An artistic revolution is taking place before our eyes. Recent advances in AI-generated images, text, and videos are stunning, and improve every day. These digital images are no ClipArt or Etch-a-Sketches—they are convincingly original pieces, composed by software that has become increasingly effective at mimicking human styles. I have neither the expertise nor the authority to explain exactly how this code works, but I encourage my fellow humanities nerds to understand that this technology will have consequences for our fields, even if we do not understand it. There are three key threads.
First, AI systems have been coded to mimic—to mime, re-contextualize, remix, and reframe the billions of pieces of artistic talent that have been fed to them. These are data machines, immense digital black holes, programmed not to directly produce great art, but instead to approach desirable images through trial and error, with the aid of immense data sets. I’ve heard the process described as “making marionettes of us all,” and that’s not far off the mark; this software can effortlessly mimic styles, adapting and extending the creative work of an individual. Dead or alive, with permission or not, anyone whose work is extensively available in digital form runs the risk of being mimicked at a rapid clip. Second, this technology is moving very quickly. It was only in July of this year that the code really began to bear fruit, and technology tends to follow an exponential curve. All indications suggest that we are only beginning to rocket up that incline, and that the ability to mimic written or visual style flawlessly is quite possibly in our near future. Third, this technology is cheap. Anyone can sign up for one of the best generators, DALL-E 2, and produce hundreds of images for free in an afternoon. While it takes a bit of practice to get the prompt one provides to get the wordings right, it is hardly difficult to generate a suitable image for, say, a Haverford Poetry Club poster in record time. A watercolor painting that might have taken an artist hours to finish is now automated effortlessly—and this has huge implications for the artistic economy. If you’re a business, and you need a quick image of, say, “a close-up shot of an astronaut playing poker in a casino on mars, neon lights, detailed, canon ef 105 mm f/2.8,” what makes more sense for your bottom line: days spent commissioning a costly freelance artist, or minutes spent using DALL-E 2 at no cost?
I care about this subject because I genuinely believe it to be one of the most important developments in the history of artistic creation. I don’t write those words lightly. But authors and artists must take it seriously. Traditionally, knowledge workers have held themselves aloof from the forces of automation. “There’s no way I’ll lose my job in graphic design before my lawn worker or housekeeper is automated out of existence,” they say. Yet here we are: creativity is officially on the chopping block; the ability to generate compelling artificial art at a mass scale is coming. So. What to do? How can we approach these images as artists, critics, and interpreters?
It is in part such a daunting challenge because so few of our well-worn theoretical tools can account for this new reality. While artistic theory has gone back and forth over the role of interpretation and authorial intent for decades, what do you do when there is no “author” at all—merely an inanimate string of code, fine tuned to mimic authorial style? Undoubtedly these pieces will grow good enough to cause genuine emotional reactions in human beings (if they don’t already); to what extent are those emotions less valuable when triggered by a ghost in the machine? If someone feels a powerful response to elements of a piece which were not placed there by conscious design, but semi-random fluctuation, is there still value in “reading” the piece for meaning? Say, for example, that an AI program creates a heart-wrenching image of a mother grieving her child. What are we to make of the particular slant of her eyes, the way the shadows fall on her face, the tones and tinctures of the paint? These are but variations in code. No artist consciously imbued the piece with vision, emotion, or soul; it is empty of us. How can we ascribe value to any sort of interpretation of meaning when we know definitively that there is none—that the works we would analyze are imbued with about as much “meaning” as the patterns of a Roomba?
One solution out of this conundrum is to eviscerate the role of the creator entirely, and argue that the only important meaning in art lies with human emotional response. Much as a canyon—with all its arbitrary randomness—is considered beautiful, perhaps these paintings and pieces can gain meaning purely from the responses that they invoke in viewers. But such an interpretation leaves little room for the value of human creators in an increasingly automated future. Unless we consciously assign particular worth to art that has been made by our fellow people, its creation runs the risk of going the way of the cobbler; an interesting hobby, perhaps pleasant for a weekend, but a wildly inefficient use of one’s time. How often do you buy shoes from a traditional cobbler, an artisan whose craft was once genuinely valuable, but is, today inordinately impractical?
Perhaps my greatest concern is that our natural affinity for creation will be damaged by AI technology. I write poetry not because I feel that I’m the best of the best, but simply because the act of artistic creation gives me purpose and satisfaction. But can this rewarding feeling survive a world where the art we generate for our own contentment is twice as bad as what we could do in seconds online? What about ten times as bad? Would I gain as much love from poetic expression if my style could be effortlessly ripped off and improved upon in the digital sphere? We have so little precedent for this in previous histories of automation; traditionally, we worked to automate the tedious and time consuming tasks of our lives, not the enjoyable ones. Chopping down the occasional tree may be nice on a brisk morning, but the chainsaw ultimately saves more pain than pleasure; I struggle to see how AI art will do the same.
A final element of AI art that seems relevant is an acknowledgment of its disposability. There is something in the mass production of original images that cheapens creativity. images that could represent the heart and soul of a week’s labor for a conventional artist will be deleted in seconds in favor of a slightly better alternative. We live in a disposable world; objects like clothing or dishes, which were once painstakingly repaired, are now tossed every couple of years. I remain doubtful that extending this sort of constant churning to art is a good thing.
To be in uncharted waters is both electrifying and terrifying. I struggle to think of the last time that such a fundamental shift in the way we experience human creation swept over the world.he closest I can muster is the spread of the printing press, a process that took decades. Now, we must contend with a greater change in a fraction of the time. There is no definitive philosophical guide to draw upon, no Barthes or Foucault of the AI revolution. Instead, it is one of those rare intellectual concerns where we are in entirely uncharted territory, where we claw at the gaps and fringes of our understanding. I propose these questions not to develop a definitive framework, but to begin by taking the first tentative step toward acknowledging that we have a fundamental challenge on our hands, We must confront as both artists and thinkers, and I cannot travel uncharted these waters by myself—so perhaps you will come with me.
