This image was created by an AI, MidJourney. All I had to do was type in a prompt (“wildfire”) and aspect ratio. This AI is pretty good, but nowhere near the state of the art, and AI like it are, over the next few years, going to make art like this available within seconds at a cost of pennies. This applies not just to “art” like the above, which is going to accompany my prose and worldbuilding projects, but to almost every area of life where you see pictures of any kind. I think it’s hard to understate how big of a deal this will end up being, and this blog post is largely my attempt to collate a lot of the arguments under one roof, in part because some of the arguments aren’t actually arguments at all.
Artists will be put out of jobs. This is pretty much inevitable given that work which once took multiple hours will now take seconds, or maybe minutes if it’s difficult to get a good generation. I really do need to stress that the technology is in its infancy, and 95% of the obvious problems that it has now will be solved with larger models, different approaches, or better UI. If you’ve played around with Stable Diffusion or MidJourney or DALL-E 2, then you know how hard it is to get a good result for a specific idea you’ve had. I’ve been keeping up with the papers, and these problems are going to disappear. They’ve disappeared already in the current crop of non-public models, and they’re going to disappear from the public-facing models as well. Specificity is one of the key things that human artists have going for them right now, but it’s not something that’s going to continue.
The economic impacts are going to be unequal. The most impacted people in terms of actual jobs and labor will be those at the bottom of the market, people working pretty cheap commissions on Fiverr, doing mercenary artwork to specification. If you’re one of those people who supports themselves in e.g. D&D character commissions or book covers, get ready for at best a paycut. If you’re at the upper end of the art market, working graphic design, I think there’s probably still time for you, though it will change. We’ll see these models incorporated into workflows and used as tools, but being able to do a lot more with a lot less labor inevitably means that there will be less actual pay to go around. To some extent, there’s demand that’s being unserved, and that will allow people to make money, but I’m a bit skeptical. What I think is more likely is that prompt engineering and image manipulation will become go-to skills, and the artists with those skills will displace the artists whose primary skill is in the manual craft.
Lots of artists are likely to lose their jobs, have trouble finding commissions, commission rates are going to be pushed down, etc. These artists will need retraining, which they’re really unlikely to get. For some, their primary method of putting food on the table will instantly become unprofitable. This will be bad for them.
But that said, the economics are both obvious and uninteresting. It’s the other stuff that I’ve seen most artists talking about, and what I want to at least touch on here.
Will People Still Make Art?
I’m a writer. If I wasn’t being paid to write, would I still write? Yes. Demonstrably yes. I wrote when I had no audience and wrote when it was put up totally free online, and it mostly is still free online, supported largely through Patreon.
One of the big arguments I see, once we throw economics to the side, is that if people were creating art for the sake of making art, they’d still make art. Nothing will have really changed for them. People will lose their jobs, but AI art won’t stop people from picking up a pencil and doodling on their paper. Better, people will use AI art to make new things, and it’ll become one more tool in the toolbox (see below).
I’ve seen an argument that people still have horses, that new technology displaces but doesn’t murder old technology, that horses weren’t eliminated by the car. I think it’s an interesting analogy, and really wrong in a lot of ways, but feel compelled to take it apart for a bit. Horses were a big deal. At their peak, around 1915, there were something like 25 million horses in America. They were on every city street and entire industries were built around all the problems that come with horses, namely all the horse poop, feeding the horses, caring for the horses, all the horse equipment, etc. By 1960, the number of horses had declined to 2.5 million, and they weren’t welcome in many places, having been entirely replaced by automobiles, tractors, etc. They went from something average and everyday to something that was almost never seen. This was also at a time when population was increasing, and in that context, it’s maybe even more dire, because the number of horses per capita was going through a precipitous decline. So in that context, saying “horses stuck around when the automobile came” is true, but if you went up to a painter and said “hey, within your lifetime painting will see a 90% decline, stop being taught formally, disappear from daily life or awareness” I think that should shock them and be treated with some certain amount of gravity, even if painting isn’t going to be completely killed off forever. I definitely think that it’s the sort of thing that warrants a conversation of some kind.
I think the general thrust, that people will still make art, is correct in the broad strokes, but it’s missing a little something.
Why Do People Make Art?
Let’s go through a few possible answers, though I think the answers for any given artist will vary, and most will have some combination of answers, if they can pin it down at all.
First, money. Money can be exchanged for goods and services. We need money to live! See economics, above. I don’t really want to talk about it more, but yes, art is a livelihood for some people, and they’re understandably anxious about losing it, especially if they have no other skills at the ready. It’s a brutal blow that seems like it comes from nowhere, because only a couple years ago, this technology didn’t seem like it was on the horizon, and art seemed safer from AI advances than other fields.
But like I said, I made my own sort of art without any money coming toward me, nor any expectation that it ever would. Why did I do that?
Art, at its core, is an attempt at expression and evocation. It’s a way of talking about things that are hard to put into words, a way of making arguments that don’t have their own formal logic, a way of connecting with others, of sharing. Some people paint or write because it’s a way of letting something out of themselves, of working through thoughts or trauma, or spending some time feeling some particular emotion or working on a thought. I’ve read plenty of novels where it was clear the author was putting a piece of themselves out there, and I’ve written a few that have pieces of me in them.
I can see where people would say “but you can still do that if AI is doing its thing, the AI won’t stop you”, but I’m not entirely sure that’s true. In a strictly mechanical sense, yes, an artist can still create, knowing that the art could be done faster and better by AI. There are still ways in which the artist relates to the art that are impacted though, which might not be apparent at first blush.
The expression aspect of art doesn’t necessarily need an audience, but I think it does help to know that you’re putting something out there that people will relate to, something that will make them feel something. I don’t think I would write as much if I knew that what I wrote was never going to be seen by anyone, that it would languish on my computer in some file somewhere, only ever seen by me until I died, at which point the hard drive would collect dust somewhere until it was reformatted or thrown out. That, to me, would feel like howling into the void, like rolling a boulder up a hill only to have it fall back down, pissing into a sea of piss. Art is already like that sometimes. It’s a common thing to release something, to tell people to look at it, that you want them to see this thing that’s a piece of you, and to have that be met with deafening silence.
I think that in modern times, this is already a huge problem, and AI art is going to make it so much worse. I walk into a library or a bookstore sometimes and see how many of these things are on display, and I feel that same feeling, that there’s so much content you could drown in it all. It makes that howling feeling more intense, as my own contributions to the culture are rendered insignificant. When I think of AI art, and AI prose (which is probably coming, just not really here yet), I think about how I would feel if someone said “oh yeah, next year there’s going to be a million times what’s out there, done in one thousandth the time, to specifications, and it’ll all be done without any real human involved”. It’s a feeling of existential despair, as though a cornerstone of my identity and how I orient myself with the world is getting ripped out.
This is an emotional argument. I think I have to point that out, because some people see a person saying “this is how I feel” and other people want to come in to talk about economics, or engage in problem-solving, and it’s really not about that. It’s about the feeling of being without a moor, of being small and insignificant even beyond how being a member of a society of billions makes you feel small and insignificant.
The other half of the expression angle is connection with other people. Art is a communicative act. It’s a conversation. You see a picture and it makes you feel a certain way, and yes, sometimes you silently process that art, but most of my favorite aspects of art as discussing it with other people, wrestling with the art in public, teasing out what it’s trying to do, or what it’s doing without trying. I generally think that this is one of the best parts of being an author or an artist, this very public back and forth, sometimes with the art having to defend itself as the critic shadow-boxes, and I think it’s something that’s going to essentially disappear for most artists, even if they continue making art. The nature of criticism will change when the creator is an AI. I think it has to change.
The dumb form of this is “kudos”, engagement with a work on the level of numbers of likes or shares. I think this is the kind of stupidity that comes from wanting metrics on everything and not being able to create metrics for stuff that most people actually care about, and I have relatively little sympathy for someone worried that they’ll get less likes on Instagram or hearts on Twitter or whatever those websites use. I also haven’t seen anyone actually express those concerns, but given a wide world, maybe they’re out there. What concerns me, and what I think concerns a lot of artists, is the impending loss of discussion, engagement, conversation, etc.
So, would people stop making art? Not all of them. But the reasons that people make art will evaporate, at least a little, for certain artists. I think there’s something incredibly demotivating about creating art when no one will see it, or when your expression has been “done before”. I’m not trying to argue that AI art will be bad for humanity or society, but this is at least something that a lot of artists are feeling, and I think they’re right to feel that way.
Cheap, Available Art
I think it’s time to point out some good things about AI art. The first and biggest is that art will now be cheap and available. Putting aside the artists for a moment, I actually do think that this is a net win. If you can talk to a computer and get art from it, there are huge gains to be had. The floor for what it takes to create art is going to drop like a rock, and anyone with access to a computer will be able to make (or “commission” for pennies, if you prefer) decent artwork.
Insofar as I feel something from art, I think this is great. As someone who was not actually able to make art before, suddenly I can, and I can add it to the things that I’m making, especially words, to say “this kind of thing!” or “here’s some help on the visuals” or just “isn’t this thing that was in my head neat?” And I do like all this. Prose is different from artwork, and complementary to it. In my ideal world, there would be illustrations for all my work, one or two big splashy pictures per chapter in order to set scenes or punch hard at some specific moment. AI art is almost there for that. No real artist is being displaced, because I would never have had the money to actually commission artwork, nor the time or skill to make it myself.
I’m working on a big worldbuilding document right now, one with 70 different places, and for which I want 70 illustrations. To do that through conventional commissions would cost something like $7,000, which I don’t have for this project, which will be seen by maybe a hundred people, if I’m lucky. On top of that, $100 per commissioned piece is at the low end, and would represent relatively low quality artworks just because of the labor costs involved. Because of AI art, there’s now art that would never have existed. I’m genuinely thankful for this kind of thing. I genuinely do think that it’s good for society and culture. When people talk past the concerns of artists, it’s because of stuff like this, and I think the good needs to be acknowledged.
AI as Tool
The AI can make art. I’ve seen people dispute this, since what the AI makes by definition may not be art, under at least certain definitions. I think those definitions are stupid though, and will elect to ignore them.
The “AI as tool” argument is a line of thinking that the AI is basically just like a pen, or a palette, or any other tool that an artist uses to create. There are strong and weak versions of this, and I think overall some confusion about the ways in which this does and does not apply, which I’ll try to break apart here.
If the argument is that the AI is a tool and will change art no more or less than a new type of paintbrush, I think that’s incredibly false. The biggest difference is that the AI does things far beyond what any implement would do, adding in composition, content, color palettes, picture elements, and all sorts of other things. If used as a part of the creative process, it’s bringing so much more to the table than any other tool possibly could.
If the argument is that AI will be a springboard for artists, I think that’s much more defensible. A lot of artists will slot it directly into their workflow, taking the place of reference photos, stock images, and things like that. Artists already look at inspiration before starting on a project, and using the AI to generate that inspiration is a no-brainer, a way in which AI art will become a part of the artistic process without actually changing it all that much.
There’s a lot more that this technology can do though, and I think that some people who say “the AI will be a tool” don’t quite grasp the breadth of work that they’re capable of doing. They’re text-to-image models, mostly, meaning that you put in text a text prompt and get out an image, but because they have some “understanding” of what an image is, there’s a wide range of options for what they can actually end up doing, some of which has relatively little to do with text. The AI can make variations on an image, keeping semantic and compositional elements the same. It can blend the elements of two images together. It can apply the style of one picture to another. It can lift styles from any picture and use them on any other picture. It can employ in-painting as content-aware fill on steroids. It can do uncropping and changes to composition after the fact. It can do upscaling of images. It can bring images into focus by inventing details. Most of this is stuff that it can do now, and in a year or two it’ll be able to do more just by virtue of new UI, rather than entirely new models or techniques.
So I think the “AI as a tool” angle is also apt from the perspective that artists are getting ten or twenty new tools to work with. I also think that this is a relatively small slice of the impact, relative to cheap, plentiful art.
Another major concern is what impact this will have on “the culture”. This is a really nebulous thing, and I think difficult to argue about, but I think we can at least lay out some of the concerns and their rebuttals.
First and foremost, there’s a concern that AI is not actually “creative” in any meaningful sense of the word. What it does is take in a huge amount of artwork and “learn” from it, and when it outputs, it defaults to the most incredibly generic stuff you can imagine, especially in terms of composition and subject matter. It’s pretty rare of the AI to surprise you, and most of it is soulless, not because it was made by a machine without a soul, but because it’s safe and sanitized, similar to everything that has gone before, rehashing and remixing without actually adding something new.
I think there’s some truth to this. The current models work best with the things they’ve seen a million times before. They’re better at depicting Homer Simpson and Darth Vader than some unique new creation. They know the standard poses that have been endlessly copied, and won’t come up with new poses except when they screw up. They bake in biases, which is its own problem and beyond the scope of this post, but one of the things that they bias toward is stuff that’s popular in the training data, which includes a bunch of derivative stuff.
I’ve had a lot of struggles with this. I have a specific image in my head, I’m trying to prompt for it, and the AI just does not want to do it. The most trouble that I’ve had so far has been with trying to get a tavern running across the plain with chicken legs. I’ve tried hundreds of prompts and a few different models. Why the AI doesn’t want to do it is interesting: it’s seen taverns, and it knows that taverns are supposed to go on the ground. It knows what chicken legs are, and knows that they’re not supposed to be attached to the bottom of a tavern. This is probably something that will be partially solved by future models, but the ways in which the models are derivative is a key feature of how they’re trained. They are creatively constrained.
So I think if you’re an artist, or even just someone who’s concerned about “the culture”, you can look at the current models and their limitations and think to yourself, “Isn’t it going to be bad to have all this AI generated art rushing like a tidal wave to wash away all the human art? What does the culture look like when everything is a remix of a rehash of a cliche?” I think that’s a legitimate concern, but I don’t really know the answer.
One thing to consider is that the technology now is not the technology we’re going to have in a year’s time. Some of the dullness and sameness of current outputs is going to be overcome by the ability to prompt things into place, or to add vocabulary to a model, and if none of that ends up working like I think it will, we actually have an upside: there will be a niche for artists, and it will be in creating things that are fresh and new, beyond the bounds of the clichés.
Obviously I don’t know how any of this will end up shaking out. I think anyone who claims to see the future is probably full of it. But my feeling is that at least as far as the culture goes, the ways in which the models regurgitate things are probably not going to be a major impediment to the health of art as a whole, and in fact, the specific ways that the models are generic allow some better understanding of style and content, since you can have the same subject in a few different styles and see the specifics of what makes an artist or artistic movement unique.
I want people to have more empathy for the artists, and for artists to have a better understanding of the technology. I think lots of jobs are going to be lost, and it would be great if there was retraining, but I don’t think there will be.
I think the worries about art as a whole come from a place of emotion, but they’re about the future of expression, the way we connect to each other through art, the conversations we have, the ways we interpret, and on the whole, are things to at least be thinking about.
And I think that the future is probably pretty bright in at least some ways.
(Also, see the addendum on “theft”.)
(Also, if you want a short story, Eager Readers in Your Area!)