Letter 7: The Case For and Against AI Art
On new technology and the future of artists in a web3 world
A solitary monochromatic figure stands against a speckled wall: their loose, oversized clothing draped freely over their slim figure; their face, partially obscured by the shadow cast off their wide-brimmed hat. They stand there ambivalently, their pose a juxtaposition between reclusiveness, curiosity, and a quiet confidence that envelops the viewer with a broody, mysterious aura.
An absolutely stunning piece of work that any artist would be proud of.
The catch? A human didn’t make it.
What you’re seeing above is the result of a word prompt, specifically, A Yohji Yamamoto look inspired by monochromatic shapes, which was processed through an AI program called Midjourney.
In this letter we’ll be looking at AI art, which is at the center of a rather toxic battle raging online between supporters and detractors of this new method of expression. The discussions have become especially heated as the topic frequently divulges into philosophy (What is art? Is art inherently human?), the law (Is this legal? Who owns the rights to an image?) and ethics (Is this fair to artists? Is this stealing?).
I’ve been playing around with Midjourney for a couple of weeks now, and I’ll be interspersing the article with a variety of images created using their platform as well as the writing prompt I used to create them. I’m doing this to demonstrate some of what AI art can produce these days. Note, I am a complete noob and haven’t even begun to experiment with the various advanced functions and settings you can use to more specifically manipulate images. All of my images have been from a writing prompt alone, with minimal (if any) further direction.
The Case For AI Art
Society has reached a point where AI art can be created on a whim at a quality that many would consider pleasing, requiring significant talent, and most importantly, able to stir a reaction from the viewer. While the uncanny valley still plays a part in many of the more photo realistic attempts at people (and faces in particular) there is no doubt that AI art is at a point that many people find enticing and interesting.
At the moment, the two main contenders in the space are Dall-E 2 (a product by the research and development company OpenAI, in which Elon Musk is one of the founders) and Midjourney (whose team is led by Leap Motion co-founder, David Holz).
Let’s take a few moments to consider the pros of this new artistic endeavor.
Inclusivity & Accessibility
One point that I feel has been consistently ignored and gone unspoken is how AI art can help bring about the democratization of art, most notably for those who have been unable to participate in the creation of art in any traditional sense up until then.
What do I mean by that? Well, with AI art generation, many people who may not be classically trained (or even proficient) in art or who perhaps lack the technical or mechanical ability may now partake in something which felt challenging or otherwise impossible beforehand.
This is especially promising for the millions of people who live with intellectual or physical disabilities which may limit their ability to execute their dreams to the extent that they would like. A simple, easy to use program with minimal prompts could be a great boon to those individuals and help them realize their artistic vision.
A Reference Tool for Artists and Non-Artists Alike
Conversely, this point is one which has been brought up many times already but bears worth repeating.
Artists may use prompts to help visualize concepts or styles they are considering and use it as a source of inspiration or as a point of reference. These prompts can further serve the artist as a means of refining possible directions that they may wish to explore or avoid, saving them valuable time.
Until now, many artists have used things like Pinterest for this purpose, pinning references of a variety of subjects they enjoy for potential further use. AI art takes this one step further: it’s like Pinterest, but instead of pinning the things they like, artists can now conjure them at will with surprising specificity.
It could also be used for non-artistic purposes as well. I have found Midjourney to be quite talented at item generation in particular and have used it for magical or enchanted items that our party may encounter in D&D sessions. Need a Flask of Crimson Tears and a Flask of Cerulean Tears?
We can then print those photos and use them in our campaign, eezy-peezy.
This AI art generation seems especially well suited to those who dabble with the abstract, fantasy, sci-fi, the unreal, and horror. The Lovecraftian or Junji Ito inspired nightmares which you can conjure with these generators is equal parts awesome and terrifying.
A Content Creator’s Dream
As anyone in any form of content creation can tell you, it can be a real pain to only include free/CC0 images. Look, I adore Unsplash and other free to use stock image sites, but they all have significant limitations, and many of them gate their best content behind paid tiers. Likewise, Shutterstock is well known in the industry, but it is not always affordable, especially for those on a budget.
AI art generators present an affordable alternative for content creators, ensuring that they have steady and readily available access to whatever images they need, without worry about exorbitant costs or potential copyright concerns.
Accessing the Masters
As a side note, I don’t think it has been considered enough the practical application of AI art in its ability to mimic and assess the work of masters from the past.
AI art prompted to render things in the style of certain artists—particularly artists who are no longer with us and whom we otherwise could not access—may yield clues into certain patterns and qualities, imperceptible at first perhaps even to art historians and students of the artist, about what made their work so distinct and unique. Artists could use this to gain new insights and a better understanding into what it is that makes their favorite artists stand out. This knowledge could also be useful to those who are trying to replicate or imitate the style of their favorite artists in new or unfamiliar ways.
A caveat: I don’t believe the technology is quite there yet in all instances, but we are rapidly approaching such a date. Already, in many AI art generators you can prompt them with “in the style of x”, and the results are often impressive.
There are many more positive applications of AI art that we can consider—and perhaps we shall in future letters—but in an effort to keep this from becoming a novel let’s move on to some of the arguments against AI art that you may have already seen in the space.
The Case Against AI Art
While perhaps not quite as contentious as being pro-crypto, I’ve definitely seen a large swath of the population expressing overwhelmingly negative comments about AI art online (to be fair, negative feedback tends to dominate social media either way, and that hatred may not be quite as palpable to the general population).
Just recently an article by Charlie Warzel of The Atlantic in which he was discussing the trial of Alex Jones, came under fire online as Charlie chose to use a couple images he made in Midjourney. Quickly, the internet jumped in to make baseless assumptions about the author and the platform.
Eight days later, Charlie released a followup article explaining how he went viral for his use of the AI images and tried to clarify his position as well as quell some of the rage.
However, I don’t want to paint a picture that the internet is united in hatred against this new technology. There seems to be a general permissiveness (or dismissiveness) that permeates the public about AI art, so long as people are treating it as a lark and enjoying it as a laugh on their own time. Jon Oliver even released a ten minute segment about AI art towards the end of August, gleefully poking fun at the random ambition of one Midjourney user who got high and made hundreds of surprisingly detailed prompts about Jon Oliver and his relationship with a head of cabbage (I never thought I’d write that sentence, but here we are).
Cabbage weddings aside, the fact remains that there are a vocal group of dissenters who seem poised to lash out at the use of AI images in any reputable or official capacity.
Let’s examine a few of their arguments.
While the legality of the personal use of AI images appears to be largely a non-issue, there is a lot of murky water surrounding the commercialization of said art, and who owns the images — is it the person who wrote the prompt? The company which made the AI? The countless artists which the AI studied to develop its model?
As we move forward, the legal standing of AI art remains to be decided. However, in America, one of the most pressing legal issues surrounding an AI’s ability to patent artwork has already been decided.
The case Thaler v. Vidal was the final showdown over a pair of patent applications Thaler had filed in 2019 and had been rejected, in which he listed his AI DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the inventor. After his rejection, Thaler filed a complaint in Virginia which was roundly dismissed, at which point he took his case to the Court of Appeals for the Federal Circuit. And, on August 5th, 2022, they reached a verdict. Its preface read:
This case presents the question of who, or what, can be an inventor. Specifically, we are asked to decide if an artificial intelligence (AI) software system can be listed as the inventor on a patent application. At first, it might seem that resolving this issue would involve an abstract inquiry into the nature of invention or the rights, if any, of AI systems. In fact, however, we do not need to ponder these metaphysical matters. Instead, our task begins – and ends – with consideration of the applicable definition in the relevant statute.
The United States Patent and Trademark Office (PTO) undertook the same analysis and concluded that the Patent Act defines “inventor” as limited to natural persons; that is, human beings. Accordingly, the PTO denied Stephen Thaler’s patent applications, which failed to list any human as an inventor. Thaler challenged that conclusion in the U.S. District Court for the Eastern District of Virginia, which agreed with the PTO and granted it summary judgment. We, too, conclude that the Patent Act requires an “inventor” to be a natural person and, therefore, affirm.
And in its closing, the ruling reaffirmed:
Here, Congress has determined that only a natural person can be an inventor, so AI cannot be.
And so, at least for now, the matter appears to be settled (even if the answer is essentially “inventors/patent holders are humans because we say so”). AI cannot be listed as an inventor on any patent in the US, for now. But that didn’t stop Stephen Thaler from trying again in the U.K., Germany, Australia, and the European Patent Office, all of which came to the same dismissive conclusion.
It remains to be seen how such claims will be resolved, but we'll continue to explore this subject in future letters should new developments occur.
At what point does inspiration or imitation become theft? If you ask five different people, you may end up with five different answers. With the current focus on AI art, a spotlight has been shone on this evergreen question.
When I reverse image search my Midjourney creations, I never find direct or apparent images that they were pulled from. Thankfully, I have yet to find a 1:1 artist appropriation (theft), at least in my limited, anecdotal experience. It really does appear that these programs aren’t lifting identical elements and planting them elsewhere, but rather, reconstituting elements and patterns in new situations and contexts that the program has identified from its study of a wide array of work.
However, just because I can’t find direct rip-offs doesn’t mean it doesn’t exist. If you recreate and sell the work of one person, it is most definitely an offense, but what happens when the work you create pulls from massive databases of information and countless reference images? Is an IP theft no longer an IP theft so long as you create something unique enough that no individual has a strong enough claim against you?
Allowing myself to answer my own hypothetical, I think the answer is yes. If you borrow or imitate enough elements from either an individual or a group of individuals and create a new work, wherein the new piece is unique enough to stand on its own, I think the law is generally on your side. I suspect that many of the same rules apply in the case of creating parodies. Additionally, if you’ll allow me to play devil’s advocate for a second, all artists imitate and emulate, often painfully obviously at first, then, as they refine their craft and define their own style, less and less obviously until their work becomes something unique enough to stand on its own. Where the truth lies is not for me to say.
But apart from patent issues, the subject of most interest to creators and artists alike is undoubtedly the commercialization of AI images. This is a subject which does not appear to have a resolution in the near future, and which continues to exist in a gray, undecided territory.
On one hand, many of the AI programs extend commercial rights to the creators of the work. In most cases (there are a couple small exceptions), Midjourney allows paying customers to completely own their assets, whereas free trial members are on a Creative Commons Noncommercial 4.0 Attribution International License—in plain English, that means that free trial members retain a similar amount of rights (even commercial) as the paying customers, but they need to credit Midjourney as well.
But there are a growing number of artists who claim they are owed royalties if their work was used—even in part—to train said AI systems, or if these works of imitation cause them to suffer a loss of sales. I’m sure it’s only a matter of time until we see such issues escalate to the courts.
What I previously listed as an asset of AI art — the ability for creators to generate images at will, bypassing the needs for licensing photos and artwork, could also be listed here as well, depending on your perspective. There exists a very real concern that has some artists understandably worried about losing the source of their livelihood if they are no longer needed to be commissioned. And they have a point: there are many instances where an AI generated image, even if not considered on the same level as a professional artist, would more than suffice. Blog writers, news sites, self published authors, journals, newsletters—all could potentially benefit from the use of AI art generation, putting traditional artists at serious risk of being unable to land enough work to pay their bills.
Much as many traditional artists lamented the rise of digital artists (many of whom could work faster, and take advantage of an undo button, among other things), there is a resistance right now against future technological change which would challenge the status quo. Unfortunately for those people, I do agree that a certain number of entry level jobs will be lost to this advancement of tech, but that has always been the case with changes in technology—you either find a niche of people who prefer the old way, adapt to the new way, or get left behind.
It is certainly not fair, but it’s also something which cannot be stopped.
An unrelated but significant ethical gripe of mine concerns signatures and watermarks. From what I’ve seen thus far, signatures or watermarks that are incorporated by these AI programs are seemingly jumbled and obscured. On one hand, I get it—you need to avoid copyright issues at all costs, especially when using elements of hundreds, thousands, or even millions of pieces, lest you accidentally create something new that incorporates a signature into it, thereby falsely attributing a new piece of work to an existing artist. I think this is a wise move and objectively the right one to make in circumstances like this.
That doesn’t change the fact that knowing these signatures and watermarks exist in works and are intentionally being hidden and subverted leaves a bitter taste in my mouth and causes me to question what sort of things these AI programs are, or could, be obscuring as well.
It’s Just Not There Yet
Part of the fun (and challenge) of AI art generation programs is that there is a certain degree of the unknown you have to anticipate. As such, keeping control and giving direction to the algorithm is very much a learning process, but even in the hands of seasoned pros, AI art will occasionally result in wacky, and wildly unexpected images.
While most of the time my Midjourney adventures would makes things more or less as I had expected, every once in a while it would birth something unholy and nightmare worthy.
The following image was NOT intended to be horror, by the way.
To their credit, it was a rather abstract prompt and the phrase “steal the life” was probably not the wisest choice, likely having directed the algorithm to tread a darker path than I was originally intending.
I consider this to be a part of the growing pains of new technology. With time, it can be expected that this technology will continue to improve at interpreting our true intentions; furthermore, we’ll likely learn how to more effectively use these new platforms.
Where Does the Tech Go From Here?
Some people have criticized the tech at the moment, claiming that the images being created are neither high enough quality nor at a resolution sufficient for their needs. First, I’d like to point out the obvious: Dall-E 2, Midjourney, and all these other programs are likely not as limited in the tech as it may first appear. These limitations are not limitations of tech at all, but of business: reducing file sizes and image quality not only to save money on their end, but also to keep the product more affordable and accessible for a wider audience. We will certainly see new (potentially higher paid tiers) in the future which allow for greater detail and size.
With that being said, it’s also worth noting that this is as bad as it gets. The tech and our ability to interact with it will improve exponentially over the years. In a decade, AI art will be almost unrecognizable from the humble origins it currently originates from. Given enough time, it will become all but indistinguishable from the work created by humans. I’m not going to lie, I’ve seen many pieces already which could have fooled me.
It’s almost unfathomable that much of the art produced—within a matter of seconds—is better than anything most of us can create. On a large enough timescale, AI art generation will exceed human limitations and create something that even skeptics have to admit is uniquely its own thing.
I do not believe this to be a matter of if, but when. Whatever “purely human” nuances AI currently lacks, it will one day reach a stage where it can replicate us to such an extent so as to be indistinguishable from anything we can produce.
Marrying Art & Technology
I never expected to quote George Lucas, but he’s right.
It may appear to be an oversimplification at first, but upon examination, you would be hard pressed to find many areas of art which are not predicated on tools and technology. Be it a brush, a pencil, paints, tools for cutting, setting, marking, recording, displaying, embroidering—there is very little in the art world that is not beholden to a history of technological advancement.
Before the 1800s, paint used to be either expensive to acquire, or time consuming to produce: you would need to find things in nature which provided the correct pigment, grind them down, apply oil, and fold them in until they reached the correct consistency and texture necessary. Repeat for every color required. These homemade paints also tended to spoil relatively quickly, requiring frequent fresh batches.
Think about all the art forms you enjoy. How many of them would you pursue if you had to create every single aspect from scratch? Technology eventually came along and refined the process, making the pigments more lustrous and vibrant, having them last longer, and most importantly, increasing the quantity that could be produced to such an extent that the products became more abundant, and affordable.
So yes, although I believe this technology will one day mean the loss of a certain percentage of jobs, it likely will also have a strong net positive in allowing more people to flourish in a number of new artistic fields we cannot even comprehend as of yet.
Technology democratizes art. I would make the argument that programs like Midjourney which use prompts and AI to create works of art is just another evolution in that democratization.
This is a system trained on art, having studied patterns and intricacies to such a thorough extent that no human could manage to learn the equivalent amount of information in a lifetime. But it was programmed by humans, directed by humans, and made for humans. Rather than dismissively rejecting AI art as soulless computers mindlessly engaged in the act of imitation, I would flip that thought on its head and suggest that perhaps this is actually an innovative, collaborative art form, one that marries art and technology; the creator with their tools of creation.
This is an art form that is both uniquely human AND inhuman, and that is, well, pretty damn cool.
But Is It Art?
Is any of this art, though?
With such a subjective subject, I think it’s impossible to objectively claim an answer one way or the other. What I do think is possible, however, is doing some self reflection in order to get at the heart of why our beliefs are our beliefs.
As a thought exercise, I’ve prepared a list of 15 questions about art for you to consider. Hopefully in reading and reflecting on your answers to these questions you will help to solidify your position on AI art, as well as providing some reasoning or rationale for the beliefs that you hold (whichever end of the spectrum they may fall on).
Is AI Art Art?
These are the same questions I asked myself when I was wrestling with the subject:
Does art need to be created with intention?
Does art provoke thoughts and/or emotions?
Does art conform to conventional standards, or seek to break them?
Is a piece of work not art if the artist studied or imitated the work of other artists?
Does art require a minimum level of time/effort? Can a five minute doodle on a napkin be art? Is a piece painted over a decade art? Why?
Is art inclusionary or exclusionary?
Does everyone have the ability to make art?
Who determines what is and is not art?
Is it not art if “anyone can do it”?
Is art an inherently human creation? Can apes, elephants, corvids, or other species create art?
Does art require a certain level of intellectual aptitude?
Can art be created without a purpose?
Does art need to express a message?
Can you create art by accident and/or experiment?
Are algorithmic creations (like Artblocks) art?
Having thought through your answers, how does this reflect on your views of AI art? Has your position remained the same? Begrudgingly shifted? Done a 180? I’d like to know.
As for myself, I believe my position has shifted from generally dismissive to cautiously optimistic. If this is only the beginning, I’d say we’re in for quite the journey.
As a small postscript, I just wanted to let you all know that we’re changing the format of these letters going forward. Weekly letters have been a fun, but challenging endeavor, particularly in delivering quality content on a tight schedule.
Brevity is not my strong suit, and it looks like y’all don’t want it to be either: in looking at the metrics of our previous letters, I’ve noticed a preference from the audience on deep dives, abstract thinking, and even lengthy thought pieces—all of which I initially expected to work against me, not in my favor.
In light of this, we will be releasing a new letter every other week (typically on Fridays whenever possible, but don’t hold me to that!) So yes, the bad news is that this means a reduction of four letters a month to two. On the plus side, I believe this experiment will give me the breathing room necessary to go as in depth on these subjects as I would like, while keeping the quality of content consistently high for our readers (that’s you!)
With that said, have a lovely Labor Day weekend, and I’ll see you for our next article on September 16th; I don’t know what we’ll be talking about, but I do know that it will be BEEFY! 🐄
Written by: Brad Jaeger
Director of Content @ Curious Addys (say hi on Twitter!)
🐙 If you enjoyed this newsletter and would like to be kept in the loop, feel free to subscribe! 🐙