Artificial Intelligence and Art

I would like to take some time for us to consider the importance of a more robust, critical discussion about AI (artificial intelligence): specifically, its use in the arts.

I attended several events about AI during Tech Week Grand Rapids, which served to further my interest in talking about it–because part of the discussion, I believe, is still missing. In the discussions at Tech Week, panelists generally expressed positive feelings and shared some experiences with apps like ChatGPT, where a user enters a prompt and the app draws on past data to form a reply. Panelists discussed examples ranging from using AI to make a packing list for a vacation, to composing music. One attendee talked about using AI to create what he called “less critical” dialogue for a play he’s writing (or, “co-writing with past data.”)

Of course, AI isn’t monolithic, and ChatGPT is just one type. Another event at Tech Week discussed what is considered the “holy grail” of AI, known as AGI (artificial general intelligence): a theoretical type of AI with human-like cognitive abilities. Panelists remarked on its potential to one day solve complex societal issues, talking about it almost like a form of magic. While I didn’t attend every presentation about AI, I didn’t hear any discussion about possible risks or negative consequences. And that is a very important part of the discussion that is still missing.

I definitely understand the human desire to be “in” on a shiny new thing, and I know it’s historically unfashionable to do a “Satanic Panic” about something we don’t really understand. A lot of people will say, “It’s the future, so get in the car, or get run over.” But that’s the thing: we’re all in the car, it’s going really fast, there’s no map, the road is pitch black ahead of us and I’m not sure this thing has brakes. Other passengers wearing slightly better clothes are telling us we’re going to the good place. But they don’t know any better than we do.

My experience with AI and Art

I’ve never opened ChatGPT, but I did do a deep dive into Generative AI for almost a year. We had a project that needed imagery from a location I could not afford to visit. So I learned what’s called model-training, using distros including Flux, ComfyUI, Stable Diffusion, and others. I have reached a level of competency that allows me to produce realistic depictions of people and nature. Initially, when I saw these images, I felt amazed by the potential.

Within a few weeks, however, I began to feel something else: even as AI images became closer to what I needed, when I looked at them, I started to feel an emptiness. I’m not sure I can articulate it, but something incredibly important was missing. The images were correct, but they still didn’t feel “right.”

Despite this feeling, I was not ready to call it quits. I remained open to the possibility that something would change for me, when I integrated the AI images into the project, incorporating them with actual photos of real people.

It turns out, I was wrong about that. Merging the real and generated media only served to underscore what was missing in the AI images. The generated images are relentlessly devoid of anything human or natural. It’s like touching the wall in a very cold building, drawing the warmth from your body.

I know this sounds extreme, and maybe I’m wrong about all of it. I’m just starting to believe there is an essential quality to people and nature that machines can never replicate. The outputs appear convincing, but there’s an overwhelming absence: it’s not about what’s in the images, it’s about what’s not.

Ultimately, in my exploration, I decided to delete the AI images completely from the project. Back to square one, happy to have the emptiness removed, but I was still missing that key imagery. Of course, troubleshooting is at least half of any project. I tried to remember if anyone I know had visited the areas we need to depict, so I could reach out to see if they have any photos of it. I searched posts made by friends on social media, and a few results came up. I messaged to ask if I could use any photos, even just snapshots taken on their phones, of landscapes in the areas we needed. After several attempts, thankfully, I finally got some photos that would work. I was, after all, able to solve my problem without the help of empty AI imagery. Instead, I solved my problem with the help of community.

Post-mortem of my experience

This experience caused a paradigm shift in how I interact with almost anything generated by AI. Now, when I hear about writers using AI to help create dialog, or musicians using AI to help compose songs, I feel something I can only compare to grief. I dread the idea of unknowingly experiencing art with elements that may have been created by a machine. I resent the skepticism I must now apply like a reflex, to whatever I see, read or hear. It’s strange to think there was a time when we never had to consider whether a human being produced a work of art.

In my experiences trying to use it, my attempts to apply AI to solve a problem essentially cut me off from other people. It failed. I had to go back to being reliant on others. I had to connect with people to help me with something I could not do myself. Without machines to make forgeries, I am required to be aware of my own limitations. I have to seek out partnerships with others. In order to achieve results beyond my own skillset, I have to engage and collaborate with others, and these authentic, human interactions are the catalyst for endless possibilities that simply are not possible for AI that uses past data to emulate results for me.

Human connection is the common ingredient in almost everything good we’ve ever experienced as a species. It unlocks undiscovered possibilities. In stark contrast, written or visual expressions created by AI can only show us what has already been said, written or seen. It’s categorically impossible for AI to create an original thought, or image, or anything at all. It depends on creations by humans to create anything itself. For the artist in me, it’s a doom-loop of rigidly conformed expressions that can never escape from what has been. I crave human possibility. I need mistakes and happy accidents.

I want to seek out other people for their talents and their minds, to collaborate and create new things. Which is something only we humans can do, together.

Article edited by artist and writer Holly Bechiri