At my current position I regularly lecture about new media and technology and AI, specifically generative AI such as LLMs and its popular usage in many areas of media creation, has become a bigger and bigger topic. Way back in 2021, I knew an early adopter who was ghoulishly enthusiastic about these pieces of software, making all the predictions now commonplace: replacement or transformation of jobs, sea change in art production, a revolution like we hadn't seen before.
5 years on, it's unclear if the ghoul was correct. If I was a betting man, I'd take my money and leave the casino because are you nuts? Having spent yesterday toying around with two LLMs to create concept art for a video game idea has made me less certain of the future than I was before. In 2021 I was strongly pro-human. As I saw it, this is our society and those are our jobs; why embrace something confusing, mostly unnecessary, and mostly nakedly economic in its interests and purposes?
I lecture on the potentials and I thought it might be time to actually use and get familiar with the tools. After all, I tell my students to do this. The world would do better if people took their own advice, or gave less of it overall. Students ask me for advice; I tell them to play around with the LLMs and figure out how and when they work well; I had to do it myself.
The game is about cemeteries, death, grievance, resentment. I had the idea before I was worried about LLMs, so it's connection is serendipitous. An environment in which humans bury their dead according to different traditions - an interesting side story to watch alongside what a lot of cynics, pessimists, and doomer-scrollers say is the end times of humanity. I enjoy a good Abrahamic-infused apocalypse fantasy as much as the next person. Watching doomsday cults get a foothold on Gen Z social media platforms is a little unsettling, but not worse I suppose than watching goths and punks in the 1990s predict the same levels of catastrophic collapse for society by 1999 (or, now that I think of it, all those more intense hippies and revolutionaries predicting the same right around 1968 or 1973, if the history books are accurate).
Instant Gratification
Running ideas through the LLMs was a matter of typing out what I wanted to see and watching the robot make something like it. It was instantly gratifying, like seeing magic. I recall the breathlessness of engineers and early adopters reflecting on what the bots could do. Their excitement was extra impactful I think because these are people who rarely show emotion, deadened by tech fetishism and poor social skills. No offense, engineers. I, too, suffer from 'white guy voice' - that late 20th century post-irony emotionless monotone drained of all telling tones and waves. I find that like other white Americans I have to inject hyperbole and the vocal versions of emoji into my sentences to convince people I have any kind of feeling at all. It took literal years to learn how to talk like I wasn't a cardboard box. So I have real sympathy for the legion of engineers who got excited for one of the dozen times in their life when they thought they had a magic box in their hands.
But...it was Pandora's box!
Joking. I 'worked' from around 10-5 pm with four programs designing elements of my game. I felt I had got fairly far in a short time, though looking at the images now I have a sneaking suspicion I understand my vision less than I did at first. This is how it goes with songwriting as well, I should say: one starts with a dreamlike melody in their head and if it takes a long time to get to an instrument, the melody is joined by an imagined rhythm, maybe a verse is written, then a chorus, something like a bridge. Visual imagery can even make its way in, and one imagines a room with people and instruments, lights of a certain color and tone that suits the whole thing - a music video, basically.
That working with LLMs should feel magical and one fall into a flow for hours, then come out of it feeling not certain of the bigger picture is, I think, not a sign that the whole thing is a scam. I'll write many, many blogs, essays, and articles with specific ideas and go after things with full-throated (full-fingered?) passion, then step away for a break and feel like I'd just sprinted underwater: got nowhere except tired.
We have to recognize that this is work, of a certain kind. I had to adjust my prompts multiple times, to varying levels of specificity. What I got in the end were images that I consider effective placeholders for when a real, human artist can come in and give them more life and cohesive vision.
Being the Producer
I felt as I hope a good producer feels (and not a bad one) - I could make broad and not deep decisions about the overall direction and tone of the game. The art LLMs make, by the way, is vanilla. These are economic machines, not things with voices and messages. The LLMs are programmed to add spilling text of commentary and suggestions alongside the products you ask for, and I turned that off on both of them. It was one thing to get images out of my head, into text, and see something somewhat similar. It was another to have the entire game mechanics and gameplay loop articulated by the bot. Those mechanics, by the by, tended to be both vanilla and over-complicated.
I am still sorting out my feelings of what it was like to work with the bot. It felt like a great deal of negotiation. Instead of playing with the voices in your head, you're playing with the voice on the screen. What it produces is only as good as what has come before, which in terms of human art and the species is not nearly good enough. But a great number of us humans are perfectly fine consuming comfortable and enjoyable mediocrity (myself included - I play plenty of uninspired games that nonetheless provide relaxation). There are many human-only companies of the past and present that made games without inspiration and with nearly purely economic interests in mind. Game producers are a large part of that. I think in making decisions about framing, outlining, paneling, and arrangement, they decide how much water to pour into money-making, and how much to pour into art-making. They get to say when enough is enough on designing a character, and when things should move on to the next project, step, stage.
The Soul is Somewhere Else
As I look at the images I got from the bot yesterday, a dissatisfaction rides alongside my motivation. I am motivated, this is true. I could drop this essay and pop back into the bot chat now to drum up more images and explore more ideas. What they say about the bots helping to brainstorm is, I think, true. It's a simple reason, too: the bot makes stuff. When I make cookies, it doesn't like anything's happened until the first batch comes out. Then we're cooking. Usually that's the most exciting time, because everyone can run in and grab one, the first fresh batch, and eat up because we all know there's more to come. It's that in-progress stage that's done as well as undone, complete as well as on-going. Very exciting stuff. The verse between the choruses of your favorite song; the time between the introduction and rising action of your favorite film; pouring your coffee or tea before the first sip. It's the flow, that wonderful flow where your actions are strictly limited, you must carry on and go through the action, and you can be conscious of being alive in your motion.
My stress over the idea that AIs would take the heart and soul out of art were not as intense as some other peoples'. Life's ups and downs have bruised my ideals a little bit; I'm slightly dead in that department. Heart broken more than once, I find it less stressful to accept that the universe is ultimately made of chaos, and the existence of purity as souls or whatever else, runs counter to the way I've seen life stolen from the living.
Yet the concept is a funny thing. It's not limited to one culture or language, though it's particular manifestations and characterizations do vary greatly. Born with it, acquire it, find it in others, find it in metaphysical spaces, find it in work...people have a lot of ideas of what it is. That it should be culturally transient makes me think humans just like the idea of a pure version of themselves, or something pure they can contact from time to time, as appealing and philosophically kind. Against the chaos of the universe, a pure soul is a refuge.
Bots are another in a long line of technologies that appear to some to threaten what makes a soul pure, specifically the perceived 'uniqueness' of human artistic voices. You've heard this conversation before: only humans can (or should) make art, not robots. If one consumes robot art, one is 'giving in' to something impure, and wrong. One is losing their soul. Is it funny that it reminds me of kids I grew up with swearing off things like alcohol and drugs because they were convinced they would lose something if they consumed intoxicants? I think it's funny but telling. Imperfect idiots, we're always at war between the ideal and the real, what we want and what we get, our personal cosmology and the chaos that awaits us.
I'll say something in support of the soul-lovers: working with the LLMs, I felt strongly and consistently that my 'soul' was someplace else as we negotiated this hobby project. What I mean is, the art the bots made satisfied but did not 'move' me. Now, perhaps cheap pixel art ghosts are not really the kind of thing to 'move' someone, and my intentions with the bots are mostly programmatic or pragmatic (I can't code) rather than artistic. If I want to draw a ghoul, I'll do it.
What I think is happening is this: the process of working with bots is so overwhelmingly tinged with economic and pragmatic satisfaction rather than artistic expression and bliss, that I do not think they have a chance to replace the process of art-making and art-consuming for humans. It's very possible human consumers can accidentally or unknowingly view AI art and enjoy it. One might argue though, that the consumer's interpretation is itself a form of art and soul-like expression; what we make of what we see still requires us to make something. But set that aside. What I really care about is the experience of the artist. Bots can churn out things that resemble all that's come before, but it's not the product that really matters so much as the speed.
While it was satisfying to see what I thought of as my vision for the game come into basic form so quickly, I'll say again I felt more like a producer arranging product and I got no satisfaction from making the thing myself. It's more like I'm putting puzzle pieces together rather than drawing something that never existed. Humans make art for the process of finding something new coming out of their hands. That's the fun part, whatever comes of it. Why do kids draw pictures? To see if they can, to see what comes out of it.
Bots rush to the finish line; they are economic programs, designed to make outcomes. If humans have souls, I'd wager they are expressed or touched or made in the moments when we work without considering too much the outcome. Souls are not, putting it simply for the neoliberals, economic agents or outcomes.
The Missing Birth of AI Art
Back in 2021, when I was exposed to the image generators and chat bots, much was said in my circles about how these things could facilitate so much. One could make their own video game, say(!). Intriguing stuff. One would still need a vision, and a feel for how a game should function, lest one wanted to make a game for the 'slop' genre at low cost and low price.
What was missing then, and what I feel is still missing now, is a discernible 'birth' of AI art. Images with extra fingers, impossible bodies, hilarious surreality and post-real physics don't constitute, to me at least, a genuine artistic convention with ideas and messages embedded in the process of making the product. Underestimating human artists is dangerous of course; art is the creation of what didn't exist before. Even now I expect very creative people are figuring out how to recreate these abnormalities of early 2020s AI art in pieces that have, through their human hand and work, messages and ideas. A simple one I've seen already: people have compared the difficulty for early AI in making human hands and the fact that humanity's earliest art pieces consisted of painting our own hands and slapping walls. Hands were apparently one of the first things humans learned to recreate in art, and one of the first things AI was really bad at (maybe it's okay now - but there's nothing interesting about perfection). It's worth mentioning here that painting hands and slapping cave walls is a process. I'd bet a large part of the fun in making that art was getting together with others and getting painted. That the product was really pretty and very human in its imperfections and chaos is the echo of the enjoyable, collective process. Even if it were one person in a cave, working for hours with different paints and different slaps - well, I'm jealous enough that I'd like to try it.
AI imagery is everywhere and maybe in time a discernible 'birth' moment will become clearer. Maybe the fantasy of AGI will come true, and a conscious AI will make art just for the sake of it. I'll end on arguments against this.
Firstly, AI is immortal in a way humans are not. Or we might take it the other way around: humans are limited in ways AI is not. Not just in our knowledge base and in our image-creation capacity, but in our bodies and our lifelines. Art is a protest against death. Time spent on art is time not spent on economics. It is time 'thrown away' to a pursuit with aspirations related to metaphysics, or the soul, or a place in history. Art is an intentional 'waste' of life as a way to celebrate life. Having no capacity to die, AI has no life, and having no life it cannot 'waste' it away. It can simulate play, art, conversation...just as I can simulate the birth, life, and death of a housefly living just one week of time. Compared to a housefly, I'm nearly immortal. My simulating its short life can be nothing more than a performance that, because I'm human, would represent art because I'm still 'wasting' time acting like a fly. (I think I'm sort of talking myself into it just to prove my point). Anyway, we are much less than houseflies compared to non-biological, functionally immortal AI software that is not embodied. Long and short: art is bleeding and AI doesn't need blood.
Secondly, art is social. Our species is social, we need one another. Art is a continual commentary on the tragicomedy of that dependence: we need but lose our parents, we fall in and out of love, we belong and are ousted from groups throughout life. The chaos of our social groupings is barely contained by the narratives or lies we tell ourselves: we are a family, we are best friends for life, we are a nation, we are a people. These social imaginings are immensely important psychological guards against the reality that we do not know if death is social; can we make friends in heaven? Or is heaven another invention, and only the void awaits? Can we make friends in the void...? AI lacks these problems. Generally, AI lacks human problems. Human problems are the source of human art. We make art not just to please ourselves, but to comment on the perishability of pleasure itself. Tech CEOs are visibly artless people; it's not their fault their intelligence in regards to aesthetics is pathetically low, and they likely don't think that much about the concept of a soul (or self-reflect even at a surface level). When we hear techies make claims about art, we should compare it to used car salesmen making claims about family bonds.
Thirdly, pain defines the human experience and AI lacks embodied forms that can feel pain. Human nature is bizarre, consciousness limited, much of what we desire would hurt us (endless pleasure, boundless life, metaphysical abilities), and much of what we need we are constantly ignoring (health, rest, equilibrium). Many debates over the existence of a soul mention "torture", as in tortured souls, misery, life being a long and uncomfortable experience of pain against pleasure. No pain for AI; simulation thereof perhaps, and I think the bots are getting as spooky as the phantasmogoria of the 1800s. We like to think there's a ghost in the machine. The illusion pleases us, draws us to wonder. AI does not make art, but it is made into art by our imaginations. This is the evolutionary trick of humanity. We keep inventing ourselves into new imaginative states.
Ending soon. Techno-solutionism has at its root a strange often overlooked assumption that human bodies and human nature are obstacles to be overcome via technology. Lots of rich people today believe they have discovered new questions that philosophy has asked for millenia: what is a soul, what is labor, what is happiness, and what impedes it? Their answers are unimaginative and unsurprising: happiness is selfish wealth and the barriers are people not buying their products. All else flows from the basic fact that these people wish to have more money. Anything they might say besides that is a deception not only aimed at the audience, but aimed at them selves. Inventors, or more accurately their bosses, have liked to imagine they are philosophically significant since time immemorial. For my money, technology serves to rotate humanity around a wheel of the same questions, values, and outcomes. The wheel does not change. Sometimes we strive for greater democracy; sometimes for greater wealth; sometimes or greater war; sometimes for greater metaphysics. Societies pick from these four choices, with a few smaller ones between, and technology flows or blocks that energy. AI is certainly a technology of wealth, decreased democracy, decreased metaphysics ("AI art kills souls"), and I'm not yet sure of it's capacity for increasing war. One would hope AI's promised economic returns will increase prosperity which would decrease the desire for war, but we'll see. Most of the wealth might flow up to the top.
AI is sold as a solution to problems people do not have, and the companies are in a hurry to cause problems so they can solve them. Erase jobs so that they can be brought back. Drive up hardware prices so that automation can expand and more manufacturing can occur. Push further into overproduction as a solution to overproduction. AI is not a revolutionary moment that disrupts a chain; it's an acceleration of a system addicted to acceleration. I'm not the first or last to say this; any coherent critique of capitalism dances to the same tune. That we so readily forget this when the new inventions are shinier, faster, and more convenient than the old ones doesn't mean we're stupid, it means technology is indeed helping us to spin the wheel we're bound to. From era to era, generation to generation, we pick from a limited number of questions and answers. Right now, the 2020s is something like the 1100s or the 1770s; magic technologies, mystical thinking, political unrest.
Now to get back to my hobby.