The problems with AI art

[Oniya]

So, a couple of places that I frequent have been having this discussion, and I wanted to bring it up here.  There are a number of web-programs that will generate art/faces/etc, based on user prompts and things they have 'learned from'.

There are a couple of potential issues with this.  One that is a little scary is that if you put art/pictures on the Internet, these web programs can appropriate them as 'learning material'  One generator, called Lensa, has been called out for 'scraping' copyrighted images

From that article:

Adding a watermark may not be enough to protect artists — in a recent Twitter thread, graphic designer Lauryn Ipsum listed examples of the “mangled remains” of artists’ signatures in Lensa AI portraits.

One artist, trying to check to see if her own works had been scraped, found pictures from her own medical records in one of the 'training data sets'.  From an end-user point of view, there's no way of knowing what images the generator uses to compile an end-piece, and whether the subjects of those images knows or would want their image included. 

I don't know about folks here, but that's kind of creepy to me.

[RedRose]

This is worrying. How does medical data end up like this? I mean, I've heard interesting things from docs about patients - but this seems one step worse.

[Oniya]

The article says that the woman's doctor had died sometime between when the picture had been taken, and when she found it in the dataset - there's no way to definitively stamp when the image made it onto the 'net, but she presumes that it was after his death.  (Mind you, when I worked IT at a lab, we zero-formatted any computers that were retired prior to disposal or donation.)

[Callie Del Noire]

Dead or not someone is guilty of a HIPA violation.

[Azy]

I'm on the older edge of the Millennial generation and I should be all pro technology.  The average person my age is.  I admit that technology is useful in ways, but this is one good example of why I don't like  everything turning digital.  It's so easy for private information to be stolen.  I just had my credit card information stolen a few days ago.   

[RederReds]

Oh boy, as an artist I have opinions on AI art. The negatives far outweigh the positives, in my mind, but I think that ultimately the technology is neat but absolutely misused. It's good for quick ideation of ideas for artists and non-artists alike but... that's about it. That and it's not really AI, it's machine learning. Many people hear AI and believe that the program can think on its own when it's really not. Machine learning works on labeled datasets which have to have hundreds, if not thousands, of images to function correctly.

So, for example, you create a dataset with a bunch of pictures of college football games with them labeled based on the college. From there the program does any number of different math computations (convolution, mainly, but dear god I do not understand it despite taking a class on ML) and generates, essentially, an equation of what the optimal 'values' for any given feature on an image are. This allows the algorithm to now look at an image and identify key features such as color, shape, textures, edges, etc as it has an idea of the average 'values' of a given image. Now, art generation isn't the exact same, but the general process is. You get a large dataset of labeled images, run them through some sort of machine learning algorithm, and then you now have an algorithm that has an idea of what an 'apple' looks like. The only problem is you need a LOT of data to get anything near good so... Thus come the issues.

The datasets that these algorithms use are gigantic and sourced from scraping the internet for art without artists consent and without care for what it grabs. This, of course, has many artists unhappy because it's essentially mass larceny. There is no way to protect your art if you post it anywhere and even generators who are trying to be more ethical only have an "opt-out" feature where artists have to go piece by piece and tell the programmers to not use their work - talk about draining. Another key part of this is artists aren't being compensated in any way. Their art is being used to train an algorithm which is then being sold to others for use or, worse, being used by others and then re-sold as "original art." People have also done studies on the datasets used to train some of the most widely used generators and it's problematic to say the least. (Twitter thread link) Like with a lot of technology these algorithms often show the prejudice of society in them including sexism, racism, etc. For example if you search "flight attendant" in the dataset you are presented with women who are often sexualized or outright pornographic. Meanwhile if you search for a common male name you get no such results.

This doesn't even cover the general cultural attitude problem with the vocal subsection of the internet that tends to utilize these generators. I want to preface this by saying not everyone who uses an AI art generator is this way, but, there is a vocal subsection who are. There have been art models trained on the recently passed Kim Jung Gi hours after it was announced that he had passed on, models trained on multiple prolific living artist's work against their very vocal and very public will to not have their art be a part of these models, and models trained out of spite on the previously mentioned artists' work as a way to 'put them in their place' and impose 'supremacy' over them. Hopefully it can be seen why these decisions are not only anti-artist but also super disrespectful to the many affected. Beyond that, the general attitude these people show to artists speaking out about their work being stolen ranges from misinformed to outright threatening and toxic. I've seen people tell artists that they deserve to have their art stolen because they're "gatekeeping" art (?????) as well as people outright tell artists that they will be homeless and out of a job once the "AI revolution" comes. Now, some of this is general internet culture, but in my opinion it doesn't excuse the behavior.

One thing I will note, however, is that AI art tends to only be an issue for a few reasons:
1) The training data is taken from artists without consent or reimbursement
2) People are using AI art to phase out paid roles for artists
3) No one single person has the time to verify training data so you get, as mentioned, medical photos in the dataset as well as general prejudice being proliferated

[Kuroneko]

As another artist, I agree with RederReds completely.

I love technology, and I use digital tools, programs, etc to make artwork, concept art, and costume renderings, but the outright theft of images without the consent of artists is theft, plain and simple. It's unethical. The argument that all 'all artists learn from the work of others' presented by many AI defenders is also disingenuous. Do artists learn by doing studies of existing art? Absolutely. I have my own students do it to learn techniques and important art concepts. Does that mean I can say it's my own work? No. It's a study. It's a learning tool. It's practice, and artists spend years training to develop their own style. If it's theft for a human artist to copy another artist's work, then it's also theft for AI programs to do it. Many artists have shown that they've lost jobs because these programs harvested their work and can now replicate their signature style.

The best, most ethical use of this technology I've seen is using your own image as a base to quickly generate variations. I can see how this would be useful for a concept artist to speed up their workflow. But until the ethical questions regarding the image scrapping are answered, it's too problematic. Once you know these images were created using stolen images, they're very hard to appreciate, no matter how beautiful.

The fact that images of medical records were part of the Stable Diffusion image learning set is frightening.

[Vekseid]

The genie is not going back in the bottle. People need to be aware of what it means and what sorts of discussions need to be had, given the new reality.

To me, AI is a force multiplier. For artists, the end result I see is they will make images and use AI to generate thousands of permutations of each of them. AI art is still very much in its infancy, and 'prompt engineering' can only get you so far with it.

Oh boy, as an artist I have opinions on AI art. The negatives far outweigh the positives, in my mind, but I think that ultimately the technology is neat but absolutely misused. It's good for quick ideation of ideas for artists and non-artists alike but... that's about it. That and it's not really AI, it's machine learning. Many people hear AI and believe that the program can think on its own when it's really not. Machine learning works on labeled datasets which have to have hundreds, if not thousands, of images to function correctly.

This sort of objection has been made to every advancement in artificial intelligence over the past sixty years. The first artificial neural networks couldn't even perform a XOR operation. Now I can tell one to show me a picture of a chitinous apple being fired out of a golden cannon.

If you can tell me the difference between creating images from words and thought, I am all ears.

So, for example, you create a dataset with a bunch of pictures of college football games with them labeled based on the college. From there the program does any number of different math computations (convolution, mainly, but dear god I do not understand it despite taking a class on ML)

A class on image processing would have helped you more. : )

Every global filter you can apply to an image is a form of convolution performed on that image. Blurring, edge detection, sepia tone, whatever. Take two functions/objects and combine them to produce something new.

So when you train a convolutional neural network, it is picking out what convolutions reveal interesting aspects of an image, and permit it to recognize certain things.

You can then invert it and spit out the reverse. 'Apple' points to a number of convolutions, it picks an appropriate set and presents one for your enjoyment. In the case of more advanced models it is clear they do understand it as a 3-dimensional object.

And if it hasn't been trained on that data it just ignores those tokens. I could not get Dall-E to output certain types of hats no matter what I tried.

The datasets that these algorithms use are gigantic and sourced from scraping the internet for art without artists consent and without care for what it grabs. This, of course, has many artists unhappy because it's essentially mass larceny.

Fraud like this is only the tip of the iceberg here. I have also seen a number of 'text generators' that are basically glorified search engines. Some AI art generators could easily be doing what amounts to the same thing is they overfit their input.

[RederReds]

The genie is not going back in the bottle. People need to be aware of what it means and what sorts of discussions need to be had, given the new reality.

To me, AI is a force multiplier. For artists, the end result I see is they will make images and use AI to generate thousands of permutations of each of them. AI art is still very much in its infancy, and 'prompt engineering' can only get you so far with it.

This sort of objection has been made to every advancement in artificial intelligence over the past sixty years. The first artificial neural networks couldn't even perform a XOR operation. Now I can tell one to show me a picture of a chitinous apple being fired out of a golden cannon.

Yup, there's no stopping the advancement of AI, that much I agree with, but I am a firm believer that it can be done ethically. The current way of training these models is sketchy at best and outright illegal at worst. However, part of the reason I believe AI art has been taken so poorly is because it threatens careers for creatives. While I think adapting to AI generators is good and using it as a tool even better, the fact of the matter is many companies and commissioners will see AI generated art as "good enough" and protect their wallet. AI generated art wouldn't be as scary if it didn't literally threaten people's ability to live through their career. Automation is inevitable and we need to adapt but that's a whole other conversation about capitalism, socialism, the virtues of both, etc.

If you can tell me the difference between creating images from words and thought, I am all ears.

Not sure I follow here, apologies. If you want to expand more I'm more than willing to try and elaborate on the point.

A class on image processing would have helped you more. : )

Yeah I mean... It was a Mechanical Engineering class on Machine Learning so take that as you will :P I will not call myself an expert, just someone vaguely aware of the parts under the hood so I appreciate your more though explanation.

And if it hasn't been trained on that data it just ignores those tokens. I could not get Dall-E to output certain types of hats no matter what I tried.

Fraud like this is only the tip of the iceberg here. I have also seen a number of 'text generators' that are basically glorified search engines. Some AI art generators could easily be doing what amounts to the same thing is they overfit their input.

That is the downfall of AI art which is why, for example, we see a lot of the more popular generators stuck in a certain style or some can't draw hands or matching eyes. If they aren't trained on something then it just falls apart. It's the one light in the tunnel for me, personally, however I do believe that at the end of the day we'll see people take it as good enough and/or (hopefully) hire someone to make minor fixes. It's why I still feel secure offering commissions - these generators can't take a "just make the hat red" from a commissioner and tweak a piece the way that's desired. But, I've also seen friends lose their jobs because "AI will do your work :)" which is short sighted and shitty so I have a bit of a personal investment. I don't think these generators will ever disappear, but I would like to see something happen to respect the time and work of artists more.

[Oniya]

That is the downfall of AI art which is why, for example, we see a lot of the more popular generators stuck in a certain style or some can't draw hands or matching eyes.

If you want a laugh (or inspiration for alien life forms), tell one of them to produce a 'hand drawing tutorial'.  I've *seen* some things.

[Vekseid]

Yup, there's no stopping the advancement of AI, that much I agree with, but I am a firm believer that it can be done ethically. The current way of training these models is sketchy at best and outright illegal at worst. However, part of the reason I believe AI art has been taken so poorly is because it threatens careers for creatives. While I think adapting to AI generators is good and using it as a tool even better, the fact of the matter is many companies and commissioners will see AI generated art as "good enough" and protect their wallet. AI generated art wouldn't be as scary if it didn't literally threaten people's ability to live through their career. Automation is inevitable and we need to adapt but that's a whole other conversation about capitalism, socialism, the virtues of both, etc.

And yet you don't see programmers complaining much, which this stuff is already much better at.

It's mostly "I can do so much more now!"

Can automate so much tedium.

That said, there will need to be a conversation about how our economy functions, eventually.

Not sure I follow here, apologies. If you want to expand more I'm more than willing to try and elaborate on the point.

Neural network models are all built off of our own brain. Our brain is still vastly more complex than any current network, but the way data passes through layers of a NN is not so different from thought as we experience it.

Meanwhile

That is the downfall of AI art which is why, for example, we see a lot of the more popular generators stuck in a certain style or some can't draw hands or matching eyes. If they aren't trained on something then it just falls apart. It's the one light in the tunnel for me, personally, however I do believe that at the end of the day we'll see people take it as good enough and/or (hopefully) hire someone to make minor fixes. It's why I still feel secure offering commissions - these generators can't take a "just make the hat red" from a commissioner and tweak a piece the way that's desired. But, I've also seen friends lose their jobs because "AI will do your work :)" which is short sighted and shitty so I have a bit of a personal investment. I don't think these generators will ever disappear, but I would like to see something happen to respect the time and work of artists more.

Well it still breaks down the scene into its convolved components, in theory it should be able to maintain state awareness on an image in the same way ChatGPT does conversations.

Another thing I've noticed with the Stable Diffusion crowd, they have an unhealthy obsession with Emma Watson and I suspect it is a matter of time before legal precedent gets set for that as well.

[RederReds]

If you want a laugh (or inspiration for alien life forms), tell one of them to produce a 'hand drawing tutorial'.  I've *seen* some things.

Oh man yeah that sounds like it would produce some fun results. I wonder if it would replicate the "draw the rest of the owl" meme if you asked it for a tutorial on how to draw an owl.

Neural network models are all built off of our own brain. Our brain is still vastly more complex than any current network, but the way data passes through layers of a NN is not so different from thought as we experience it.

Ah so, as a tangent that might not address your question (which, apologies if that's the case) - I think the distinction in NN versus Artist is the process. While they may process information similarly the process in which art is created and the intent in which it is created are vastly different. Many artistic types tend to be almost spiritual about their work to some degree and I would have to say I'm one of them though minorly so. I enjoy the process and the concept of cutting any of my years of work out feels like I would be sullying what art means to me. The tedium in creating a piece is part of what I enjoy when it all pays off and perhaps that's the masochist in me but I digress.

Really it comes down to the abstract concept of 'soul' in work which there are endless debates on. From a material standpoint there are no differences between art created by a machine and a person, but from a standpoint of soul and intent there are differences. I think that's another part of the vitriol from artists - their art is personal to them. It's a part of who they are. Taking that without their permission is taking something that's a part of their person. That's part of the difference between the reaction programmers have versus artists, I think, but I could be very wrong. I've done both and I absolutely understand the appeal of taking tedium out of programming, but I've also only programmed lightly for a few years so I am nowhere near an expert on the subject.

Another thing I've noticed with the Stable Diffusion crowd, they have an unhealthy obsession with Emma Watson and I suspect it is a matter of time before legal precedent gets set for that as well.

Another thing I've seen is people suggesting people train a model on Disney's work as a way to force some sort of legal precedent. Any model doing that would get slapped down real fast by the big mouse.

[Kuroneko]

Please excuse the flip meme.

I snipped this bit of RederRed's post because this is also part of what I'm seeing from artists, and it's also something that resonated with me.

Really it comes down to the abstract concept of 'soul' in work which there are endless debates on. From a material standpoint there are no differences between art created by a machine and a person, but from a standpoint of soul and intent there are differences. I think that's another part of the vitriol from artists - their art is personal to them. It's a part of who they are. Taking that without their permission is taking something that's a part of their person. That's part of the difference between the reaction programmers have versus artists, I think, but I could be very wrong.

I can't speak for programmers either, beyond saying that I'm not sure it's an apples to apples comparison. Programmers and artists likely have different attitudes towards these issues, but maybe I'm assuming.  I personally can't see why would I want an AI program to do my art process for me. It's the process of creating it - and everything that goes into it, which sometimes include longs hours - which makes it enjoyable in the first place. It's the difference between buying a meal and cooking it yourself, I think. Can I buy a better meal than one I might make myself? Sure. But buying a meal will never replace making one of my mom's favorite recipes, or one that we used to make together as I grew up, where she taught me how to cook, season, and prepare a meal. There's more personal investment in the latter. The sense of accomplishment and satisfaction that comes from that isn't the same as buying dinner. Creating a work of art is the same, at least for me. I don't get that same experience from stringing together a set of prompts to feed into Midjourney. Maybe it's not the best analogy, lol. Of course, I'm only speaking for myself. I'm sure that there are plenty of artists who would love to cut down on some of the repetitive tasks that their work involves. I know lots of other costume designers that use their previous drawings of human figures for their renderings, rather than draw out new ones to speed up their process, for instance.

Other professors I know are now seeing papers written with chatGPT turned in by students, which means we'll soon have to come up with language in our syllabi and grading rubrics to address the use of AI text generators by students, as well as ways to determine if something was AI generated. I anticipate we'll be discussing it quite a bit at my university.

[firepyre]

I don't much think this will be a popular viewpoint amongst artists, but from a layperson's perspective, having access to AI tools that we can use to spit out art-like images is fantastic. No need to deal with copyright, and even someone with minimal skills can get a pretty decent looking result. For most uses where I just want to convey a concept or idea, I don't really care if it has soul or not.

On the other hand though, or something you hang on your wall, or in a gallery, something intended to be the center of conversation, then intent does matter. There is a place for art created with intent, and I think AI largely lacks that. That said, there has been art competitions won by AI compositions, so maybe I place too much importance on that. As the beholder, if I can't read into the artist's intent, then does it really matter if the piece was created by a machine or a human?

While it sucks that some artists might lose their jobs, that's really no different to any other industry -. In my opinion, neural networks have the potential to shake things up on a level similar to that of the industrial revolution. We've probably only seen the very tip of the iceberg yet. And like any big change, there's going to be a whole slew of new ethical and social issues to deal with, and a bunch of growing pains while society sorts out some kind of new normal. Exciting and scary in equal measure.

[Oniya]

The thing is, with the way the programs 'learn', there's no way to opt-out.  You upload prom pictures, and they could get scraped and conglomerated into one of those 'flight attendant' pictures that Reder mentioned.  Emma Watson might have access to good lawyers, but what about Judy Smith?

[GloomCookie]

My opinion is simply this. AI artwork does scrape from images, but then... everyone does. How do we not learn than from observing the classics and giving our own interpretation to it? For example, this:



Now I ask, is this one any different if it was made with AI or by a human hand?



The point I'm making is that all art is built on something else. What's the cutoff before you can claim that it's no longer an original piece and is just an amalgamation of other works? Is there a percentage of 'original' art that needs to exist? If so, does that 'original' have to be so unique that it has never existed before or since in our universe?

Also, there's a lot of artists mad that they can't charge as much for commissions because someone can put together an AI portrait of their D&D character. That's legit, but then not everyone wants to spend $50 for a character they might get to play three times before the game dissolves. And if it does go longer, so what? They have a representation of their character. If they want to bad enough, they can then take the AI image to an actual artist and say "Here's my character's base image, can you please draw them doing this epic thing?"

I see AI art as a tool, nothing more. It's useful for me to create something to tell the story, such as a beautiful city or a dark castle, something I can throw together in 5 minutes with some level of consistency (which is a huge deal for me) without having to do the same thing on Deviant Art or Google or Art Station for over an hour finding just the right picture that matches the art style I have going. I can instead type a few parameters into the AI program, get a few images that are fairly uniform in appearance, and gets the point across better.

And finally, I use AI to create a custom avatar for myself, and I don't have a fortune to spend on having every possible emotion or clothing combination just to put out a little blurb about how excited I am or that I'm munching breakfast with the family.

Or holding a kitty.

[Kuroneko]

There is no getting around the copyright issue. The issue is there whether we want to ignore it or not. I have no issue with the technology at all, only the way images have been taken without permission or compensation. As mentioned, the tech is here. We now have to find a way to use it ethically. Has an AI piece won an art competition? Yes, at a state fair, which is not exactly the pinnacle of art competitions. However, I have no doubt we'll see AI submitted to competitions with more reputation, leading to more discussion surrounding its use.

Now I ask, is this one any different if it was made with AI or by a human hand?

I'm not sure that the Mona Lisa and a derivative image is the best example, for a coupe of reasons. One, the Mona Lisa itself is in the public domain (though photos of it may be under copyright). No artist will have their income, potential income, or reputation affected by AI or an artist inserting a new face into the image, or making otherwise derivative works of it. Two, are these images any different? Astonishingly so, in my opinion. While the second image is beautiful, it has none of the characteristics that make Da Vinci's work a masterpiece. There is no chiaroscuro, no sfumato, the very things that made him the master he is, the things that make the Mona Lisa the masterpiece it is. To me, aside from the obvious copying of the composition with a new face and background, the second image looks more like Edward Burne Jones' style. But, his work is also in the public domain. So, no harm, no foul. Whether the second image was made by a human or AI, it's a blatant copy of Da Vinci's work, so I personally can't find the same appreciation for it as I do for the original. If a human made it, I could see the value in it as a learning exercise, but not as a final piece for their body of work.

The point I'm making is that all art is built on something else. What's the cutoff before you can claim that it's no longer an original piece and is just an amalgamation of other works? Is there a percentage of 'original' art that needs to exist? If so, does that 'original' have to be so unique that it has never existed before or since in our universe?

I don't think amalgamation isn't quite the right word. Do artists hone their skills by learning the techniques of those that have gone before? Do we do copy work as exercises? Absolutely. I'd never sell them as my own work, however, ore represent them as anything more than a study. I recently took a watercolor class from one of my favorite artists, Aria Fawn, where she generously taught her painting techniques. Did any of my work from that class look like hers? Not at all. I don't want to copy her work, I want to improve my own. I don't know a single artist who wants to make their work out of an assembly of other artists' work. Instead, we look to interpret the world around us in a unique, individual way. There's no need to do the Mona Lisa again; it's already been done. We want to make something new.

Artists aren't mad that they can't charge a set fee. They're mad that they're losing jobs, and that the work of artists in general is being devalued. They're mad that their work was used to train a program without their permission. Their mad that filters are being created using their signature style, which was the case for Hollie Mengert. But you know what? Artists deserve to get paid for their time, like any other worker. So, if it takes me an hour to do a finished piece, I'm going to charge the client for that hour. If it only takes me ten minutes to make a sketch, I'll charge accordingly. If I used an ethical AI tool to generate variations of one of my own sketches and it took five minutes, I'd only charge for five minutes.

By all means, use the tool. I understand the appeal. Let's make sure it can be used ethically.

[HannibalBarca]

The first semester I went to college, I worked for a newspaper.  This was 1988-89, and the printing press was a newer version that had a computer connected to it, so that the pages could be type corrected on the computer and fed directly into the press.  Before that, a typesetter had to arrange metal type in rows, by size and font...by hand.  Yes, just like Gutenberg did hundreds of years before.  When the computer was combined with printing presses, it was a huge leap forward in printing technology.

Now that first semester, my major was photography.  My grandfather had worked with me in his home darkroom, and I knew how to mix chemicals, develop film and prints, enlarge them, even do some tricks like burning and dodging to make double exposures and such.  During the semester, our photography teacher took us over to the computer lab, to meet a professor who was working with something cutting edge--digital cameras.  I'd heard about it, but seeing someone simply pull out a data card and stick it into a computer, then call up the images they just took--it was stunning.  Even more printing them out, even if at that point printers couldn't match photo paper pixel quality.

It was the talk the computer lab professor gave us that changed everything for me.  He told us the change was coming, where not only digital photography, but computer imaging and processing, was going to do away with thousands of jobs.  He told us how years before, he had to go to representatives of the typesetters union, who worked at newspaper and magazine companies, and tell them computers were going to do away with their jobs.  They laughed at him.  Five years later, there were no typesetter jobs left, anywhere.  He said that the same thing was going to happen to darkroom techs and photo finishing.  Computers would do all of that.

I was gobsmacked.  I'd been planning to be a photographer, but was currently employed at the newspaper as a darkroom tech.  I wasn't a good enough photographer or journalist yet to have a full-time job with one of those, and I took him seriously.  So I changed my major to desktop publishing, computer reprographics, and computer animation.  That professor wrote the curriculum for those degree tracks for California junior colleges, and I earned two of them from taking his courses.  I worked on Photoshop--back before there was a 1.0, 2.0, or so on.  Every day he'd tell about a new effect on the program he'd learned from last night's experimenting on it.  I learned the basics of video editing, multimedia...so much of it was just getting off the ground.

And I lost my job at the newspaper as digital cameras became the norm.  Not because they replaced me or fired me--because the newspaper went under, when their costs with traditional film and print and other costs were too great compared to the other town paper, that went digital, saved money, and lowered the price of their paper accordingly.  I, therefore, have first-hand experience about technology coming in a wiping out your job.  It's ironic that I was learning the new technology--but I couldn't find a job at first that would hire me to use it, because I lived in a small town, and wasn't willing to move to a bigger city, where I could get an in-demand job with an advertising company, television network, or magazine.

I'm a visual artist as well as a musician and writer.  I've done pencil/pen/graphite illustrations for decades, and am a competent portrait artist.  I've also begun using MidJourney, the AI art generator, too.  I can see the same issues here that I saw when I was a young man, and the same issue that has cropped up for millennia, when technology leaps forward.  Blacksmiths lost jobs when automobiles were invented, telegraph operators lost jobs when telephones were invented, kamishibai men lost jobs in Japan when televisions were invented--there are countless stories of history running roughshod over entire career fields when time moved on.  Humanity is relentless to itself when we develop new stuff.

Where it becomes brutal is when money becomes involved.  I don't use pictures I create for income--only for pleasure.  I already think the works of dead artists should be in the public domain, so if I make an image in the style of Frank Frazetta it doesn't cause me any anguish.  For those who are alive--I wouldn't sell anything in their style, if I do use it.  But--I wonder if simply paying money to an AI generator's creator is feeding the beast by rewarding them for copying art I don't have the skill to replicate in my own hand, especially with me being older now, and losing some of my manual dexterity.  By simply being a functioning member of society, I am already assisting in the continuation of so much of the end-stage capitalism I already despise, but life has to be lived as you can, when you're not a plutocrat or even member of the comfortably middle class.

I don't want artists to lose incomes on work they deserve to have, but at the same time, how much of it is inevitable, like what happened to me when I was younger?  And what happens when software and hardware can do the work of what 95% of us in the general population can accomplish?  Machines already run assembly lines, grow food, manage electrical grids, organize supply lines, direct traffic on roads, skyways, and railroads...will it come to the point where the only jobs necessary by humans are those of repairing the computers and machines that do all the other work?

I do know, however, that much of this is driven by the profit incentive, which boils down to someone wanting more than what others have.  We need to get a grip on letting the sociopaths and narcissists into positions of power and control, and spend less time focusing on how the little people spend their free time making images they couldn't afford paying a professional artist to draw, anyway.

[Timeless]

~snip~

I'm gonna say this, but this post really disheartened me. Kuroneko basically nailed what I wanted to say, but I do want to link a video which basically explains how the artists, including me, feel about this.

Why Artists Are Fed Up with AI Art

https://youtu.be/5Viy3Cu3DLk

[Dice]

I hate that photos and images (And as a photographer I am sure my work is in this mess somewhere) are bluntly stolen wholesale to make this shit happen.

I get that there is no going back, but how we got here was bullshit.

[Rinzler]

I would never of course suggest that hackers and trolls might find some way to flood these datasets with nazi iconography. That would be despicable. Even if the resultant shitstorm would be very amusing to behold.

In any case, it does raise an important point, I think: that these datasets may very well be subject to the whims of the people who curate them. Hardly a reassuring thought, given the present impetus of AI art.

[Chulanowa]

My opinion is simply this. AI artwork does scrape from images, but then... everyone does. How do we not learn than from observing the classics and giving our own interpretation to it? For example, this:



Now I ask, is this one any different if it was made with AI or by a human hand?

Couple key differences.

1. Leonardo da Vinci has been dead for five hundred and three years.

2. There is literally no one connected to modern media on this planet who will not immediately recognize the latter as a spoof of the original by name and creator.

"AI Art" is stealing from extant artists trying to make a living, using their own art and style against them, forcing them to compete against a theif with a print button.

[GloomCookie]

The concept of copyrights are disturbing in their own right. Mickey Mouse was created in 1928, yet the Disney corporation still owns the rights to him exclusively. The original creators are dead, have been for decades. Yet a corporation owns the concepts. The way Copyright works is that Star Wars will remain under copyright until 2072 at least.

Yet if I patent an idea, which is similar to copyrighting but for concepts that I can build and manufacture, If I patented an idea the moment I hit post, it will be available to anyone who wants to use it on December 27, 2042.

Why is one type of idea more protected than another? If da Vinci's work was still under copyright, how would you feel?

OR! Let's consider something else.



This image is 100% legal.



This image is not.

What's the difference? French law says that architecture is protected under copyright. The original design for the Eiffel tower is no longer under copyright, yet the lights were installed later and are still protected by copyright.

Copyright law is a giant mess and full of complications, contradictions, and idiocy.

I get it, protect your work for a period of time, but why is it that copyright law on works and images are so insanely long? Would it be ok if all the images an AI used datasets from were over 20 years old? Or do they all need to be public use and not covered by a copyright that is arbitrarily long? What's a reasonable use expectation if the AI is adapting and changing the work instead of just copy/pasting like thousands of people currently do? You post ANY image on the internet, it WILL be posted somewhere without your copyright info and will be distributed far and wide and you won't see anything for it unless you got paid upfront.

Try doing a reverse image search on Google for any random image you find. I doubt you'll find anything at all. Thousands of images are lifted from their source, so trying to credit every single artist is impossible. Trying to track down artists to do 'fair use' and 'attribute credit' will open up the developers to insane legal headaches, especially since AI might pull from random parts of one image, parts of another, and there's no telling where it will come from.

[Oniya]

*drags the goalposts back into the stadium*

The issue of corporate copyrights is quite a bit removed from the issue of artists who are alive, or people who post pictures of themselves/family members on Instagram, or medical records that shouldn't have been able to be scraped in the first place.

[Kuroneko]

Mickey Mouse as an image will actually go into the public domain on Jan. 1, 2024. However, the Walt Disney company will still own the trademark and copyright of the character's name, along with thousands of other symbols that are connected to or associated with the character, and those protections will last as long as the company uses Mickey Mouse as their company logo. So, should anyone decide to use the image of Mickey that goes into the public domain in a year, it can't be used in any way that implies any kind of association with Disney, or they'll face trademark violation. But as Oniya said, corporate copyrights are quite a bit removed from the issue at hand (and are far more complicated).

Copyrights exist for as long as they do because they seek to protect the author's work, not just for their lifespan, but to ensure that their heirs will receive the benefits of their work. Ownership of a copyright is a lot like owning shares in a company. Copyright terms provide some measure of 'inheritance.' This article explains it well - http://copyright.nova.edu/copyright-duration/

If Da Vinci's work was still under copyright I'd hope that his descendants were receiving at least part of the enormous amount of money that's made from the mass production of his work on everything from posters to coffee cups. I'd hope the same for Van Gogh's.

It's remarkably easy to reverse Google search an image. I did it for the Eiffel Tower images posted here and found their sources immediately (Wikimedia commons, and Upsplash respectively). I think there's some confusion about what makes an image 'legal' or not. The legality and fair use of the Wikimedia image has nothing to do with the fact that the copyright on the tower's architecture is in the public domain. It can be used freely because the image itself is is in the public domain, because the user who uploaded the image shared it under the Creative Commons Attribution-Share Alike 3.0 Unported license. Likewise, the second image isn't technically illegal because the photographer isn't selling it for commercial use. You can download it for free on Upsplash. While Paris doesn't want commercial photographers taking pics of the tower at night, they don't go after tourists. https://www.travelandleisure.com/photography/illegal-to-take-eiffel-tower-photos-at-night

It's also incredibly easy for artists to search for their own work on haveibeentrained.com to see if their work was used by Stable Diffusion.