A.I. vs Them

In the summer and autumn of 2023, I held an art “competition” between myself AI art generators: AI vs Them. It is “Them” because they/them/theirs are my pronouns. In about two hours, I would be given a prompt and then generate an 18 x 24 inch acrylic painting which would then be compared to the images produced by AI using the same prompt at WNDR Museum in Seattle. The museum has an in-house AI art generator, “Untitled By You” by wolfbear, which uses prompts to determine the images which are generated and then displayed onto five monitors which are approximately 20 x 30 inches large and take about 90 seconds to generate using stable diffusion.

AI artwork has been a hot topic for the past few years as the technology has become faster and more impressive. Some even speculate that brick-and-mortar artists may be replaced by their software counterparts. I strongly disagree with that sentiment however. The role of the artist is not merely to make pictures or sculptures. Artists are beings who bring meaning to the mundane, expose the truths from within and outside of themselves, and each have a unique vision to share, because no two people are alike. And while so unique, humanity loves to reflect on itself, finding our common ground and relating.

Not that AI does not allow humans to reflect on themselves. Humans can find the humanity in a non-living river stone with appliqué googly-eyes, and call it their pet rock. Like humans, AI learns and adapts through experience. Given the same prompt multiple times, it continues to produce new images.

I recognize the validity of using technology to generate art. It can make the creation of art more accessible to many people. I have used automated textures, digital brushes and transformations, and copied and pasted reiterations in parts of my artwork. AI and digital art workspaces can assist in imagining the unimaginable in a crisp, exact style, and that is very exciting.

With the rise of AI, controversies have arisen, such as issues over copyrighted material being used for AI. For example, viewing the displays from wolfbear’s project, it is obvious that the images are composed of other works, and blurred out signatures belie this in many corners of the artwork. It is not so much an all-original creation as it is a collage in many cases. Collages are valid art, so it does not entirely discredit it but it is not as though the AI is fabricating all-new images. Secondly, an artist has free will to create and think their own thoughts, while AI must be awakened and prompted, even if the prompt is to do whatever it will, it is only with permission, and yet, AI does inadvertently produce unwanted images just like human artists do.

In designing the AI vs Them challenge, I compared my artistry to AI. AI is definitely faster at creating a full-color image. In less than two minutes, this specific AI creates five images, while I need about two hours to make one image of comparable size in full color which satisfies. Art can be born rather slowly, and many techniques, such as washes, have lengthy time constraints, as applying layers too soon can lead to the lifting and removal of the first layer–-though intentionally lifting paint is also a technique I have used. Digital layers are a convenient tool which make digital artforms more efficient than the slow-drying layers of physical paint. Digital production allows liquid-like blending or strict line-abiding with perfect control, simply at will.

In any way in which AI lacks, it can be reprogrammed or directed to imitate a specific style, without the obstinance of an artist who may never change style again (though perhaps rightly so). Although the interpretation of prompts is not always what the viewer had in mind (note the AI generated cats that lack of pincers or duplicate eyes in the black cat round),it is relatively sophisticated at interpreting human language, and translating those words to images, but could, and probably will, improve. It is easy to imagine the frustration of being a new artist in today’s world, comparing oneself to a computer program a fraction of their age, and feel bested, with efficacy far out of sight. Keep trying, new artists! Your vision has value.

A competition between AI and a human draws many of the same concerns that contests between human artists casts. How can you stand two works side-by-side and declare one “better”? The curator faces difficult decisions! With specific criteria, or with popularity, art is subjective. The inhuman speed, precision, and strength of machines give them enormous potential in the realms of objective measure: reliability, how much weight can be lifted and moved for how long, accuracy, and iteration. Although there is certainly more beloved, complex, or universally meaningful artwork out there, each individual person has their own preferences. The perception of visible light itself can vary enormously from eye to eye.

Predictability is an aspect that is also up for debate in the battle for art. The human artist may interpret very differently than expected–after all, there is much to the mind which is unknown and the artist may reveal it or even conceive of it while creating. For example, one AI vs Them prompt was “a penguin coming out of a ceramic teapot”. Being transgender and queer myself, the phrase “coming out” jumped out to me, as in “they came out as non-binary".” While the literal interpretation was included, I added an enthusiastic-looking “I’m gay!” inside of a speech bubble for an additional facet of the prompt. None of the AI-generated images included this, and it is not known whether the prompt write Sean D. intended to reference “coming out” in that connotation or not.

Similarly, “pink 90’s crocodile” does not mention glasses, but I added a pair of narrow glasses with artistic license to provide more of a 90’s look, since that style was popular in the 90’s. There is value in the artist’s interpretation and novelty, and yet, AI artwork can share the same charm.

The knowledge that I bring compared to AI is an interesting aspect of comparison as well. While possessing the advantage of lived experience and the knowledge of feeling and other non-visual sensations, the human artist only has lived experience or what they are capable of finding. Even for these challenges, I needed to peek at some reference images. Stable diffusion calls upon a great plethora of reference images, from celebrities to landscapes to scientific illustrations. It is like an encyclopedia and artist in one. The overlaps and exclusions of the knowledge of AI vs human can form a Venn Diagram: AI does not know the secrets of the human artist which they have in their toolbox, nor can the human artist paint a celebrity they have never seen.

In conclusion, AI vs Them, though containing “vs” in the name, is more of an experiment than a competition. Declaring that either of us “won” would be rather subjective, and it is not exactly a fair competition either way. Perhaps in the future, AI can challenge me in acrylic painting rather than digitally generated images. Both human and machine artists bring joy and meaning to others, and AI and commissions allow people to have their visions come to fruition without having to learn to create art with efficacy, which can be a slow process. Both are valuable as collaboration, bringing something new to the world that the artist nor commissioner would create alone. The instant gratification of AI may be irresistible, but human-made art, be it digital or physical, has a special charm. Many, including me, believe that part of the human spirit transfers onto art, but I also find that AI is capable of spirited images under curation.

Note: This blog post has been edited from its original text for grammar.