on generated images

and the act of drawing

In 2020 I tried to make a poetry generator. I always romanticized the idea of the computer making poetry as the closest thing to real-life mysticism I could get to. The problem when you’re trying to make a black box generator as someone nontechnical, is you still need to construct the box itself. The most plug-and-play methods of calling an OpenAI API did not exist, and the closest thing I had to it was using RunwayML to make calls to some models, which did not return the poetic results I wanted. The end result, after fiddling with Markov chains and if-then statements, was just a project that queried poems from an existing database using an image. I couldn’t access this mystical source that was a linear regression on all the poetry the universe had to offer, because it did not exist.

It was crazy how quickly the world changed two years after that. Initially, when I heard of generative artwork, I never thought it would reach the mainstream. The images were shoegazey and distorted - a haunting hyperrealistic chiaroscuro-lighted image. In a conversation with someone in 2022, they predicted that soon, after image, AI would take over sequential film. I had my doubts, generative image-making itself had not reached the mainstream in any semblance of quality or coherence.

A year later, it started appearing in ads and also tools to make ads¹. All of which made a lot of sense, in spaces not created by artists, real art would not be present. To pay an artist, is to put respect on a person and their work, and to generate would essentially be valuing the margins over the person, an opinion held by the creator.

I would argue AI-generated art is the worst kind of cheap image (not to be confused with poor images). In a self-created definition, a cheap image is something you use when you don't have the ability to make the image yourself and can't pay someone to make it for you. For example, when writing a report in middle school, you’d google to find an image to pair nicely with word art. Or when designing your first landing page you’d use some free illustration pack. However, AI generated images attempt to hide its identity as a cheap image, as if it was one that had been touched by a person.

as a tangent, there also has been a rise of “content creators” who use AI - an evolution in art thievery which at its core remains the same

It seems like generative image market has always focused on obfuscating its true intentions. After all, the UX industry standard when designing for generative image interfaces is to have permissions to use on training pre-checked. Or even removed and it’s not even in use anymore.

And in the end, artists aren’t going accept it.²

however, there have been some roadbumps due to production costs from rapid scaling https://blog.cara.app/blog/finances-and-future-of-cara

---

As someone who has been drawing since she was 9, I would say the other kind of cheapening, beyond labor, is a further reduction of art and image-making into a pure end to means of generating revenue, completely erasing the value of the practice itself in favor of the result.

https://www.itsnicethat.com/features/becoming-the-zeitgeist-creative-industry-190624

What is the value of learning how to image-make? To make images, is to get better at seeing clearly - the world at face-value visually, beneath the surface systematically, and within yourself subjectively. Intertwining all those levels of seeing, although difficult and with no guaranteed gratification, is the best way to come into definition as a person.

In the most analogous example, I like to think about illustrating, something I haven’t done in a while because it’s hard. A picture is a thousand words, an illustration is infinite decisions. What is the focal point? How can I display it in a way that’s not too obvious? How do I make sure the rest of the image leads to that point? Did I mean that stray line? How do I ease these two colors into each other? When do I stop?

But in the cheapening of images, what is the value of these details, in these decisions, why not write a prompt and be done with it?

I have literally never used image generation (left) until 5 seconds ago where i tried to recreate this illustration i made (right). objectively most genai images look nice but it doesn’t have the nuance and personal meaning imbued into it… if you look closely….

To excuse yourself from the pull of your own actions, to give up decision to something else is choosing not to actively see the world as it is. The value we forget in images isn’t for the viewer, but for the maker: a recognition that coming into being is a process we must partake in: one that only we can define.

not exactly the same as making 1 million decisions but also something about the process is just about the physical act of image-making. also I just wanted to share this bc I love flesh.png’s work and their takes.

disclaimer: this is not a critique of artists using ai as an experimental tool to make art but hopefully that was apparent. I feel like there are a lot of nuances in general and it all depends on the intention of the person and the work itself.

sidenote: I’m in SF rn if anyone wants to hang

side, sidenote: trying not to get in my head about writing but it’s hard… (pretty apparent)

1. re: adobe’s push to generative ai

2. Glaze is another good resource for protections for artists against AI generation https://glaze.cs.uchicago.edu/aboutus.html