For whatever reason, I tasked myself with writing opinion pieces about AIs, despite the hundreds of sources already covering AI-related stories. Sources better qualified than me.
The first part of this series (HERE) covered some of the background stuff. Stuff like Copyright, Database Generation, Machine Learning, and how A.I. Art Generators work.
This post is Part 2 and covers some of the issues related to their use. This post is roughly 2,600 words or about 9 minutes of reading.
Right — here we go.
- So, what’s the problem?
- Summary (general)
- Summary (Personal)
- Conclusion (A.I. Art Generators)
- Reference Material:
So, what’s the problem?
Well, that database contains copyrighted images.
“Yes, but they were used for research and development by a non-profit.”
Yeah . . . you see, the same companies that paid to generate the database — without paying copyright holders for the use of copyrighted material — also set up a company whose programs, for example, DALL-E, use that database to generate images.
And DALL-E isn’t free; it charges users a fee for each image they generate.
So, to recap:
Set up a non-profit to “legally” steal copyrighted material ==> create a database of that material for research purposes ==> create a program that uses the database to generate images ==> charge people for using the program ==> make buckets of money.
Basically, big money-making tech companies all facilitate the stealing of intellectual property . . . but few on the scale seen in connection with the development of A.I. programs.
“OK . . . but, again, what’s the problem?”
Hmm . . . I thought I just explained it . . . OK, look at these images:
I asked Dream A.I. to render an image of a coffee cup and use this image as the source:
It can be argued these are unique images that don’t infringe on any existing artwork . . . as far as I know.
But it might. It could be that the art style derives from the style of an existing artist.
Here’s another rendering and the corresponding photo.
Still, this is a unique image that I’m reasonably sure is not similar to any current work.
And I own it. I can do whatever I want with it, including selling it. And in fact, some of the sites facilitate users/faux-artists monetizing the images generated by their text prompts.
“Reasonably sure it’s not similar to existing art?”
Well . . . OK. I’m not part of the art world. It could be those images are strongly influenced by the style of a famous artist. Still, they would be hard-pressed to claim infringement because these are, in fact, unique works.
Not being a copyright lawyer, that’s my best guess.
Suppose an artist is out there producing images of coffee cups based on Venetian masks. In that case, this might be considered an infringement, but that gets you down a rabbit hole discussion about whether I have the right to recreate a style I like and to what degree my version is unique and differentiated from the artist’s work.
If I copied the style or even an idea (Venetian masks coffee cups), it gets tricky because those aren’t — to my knowledge — things you can copyright.
If I made exact copies of their work, or even substantially similar works, that’s an issue, or maybe not (see references below). The term “substantially similar” is vague and subject to inconsistent interpretation.
So, what’s the beef of current popular — and struggling — artists?
You can and should watch a few of the videos I reference below (some have transcripts), but perhaps I can explain the issue through some examples.
Let’s start with this image of a dilapidated barn . . .
It’s unique but has a vague semblance to art I’ve previously seen. Perhaps not precisely matching a single artist, but close enough that I can say — while unable to name it — that it looks familiar. But we’re still talking style and subject, which aren’t copyrightable.
Example using artists and styles in the public domain
But what if I deliberately ask for a known style? Say, Michelangelo . . .
Well, in those two cases, I’m on reasonably safe grounds. The artist has been dead for a few years, and he certainly didn’t know about R.B. players and steampunk toilets, so I’m not copying any known works of the man. Besides, I could always deflect and say I was copying Da Vinci . . .
So, I wouldn’t be in legal trouble if I prompt the A.I. to create a bear flying a kite in the style of Michelangelo . . .
Examples of art and styles not in the public domain
But what about a bear flying a kite in the style of Dr. Seuss?
How about a boy flying a kite in the style of Dr. Seuss?
The term “Dr. Seuss”, the books, and even some of the characters are copyrighted.
I might be able to use those under fair use, or for a parody, or claim they are substantially different from anything depicted in any of the books (Note: I don’t know if any of the books have a boy or a bear flying a kite), but, based on what I’m reading, if I get sued, it would come down not just to who had the better layers, but also to how the judge would interpret what I created . . . and different judges might interpret my art differently, thus resulting in different verdicts.
It also matters if I’m posting it here or trying to make money from the renderings; for instance, selling T-shirts. Especially if I deliberately make it look like something from a Dr. Seuss story. But, again, it’s subject to interpretation.
Here’s one I’m pretty sure I would not be able to use without getting sued.
To say the Disney company is very aggressive in defending its intellectual property is an understatement. They are positively rabid about it.
In the above case, I’d be hard-pressed to argue that my rendering is substantially different, and if I tried to use it, I might get sued under Copyright and Trademark laws.
Side Note: it’s worth looking into the history of the Mickey Mouse copyright. Steamboat Willy, the original cartoon, will become public domain in 2024. There are all sorts of nuances about its use, and I’ve linked some articles about it below.
“Again, how does it relate to current artists who are not multinational corporations?”
Examples of artists being affected by AI Art Generators
Let’s say you are an artist working on commission and also offering drawing classes for drawing bipedal figures, be they hominids or monsters. You have an extensive online presence, and your images are on the Internet as examples of what you do.
Let’s say I come across your site and like your drawings.
I can buy an existing art piece, commission an art piece, or pay for your courses teaching me how to draw the human figure.
. . . or, I could ask MidJourney for a barbarian drawn specifically in your style . . .
Unfortunately, this isn’t hypothetical if your name is Steven Zapata; you just missed out on money you could have made. The money, instead, went to MidJourney.
But how did MidJourney ‘learn’ how to draw like Steven?
From a particular database that trained the A.I. program. A database containing copies of all of his drawings that appeared on the Internet. Copyrighted illustrations gathered under the auspices of research by a non-profit company.
Mind you; I could have used a different prompt but not gotten the same look if I asked for Michelangelo’s version. I noticed that “god the father” from the Sistine Chapel is typically depicted as a brawny man:
. . . so I could ask MidJourney to draw me a barbarian in the style of Michelangelo’s Sistine Chapel . . .
Notice the difference in style and visual impact of the final product.
A large body of art is considered public domain and could have been used to train A.I.s without stealing from contemporary artists but I don’t think it would have been as popular with modern faux artists. In fact, a large portion of the art that’s currently generated by A.I. Art Generators is based on art from modern games and other online graphic content.
It’s difficult to imagine that the people building the database gathered everything “for the greater good”. It’s much easier to imagine they did it to make lots of money.
But there’s even more to it than that.
For instance, take the case of Sarah Andersen. Her cartoons were first “borrowed” by Far Right mouth breathers who replaced her words for their use (promoting their views). It was then duplicated using A.I.s.
Someone even copied her font style (handwriting), making the fake illustrations more realistic.
It’s difficult to see a “substantial difference” between her original work and fakes. This has a double impact because not only can people represent her work for their aims, but what they’re publishing has the potential to harm her reputation and brand.
And that’s where the big argument lands . . . the people designing these tools absolve themselves of any responsibility related to their use. The users of these tools claim ownership by virtue of having created art that is arguably unique. Everyone wins . . . except the artists.
Annoying Side Note: there are many users of these tools that hide their prompts. In doing so, they essentially admit they don’t want others to duplicate their work, even as they claim they have a right to reproduce other people’s work.
Let me point something out: nearly every blog I read, every photo I see, and all the content I come across have an associated copyright claim and verbiage saying the material cannot be used without permission.
So, to my readers . . . how do you feel about knowing that everything you post is likely in a database used to train A.I.s, which are then used to generate content that directly competes with what you do?
I just read that artists have filed a “Class Action Filed Against Stability AI, Midjourney, and DeviantArt for DMCA Violations, Right of Publicity Violations, Unlawful Competition, Breach of TOS” was filed in the Northern District of California.
I wish them luck, despite the massive money that’s lined up against them.
But I also should mention that artists are not monolithic. Some artists already employ these tools in their work and are excited about their potential even if they recognize the associated problems. Circumstances matter, and it’s worth learning more about the issue by watching some of the videos below.
I previously stated that I saw the potential for these tools being used by artists as more significant than the potential for people with no artistic skill to replace artists.
I’m less sure of that after reading and seeing what some users are doing.
At the very least, these tools will devalue art.
Well, look what happened in the photography field. The advent of Digital saw a massive proliferation of photographs by people who basically just learned how to press a button and have an app automatically beautify their photos.
Digital also drove what the public wanted from photographs, and suddenly, photographers weren’t just competing with other photographers but with anyone with a camera or phone.
Other than for specialty stuff, professional photographers found it very difficult to make a living at it unless they were fortunate enough to build a brand or get exposure. It was no longer enough to have a great photo; you also needed to be recognizable.
For example, people have told me I should try selling my photos, but I can confidently tell you that I could take my best photographs, frame them, and offer them for sale, and I won’t get any takers.
“How can you be sure?”
Well, I did that, didn’t I? When we moved from Colorado to Hawaiʻi, we put almost everything we had up for sale, including favorite images hanging in my house in nice frames. Unfortunately, I ended up giving most of them away because no one offered to buy them.
“How much were you selling them for?”
Anywhere between $2 and $10, with most in the $3-$5 range. Heck, the frames were worth more than that!
I suspect that if I had some name recognition (local, regional, or national), those same photos would have sold, and probably for much more.
True, that’s always been the case with artists, but now artists — just like professional photographers after the advent of digital — are competing with a flood of non-artists who didn’t have to learn anything about the process of making art.
All these ‘faux artists’ had to learn was how to ask the A.I., “Hey! Make me something I can sell!”
That’s why I say these tools and the proliferation of AI Art will devalue the product.
To be sure, a few of these faux artists currently make some money from their renderings, but I think it’s because it’s early in the process. It’s early, and even so, millions of fantastic images are floating around.
I don’t see selling A.I. art as a sustainable business precisely because anyone can do it and because when you do something, even if it involves only writing a few words, there’s an immediate attachment to it.
What you create might not be the greatest out there, but by golly, you did it! It’s yours! Why would you pay for something someone else did, especially knowing they just asked the A.I. to generate the image, and you can do the same?
Side note: Sometimes, I see people using their phones to take a photo, and their photos are not that great. I then offer to take the photo for them, and you know what I hear? “No, that’s OK, this is fine.” I explain it would literally only take a few seconds. “No, that’s OK.” It’s their photo, you see; they will value it more if they take it than if someone else did.
I definitely see A.I. art as having the potential to hurt actual artists, especially when their styles and images are easily duplicated.
I’ve already read stories about companies bragging about saving money by using these tools for presentations, ad campaigns, and concept work. They seem happy not to have to deal with an artist because the A.I. is faster and cheaper.
Heck, now they don’t even need a knowledgeable employee . . . they just need someone who can type.
So, where do I stand on this? Well, I can’t deny liking the products.
I’ve already stated I don’t see them as anything more than something to play with, especially since it’s rare that I get something exactly like what I visualized. Part of the fun is the surprise of seeing what you get.
But, truthfully, after the initial deep dive into them, I’m already bored because there are only so many steampunk toilets one can render.
My intent was never to generate art and pass it off as mine. It’s just something I wanted — and still want — to explore. I’m profoundly aware that I do nothing in the generation of A.I. art.
In fact, I don’t even have to come up with a prompt since there are already prompt generators that will give me random prompts to use.
OK, I typed a few words, but those images, as interesting as they might be, aren’t “mine”.
I don’t have the same connection to them as I have with my photos and writing because I had little control over the finished product. It might be interesting, but it’s equivalent to buying a mystery box and finding something interesting inside.
Note: many users of these AI Art Generators appear to get personal validation from them as if they had generated the art themselves instead of asking the A.I. to do it.
I still plan to play with AI Art Generators, but one significant change in my interaction is that I won’t invoke any of the hundreds of artists these tools purport to mimic unless their works are in the public domain.
It’s a small step, and it could be argued that I’m still ‘supporting’ A.I. Companies who are no better than art thieves. I don’t have a counter-argument because I don’t know which data the A.I. program references in rendering my prompts.
Conclusion (A.I. Art Generators)
I can see the high argument from the side of the A.I. developers (see below). Still, the argument is massively undercut by them purposefully and deliberately setting things up in such a way as to shield themselves from criticism and liability while cashing in on the product they developed.
It’s dishonest and predatory, and no high discourse about the benefit to humanity makes up for it.
. . but I also don’t see this going away.
This research and development are, in effect, international competitions and probably have national security implications. The current use may be banal (for example, creating steampunk toilet paper), but what artists are now experiencing will likely expand to affect all industries.
Also working against the artists is that this is a shiny new toy given to a public hungry for any distraction. Add to that the pseudo-validation ‘regular people’ get from ‘creating art,’ and I fear artists’ rights will get trampled. In fact, there’s already a vociferous segment of A.I. Art Generators users defending ‘their right’ to generate whatever they want.
The best I hope for is compensation for the artists by the large corporations who stole their images and better rules going forward. But the companies involved have very deep pockets for lawyers, political contributions, and armies of lobbyists.
It might be tempting to call it a Quixotic quest, but there’s nothing foolish or impractical about fighting for recognition and compensation for one’s efforts. It’s practically the founding myth of the American dream . . . and it would be nice if we could put to rest the disturbing suspicion that it is nothing more than a myth.
Why Artists are Fed Up with AI Art (13 minutes)
The War between Artists and AI (42 minutes)
The End of Art: An Argument Against Image AIs (48 minutes)
Transcript of The End of Art: An Argument Against Image AIs
What AI Developers Want Artists to Know about AI (1 hr 22 min)
AI Ethics, Artists, and What You Can Do About It (1 hr 40 min)
Class Action Filed (1 hr 16 min — content begins at the 10 min mark)
Articles (some may be behind paywalls):
The Alt-Right Manipulated My Comic. Then A.I. Claimed It.
A.I.-Generated Art Is Already Transforming Creative Work.
You Can’t Copyright Style
Art Copyright, Explained
Microsoft brings DALL-E 2 to the masses with Designer and Image Creator
A startup is charging $1.99 for strings of text to feed to DALL-E 2
Dr. Seuss Copyright
In 2024, Mickey Mouse Will Finally Enter the Public Domain — Sort of
The Mickey Mouse Copyright – THE ULTIMATE GUIDE
What’s the Difference Between Copyright and Trademark?
Copyright or Trademark – What’s the Difference?
The next segment will be about ChatGPT . . . the AI Writing Generator. Stay tuned.
That’s it. This post has ended . . . except for the stuff below.
Note: if you are not reading this blog post at DisperserTracks.com, know that it’s copied without permission and likely is being used by someone with nefarious intentions, like attracting you to a malware-infested website. Could be they also torture small mammals.
Note 2: it’s perfectly OK to share a link that points back here.
If you’re new to this blog, it might be a good idea to read the FAQ page. If you’re considering subscribing to this blog, it’s definitely a good idea to read both the About page and the FAQ page.