up to a point were it is very likely to slap a litteral BMW logo (or that of a lesser known brand) on the hood on some of the designs it generates. That's just an example of course, but this illustrates the very real risk taken by anyone using such a "black box" tool during the design process, and IMO that's a gamble that any sane studio doing creative work will not be willing to take. It being legal or not or the fine points of how it generates the designs do not really matter - the tool is just too risky to use in production, at least in the current state of things.
Even the mere background noise such a topic generates (endless watercooler discussions, heated exchanges during meetings between the "pros" and the "againsts") could be seen as a counter-productive nuisance really.
I disagree with people calling it "tool". This is not something you spend years of training to learn all the intricates like you do with oil paints and brush, with maya/blender/zbrush, with illustrator/affiniy designer etc. etc.
Or do you call website with free models (or paid models) to download and use as "tool" as well? That's what is the closest to compare "ai art". Instead of spending time with keywords on sites with stock models, you spend time writing more precise prompts.
But in the end, you don't do "crafting" yourself. You don't visually draw like with 2d or 3d, and don't even program graphics with code to get a circle and then array it 16 time via sinusoid path.
Results that you get from ai art generation is "MACRO" instead of "MICRO". Micro is having ellipses which you then construct yourself into something else. Macro is getting completed scene at once.
-
And if you still call it "just another tool", then for whom? For art directors or producers, maybe.
I disagree with people calling it "tool". This is not something you spend years of training to learn all the intricates like you do with oil paints and brush, with maya/blender/zbrush, with illustrator/affiniy designer etc. etc.
See my example image above with the blackbird and the earthworm. Yes you do. The skill that you need is to find the right keywords :)
And sometimes you still simply loose. I spent the whole day to put the earthworm into the mouth of the bird. No luck ...
Or do you call website with free models (or paid models) to download and use as "tool" as well?
Stock material is very common these days. But asides that, that's premade content then. With AI you create.
But in the end, you don't do "crafting" yourself. You don't visually draw like with 2d or 3d, and don't even program graphics with code to get a circle and then array it 16 time via sinusoid path.
Can't the same be said for rendering a textured wireframe? You did not paint it by yourself :)
>The skill that you need is to find the right keywords :)
Those are skills that copywriters use, not artist skills)
It's different field, in the end.
Dunno. For me imagination is one of the most important artistic skills.
Found this amazing comparison.
AI image generation made a huge progress recently. The artworks it produces can be truly amazing, and it's absolutely fascinating how far the technology went.
But as someone who did arts in the past, I just can't stand people calling themselves "artists" by just typing text prompts and calling the AI generator "just a tool like photoshop". By that logic, anyone who ever commissioned art (you know, by using text descriptions) is an artist , and the artist they commissioned from is nothing more than just a "tool". I know a lot of people don't realize how much effort it takes to learn how to draw and I probably shouldn't be mad at their ignorance, but it truly pisses me off...
The AI is the artist here. You are just commisioning art from it. That's it.
I know we can't escape the technological advancement, that's not the point here. Just don't disrespect artists.
"AI does not use images" and yet above we have near-exact (down to the individual loose hairs hanging down over her forehead) recreations of that famous National Geographic photo down to most details. So how does that happen, if it "doesn't use images"? — images are data; if data is used, the image is used.
And beyond that, what if I've never seen the original, and I assume it's just amazing unique AI art? There's not always an easy way to check something like this. Tineye and reverse image searching can help when images are very close, but in many cases, I imagine it would be useless for catching copyright infractions.
And now we need an AI tool to find what copyrights our AI images violate....
"AI does not use images" and yet above we have near-exact (down to the individual loose hairs hanging down over her forehead) recreations of that famous National Geographic photo down to most details. So how does that happen, if it "doesn't use images"? — images are data; if data is used, the image is used.
That's a false claim, sorry. You can paint a equal image with photoshop too. Does Photoshop reuse the image then?
SD comes with a 4 gb file of AI data. This wouldn't even be enough to just put all the artstation images into the file.
But as someone who did arts in the past, I just can't stand people calling themselves "artists" by just typing text prompts and calling the AI generator "just a tool like photoshop"
I will nevertheless continue to call it a tool ;)
Is light installations art? Is video installations art? Is action painting art? Artist is a very broad term.
Anyways. Most of the users of these tools that i know are real artists with years of experience in other tools. Not some people from the street. We don't talk about apes or elephants here, pressing a button. You still need an artists eye to judge the result. And artists experience and knowledge to figure out the best prompts. But yeah, i can understand that it bothers you when you need a week for an image and somebody else just types in a few words into a prompt.
For me allowed is what works. And for me it simply matters if i need a week or a few hours to finish an artwork. I will use the most efficient way to reach my goal. Even better when it delivers a level of quality that i can't reach with my limited skills. And AI can create movies that are near to impossible to create with traditional art. Search for Deforum Stable Diffusion. A really nice tool ...
And beyond that, what if I've never seen the original, and I assume it's just amazing unique AI art? There's not always an easy way to check something like this. Tineye and reverse image searching can help when images are very close, but in many cases, I imagine it would be useless for catching copyright infractions.
The same as if you draw a plumber with a red cap and a black mustache. Or when you write a song that sounds similar to another song. It does not matter if you know the other art. When it is too similar then it is a break of copyright and design.
Yes, i see the danger too. Popular prompts will create similar images. And so the more people uses this tool, the more likely you will create something that looks similar to an existing one because of the sheer amount of existing similar images*. You can even create equal images with the same prompts and same seed. But again, it is the artist who is responsible just to release material where he has the rights for. And here we are at the point that even console artists are also artists. To create unique art you first and foremost need imagination. To repeat the popular prompts will not do it.
*it will be interesting what the lawyers will say here in the future. But for now we have the existing copyright laws. Made before AI was invented. First one to release is taker ...
And now we need an AI tool to find what copyrights our AI images violate....
No bad idea. Good that somebody else already thought about it. You even name it :)
Google Reverse Image Search: https://www.labnol.org/reverse/
Oh the topic of AI tool to find infringements, I have no doubt we'll start to see AI tools that do that, look for existing IPs or matching images and define how heavy it's influencing the result. We'll see AI that will give you what prompts to use to create a similar image, effectively making "being good at making prompts" become meaningless.
I'm a little surprised when creating the model they didn't filter out matching images (Afghan Girl, Starry Night, etc), I'm really curious how many times they had to be used for the model to have a heavy enough weighting to create them. You also notice this with celebrities and actors. If there's one really famous photo, poster, or scene, and you generate a prompt with them, it'll often heavily reference or reuse that face.
Some concepts everyone needs to learn if they keep discussing AI art especially Tiles.
Contributory copyright infringement is a way of imposing secondary liability for infringement of a copyright. It is a means by which a person may be held liable for copyright infringement even though he or she did not directly engage in the infringing activity.[1] In the United States, the Copyright Act does not itself impose liability for contributory infringement expressly. It is one of the two forms of secondary liability apart from vicarious liability. Contributory infringement is understood to be a form of infringement in which a person is not directly violating a copyright but, induces or authorises another person to directly infringe the copyright.
This doctrine is a development of general tort law and is an extension of the principle in tort law that in addition to the tortfeasor, anyone who contributed to the tort should also be held liable.
Vicarious liability is a form of a strict, secondary liability that arises under the common law doctrine of agency, respondeat superior, the responsibility of the superior for the acts of their subordinate or, in a broader sense, the responsibility of any third party that had the "right, ability or duty to control" the activities of a violator. It can be distinguished from contributory liability, another form of secondary liability, which is rooted in the tort theory of enterprise liability because, unlike contributory infringement, knowledge is not an element of vicarious liability.[1] The law has developed the view that some relationships by their nature require the person who engages others to accept responsibility for the wrongdoing of those others. The most important such relationship for practical purposes is that of employer and employee
Contributory trademark infringement
To be liable for contributory trademark infringement, the defendants would have had to have intentionally induced the primary infringer to infringe, or continued to supply an infringing product to an infringer with knowledge that the infringer is mislabeling the particular product supplied." Inwood Labs., Inc. v. Ives Labs., Inc., 456 U.S. 844, 855 (1982).
Perfect 10 alleged that the defendants played a critical role within the stolen-content websites' cycle of business that allowed the direct infringers to be profitable. They also claimed that the websites were using the PERFECT 10 mark in a manner likely to cause the public to believe that these stolen-content websites were authorized by Perfect 10. Perfect 10 continued by saying that Visa can choose to stop processing payments to these websites, which might have the practical effect of stopping or reducing the infringing activity.
The defendant's payment processing system was not an infringing product in itself, nor was it designed to intentionally induce trademark infringement, so contributory trademark infringement was not established.
Lawsuit
Main article: A&M Records, Inc. v. Napster, Inc.
Napster's facilitation of the transfer of copyrighted material raised the ire of the Recording Industry Association of America (RIAA), which almost immediately—on December 6, 1999—filed a lawsuit against the popular service.[20] The service would only get bigger as the trial, meant to shut down Napster, also gave it a great deal of publicity. Soon millions of users, many of whom were college students, flocked to it. After a failed appeal to the Ninth Circuit Court, an injunction was issued on March 5, 2001 ordering Napster to prevent the trading of copyrighted music on its network.[21]
Lawrence Lessig[22] claimed, however, that this decision made little sense from the perspective of copyright protection: "When Napster told the district court that it had developed a technology to block the transfer of 99.4 percent of identified infringing material, the district court told counsel for Napster 99.4 percent was not good enough. Napster had to push the infringements 'down to zero.' If 99.4 percent is not good enough," Lessig concluded, "then this is a war on file-sharing technologies, not a war on copyright infringement."
What is copyright infringement?
As a general matter, copyright infringement occurs when a copyrighted work is reproduced, distributed, performed, publicly displayed, or made into a derivative work without the permission of the copyright owner.
I am not sure what your point is. Since copyright is especially the part that is not touched by the AI solutions. Yes you can produce copyrighted material with it, like with Photoshop. No, it does not use or produce copyrighted material by design, like Photoshop. It is a tool to create graphics. And that's not my opinion but a simple matter of fact. Just ask yourself why we haven't seen a single lawsuit against DallE, Midjourney and Co. so far. And Napster has for sure absolutely nothing to do with the matter.
Proko makes this interesting point at around 12 minutes -
"Ai art is just a very small piece of research that comes under the umbrella of Artificial General Intelligence (AGI) and all the small companies making AI art picked up on this tiny piece of tech almost by an accidental open source side effect of the tools being there. Most companies are doing the “Art Ai thing” to get to the utopia which is to make AGI - Artificial General Intelligence. If you quantify AGI by a trillion times you get a trillion brain solving problems being solved. "
This thread has really helped me better frame my thoughts around AI generated art.
The biggest surprise for me is how it's happening faster than I had expected and the most valuable part of the tool (the image dataset) is the least expensive (monetarily).
While it's super awesome seeing such advancements technologically speaking, I wish it also translates into fields that would have much more real immediate impact on quality of life for humans (health/medicine).
None of these tools resolve the issue of copyright infringement, even with a "clean" dataset.
Randomly generating pixels still runs the risk of randomly generating copyrighted material.
Scanning each image for probable copyright infringement is way way too costly. It's expensive enough to generate 1 image, it'll be crazy costly to check that 1 image against all human recorded records for copyright infringement.
Super cool to see such advancements in AI.
I think a lot of the reason I dislike a lot of anti-ai artist arguments is because I disagree with their perception of art, talent, tools, and "cheating". I'm not a fan of AI art in terms of nearly finished pieces coming out of midjourny, but I still see it as a valid tool.
I don't like viewing random online collections of AI art in the same way I don't like viewing random online collections of human art. There's a lot of noise, and a lot of it becomes meaningless without context or intent. I see AI filtering you the visual noise you ask for.
Talent doesn't equal fame and pay. There's been tens of thousands of talented artists, plenty even more technically skilled than the well known figures of art history. Sometimes their fame is luck, sometimes it's from being prolific, sometimes they were just doing the trend and were ahead of their time. The craft and the skills have always changed with time, demand for realism in paintings dropped when photography became a thing. People have to adapt.
Plenty of things have been called cheating, AI, digital art, photography, various forms of tracing and projectors, etc. Work faster and smarter, not harder. I don't believe using AI to generate background details is any more cheating than photobashing. Artists aren't going to be able to stop AI art just by calling it cheating, mocking how it does hands, claiming it can't currently do certain things (a lot of times they're wrong), etc.
I do believe there's a lot of potential copyright issues, but no more than what already existed within various spaces of art and expression.
a Pandora's box of sorts, there is no undoing it, consumers, employers, will be unable to tell the difference between hours of strokes or a single click. the art techniques, processes, and philosophies and results are all publicly available online. exponentially soon.
I hate to necro this thread but I just have to, to point out that, well, NOW we have seen lawsuits.
Also, I have to point out that the dataset used to train the AI models was intended for academic and research purposes only, not to be used for commercial profit, in case you didn't know. That's why the dataset has scraped images regardless of copyright.
But both Dall-E and Midjourney are charging money now. The models themselves don't contain copyrighted images, but the dataset used to train them DOES. It's a fact. There's a website where you can check if an image is in the dataset, and I found my paintings in there which I never gave permission to use that way. And since it was intended for research purposes, the models trained on it can't be used for profit.
If we want a fully legal for-profit generative AI, it MUST be trained on public domain images, or they have to purchase rights to all proprietary images they want to use to train it. And it's a question, how well such AI would work.
Here's some good news for once, I think -
From the article:
On Tuesday, the US Copyright Office declared that images created using the AI-powered Midjourney image generator for the comic book Zarya of the Dawn should not have been granted copyright protection, and the images' copyright protection will be revoked.
"(...) authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship."
Ha. The general counsel for Midjourney trying to claim it's a victory, when it's very clearly not what was written...
Copyright Office: ""The fact that Midjourney's specific output cannot be predicted by users makes Midjourney different for copyright purposes than other tools used by artists."
Midjourney's counsel: "[they're] clearly saying that if an artist exerts creative control over an image generating tool like Midjourney ...the output is protectable."
Haha. Midjourney making up their own "truth".
that's quite funny
but.. hangon
is the comic book itself still protected ?
cos
edit : actually read the article. and the answer is yes - which is correct.
the question now is how long before we see an unlicensed remix?
hmm, it might actually improve my work :)
Kinda surprised this thread didn't get bumped with the release of Corridor Digital's Anime Rock Paper Scissors
First, I think it's a cool proof of concept and a good demo of where the tech is right now. And the response has been very divisive. People saying they should be sued into bankruptcy, they have no ethics, that it's basically rotoscoping, that it's unwatchable garbage in quality, that they are talentless hacks, etc.
Personally I feel like someone had to do it, and that this was the direction things were headed anyway. Rather a small studio do something like this than NFT bros or some giant corporation lead the way. It is still a little unstable, but looks better than most straight up or automatic rotoscoping I've seen. To me, it was more watchable than even some low budget anime or filler episodes, wouldn't mind watching a show using an approach like this in places as a shortcut if it was more stable or cleaned up by hand. For the action and emotions it often feels like it's lacking expression and intensity, but not always.