Back at the tail end of 2021, I remember stumbling upon a random image on Twitter. It was clearly something procedurally generated, a vague resemblance of an object, blurred shapes that roughly evoked a prompt that acted as the piece’s title. It was intriguing to look at, and as such when I discovered that it was generated from a mobile application called Wombo.AI, I couldn’t help but download it and play with the new toy.
At the time, I really thought it was just going to be a toy, which given historic trends and the eventual end result was perhaps foolish. This magic black box of an application was simple enough in nature. You’d give a series of words, select a style, and the computer would comb through its database and assemble a hasty pastiche. It was never particularly distinguishable, but if you asked for a castle, you’d get something like a tall building. Animals would be a distorted show. It was an interesting thought experiment more than anything, a place to give an idea a prototype form. Unfortunately, Wombo quickly announced that it would be using its proprietary to mass generate Non-Fungible Tokens or NFTs, and quickly the interest waned. It was one thing to play with a toy, it was another to willingly participate and enable volatile cryptocurrency practices that prioritize mass generation of images and pretending that they were immune from copy-paste tactics.
Early in 2022, Midjourney entered the scene, and Midjourney was remarkably more powerful than anything Wombo had done. Detailed images could be generated from oddly specific prompts. Multiple options with the ability to select and refine pieces. The computing power alone must have been immense, but Midjourney offered a taste at this fascinating piece of technology. It made perfect sense that this type of data-intensive aggregation took up processing power, but at least for me, if I were to pay for art, I was always going to pay an artist directly.
And then everything sort of came crashing down at once. Before long, the various twitter circles I inhabited began pointing out grievance after grievance. The two that constantly stuck out the most (and also deeply entwined) were the use of these generators to bypass paying actual artists and the blatant theft of different artists’ work. There was a whole book generated from AI and at least one book cover.
For those unfamiliar, AI Art is not spontaneously generated. These programs are trained on gargantuan large data sets and through user-influenced iteration, they learn how to associate certain phrases with certain images. Like with many models, these only exist because of the person feeding and extracting information.
This is not an emergent problem unique to the 2020s. Concerns of automation, artificial intelligence, and machine learning have been around since the start of the industrial revolution. It has just taken to the forefront of the conversation because the technology has advanced significantly and the constant need for engagement with social media platforms makes it criminally easy to aggregate and disseminate. To call AI Art “art” or just another tool in a toolbox is a little disingenuous. The very foundation of the technology is theft. The very process the algorithm goes through is not a creative endeavor. It’s a convoluted series of binary choices that makes for an uncanny simulation.
The original promise of automation was that the work we as humans were fundamentally bad at would no longer need to be done by humans: strenuous physical labor, intensive calculations, repetitive iterative protocols. Automation was supposed to free up time for humans to do the things that machines and computers were bad at: art, comedy, and writing. The principles of artificial intelligence and machine learning get conflated constantly, and in our pursuit to streamline processes, we lose sight of that sometimes that the act of going through the process is the entire especially when it comes to art.
I’m a well-documented robot sympathizer, and there was a time that I would side with the robotic overlords; however, the current trajectory of the technology is simply not heading out. The software is a pale facsimile that is nothing more than a poor speaker box and project.
As such, I felt compelled to compile several different stories that demonstrate how the actual artistic process compares to the endless dirge of pastiches. There are some spoilers in the next few paragraphs so be wary past the titles.
The first that came to mind was Diva/Vivy from Vivy Eyes Flourite Song. A singing robot who is an actually designed music player that eventually gains sentience over the course of a hundred some years. And even with all of the fancy programming and technological advancements, it still takes one hundred some years and harrowing experiences for Vivy to produce a single original song. And in what can only be described as a little too prescient, said song gets co-opted briefly for a robotic uprising that nearly upends humanity before Vivy willingly resets herself to course correct because of the *personal* connections she made. The deliberate decision she chose.
Then I thought about Zima Blue from Netflix’s Love Death Robots, another story about a robot (this time a pool cleaner) that takes hundreds of years to eventually develop masterful art based on a pool tile it used to clean back when it was a primitive machine. Even in fiction, true art takes so much of a larger time scale that we often acknowledge and the use of AI to somehow accelerate the past this ignores the very fact that this is not something that can simply be automated. Art is a complex process and creation is something we should trifle with so haphazardly (something Hayao Miyazaki has touched on in the past).
Of course, Carole and Tuesday deserves several mentions. In a world where almost all music is procedurally generated by advanced algorithms, the endurance of the human songwriters is still able to break through and spark revolution specifically because it comes from a place that computers will never be able to parse. One of the many truisms I have learned is that the goal is always to create something only you can create. Whether it’s Carol and Tuesday getting inspired by the circular motion of the laundromat or Ezekiel rebelling against the oppressive governmental systems, there is substantially more passion and authority than the songs that merely follow standard cookie cutter formulas, a sentiment that is reflected in our society.
Hell, all of this even got me thinking about Gearless Joe from Megalobox. Even with all of his opponents having biotic enhancements, Gearless Joe is able to stand in the arena and fight back on equal terms. There’s a factor of human grit and determination that is capable of closing the gap. As much as enhanced biometrics are capable of doing, at its core, they only augment, not fully replace.
I really wish AI generated tools were more toys or starting points. I wish that there were a means to start giving form to a shapeless figment that could be refined through human hands. But that’s not what “AI art” is currently being used for. It’s being used to churn out thousands of meaningless images for use of NFTs. It’s being used as a means to steal from artists two times over, first by literally stealing their stylistic choices and second by not paying to create art. It’s attempting to streamline the act of creation in a discipline where the entire part is learning the intricacies of creation. It’s not that the tools don’t have a place, it’s that the current application of the tools is actively detrimental and harmful to actual artists.
Computers are great at taxes, optimizing supply chains, doing number crunches for rockets. Computers can discrete problems with discrete conclusions. But art, whether drawings, painting, writing, filmmaking, is not a discrete problem with discrete conclusions. It’s an infinite canvas of possibilities and if there’s one thing we also know from watching the media it’s that computers struggle with the infinite.