autoartist

the case of midjourney

Let’s talk about Midjourney. 

Midjourney is an image-generator AI accessible through a Discord channel. After getting access to the Discord, users post a “prompt”; a string of words to evoke an image. The Discord bot takes those words and passes them to the AI, which generates an image, and passes it back to the Discord bot. The bot then posts the image into the channel. All users can see all prompts, as well as all of the generated images. You get 25 free images, after which you have to pay $10/month for the “basic” plan, and $30/month for the “standard” plan. 

My partner got into the Discord and had these images generated with the following prompts:

giant statue of warrior standing over ocean and village at night starry sky photorealistic

oil painted sunset mountains impressionist cityscape

I’m interested in remixing and collage, and against a lot of ideas about copyright, so I might have considered this AI a sort of fun internet pastime. What makes it more interesting—and troubling—is the almost immediate way that Midjourney users started asserting ownership of the images, and especially the prompts they used to create them. They are paranoid about losing. They hoard and hide. Now there’s in-fighting, accusations of “stolen” prompts, and vitriol against the “elitist artists” who think the AI is kind of fucked up. (An image generated by Midjourney recently won a prize at the Colorado State Fair.) 

I feel bad for the Midjourney users who want to be real artists. Skill-less, they attempt to enact the impulse that drives human beings: to build, construct, create. Our society has stripped this human need out of daily life, so you have to reorient your entire way of living in order to develop fulfilling creative practices.

But Midjourney users are not artists. That’s not just because the outputs are soulless amalgamations of other people’s work. It’s in the strange, possessive role the users assign themselves when suddenly faced with the existence of the images. Because they have not participated in the process of creating the image, the only feeling they can have about it is scarcity and defensiveness—which I would argue is caused by the lack of autonomy. All you can do is input a simple tag: /imagine, followed by a description, and wait for your turn in the queue. 

One definition of imagination is the making of something new out of lived experience. (I don’t remember where I heard that from.) What an artist gives to their work simply by showing up, day after day, year after year, can’t be broken down into little pieces and snatched out of their hands. (Except, of course, by Midjourney!) But many artists will tell you that the result of their work—the thing they made—is practically meaningless compared to the process of making it. That, of course, can't be autogenerated.

My own working definition of creativity would also exclude Midjourney: creativity has to do with autonomous engagement, outside commodification, that affects the perception of time.

We already know there is no autonomous engagement (you have to wait for the Discord bot and the AI to act), and the AI is a commodity: accessible through a monetary transaction. Perhaps the conflict in “perception of time” can be observed in how impatient one user became when communication between the bot and the AI crashed for about ten minutes:

If you require an AI-for-hire to enact your creativity, you are not working with a very useful tool. Our tools should assist us as extensions. We should be able to swap our tools when they fail or no longer suit our visions. Using Midjourney, you are relegated to the existence of the AI, and to its subscription plan. (You are also, at least at the moment, relegated to the existence of a Discord bot.) Using Midjourney doesn’t offer skills that can be transferred to other tools—even digital tools. Since I sometimes create completely digital art using a digital tool (GarageBand) there’s a good comparison to be made here. 

I’ve been using GarageBand as an intermittent hobby for more than ten years. I have only recently made music that I think is decent enough to be heard. The program has a ton of functionality that I haven’t figured out yet. Even though there is no tangible instrument, there is a learning process taking place. Over the years, my songs have gotten closer to what I imagine. They are still lacking: I need to spend more time with music—and my tool—to reach my vision. There's no doubt that learning to play an instrument would make my GarageBand music much better. If GarageBand stopped working or got too expensive, there are numerous other programs to try.

Perhaps most importantly, I can get lost while working in GarageBand. Time slows down and also speeds up. Time is a flow, not an anxious loop of anticipation and reward (typing in words; waiting for the bot to deliver your image; wondering if the image will be what you hoped; a brief hit of disappointment or satisfaction; repeat). 

Learning to draw or paint would not improve an image generated by Midjourney. It is the equivalent of typing a few adjectives and nouns and having GarageBand make a song for me. I would find it difficult to feel proud of the automatic song. I certainly wouldn’t be able to call it my own work. If I had no other skills to express myself, I might get a little defensive. 

What remains the most fascinating thing to me about the Midjourney arguments is the total misapprehension of the artistic process. Art is more akin to making food than making “content,” whatever that word is supposed to mean. Midjourney is like going to a restaurant run by robots, providing a description of what you'd like to eat to a robot server, and then waiting in line to get your meal. When you get the meal (assuming the robots aren't malfunctioning or overloaded), you call yourself the chef.

Without engagement, there is no art. Without a learning process, there is no art. The book, the painting, the poem, the song is not the art. The art is what happened on the way there. 

Thanks to my partner, Andy Maldonado, who helped a lot with this piece (understanding much more than I do about AI and Discord!).