Being an artist
This project developed, unpublished, over the emergence and then slight deflation of the first mainstream bubble of generative AI. One of the ideas that I feel has lost its edge in the public discussion of these tools since I started working on this is the notion that people will use image generators for the same reasons artists make pictures, and that if you do you have become an artist. Since this project is rooted in skepticism, I’m not troubled seeing that assumption decay a little. In 2024, we seem more comfortable with the idea that the most dedicated users of image generators are more interested in having images than authoring them, and that it’s likely they want to have them so they can use them as some part of a commercial enterprise. The assertion that these are tools of creativity for people wanting to be creative is still around, and accurate in many circumstances, but I don’t think we assume that anymore, which feels different than late 2022 or even 2023. If anything, I think that assertion is read as a form of marketing, beneficial to those who have an interest in their platforms being used.
However, comparisons to painters or “what painters do” are still flying around the discourse, whether to support or critique the “artfulness” of image generators. Publicly criticize the use of preexisting, copyrighted art by image-generating LLMs and you’re almost guaranteed to be met with arguments that “artists do the same thing” or maybe a catchphrase like “everything is a sample” or “good artists copy, great artists steal.” The terms we use in these conversations – art, artist, paintings – are often reduced to tokens, just like the I in AI. We use these terms like known quantities whose value is determined by a market in which they are exchanged. These are terms whose definition is anything but static, constantly debated by artists who make art and painters who make paintings. So, as one of those, I wanted to help break the framing of these conversations a little.
I’m very glad to feel the amount of credulity around these discussions fading. We’ve lived with image generators for a while now. I started this in response to working with generators based in Google Colab notebooks running VQGAN + CLIP, before the launch of Midjourney, Dall-E, or ChatGPT, and the discourse was already thick on the ground then. It may be difficult to tell whether some kinds of images are made by humans or machines, and we may still be getting used to the idea that anyone can acquire certain kinds of images whenever they want, but it doesn’t seem like we are ready to call anyone with a paid ChatGPT account an artist automatically.
At the same time, these arguments are still very active, and the motivations to ascribe artistry, creativity, or intelligence to the functions and products of image generators are still very present and evolving. To get anywhere new, they need to break open these terms – art, artist, painting – and welcome how complex they really are into this conversation. “Intelligence” too, but I’ll stick to just bumping into that one by talking about art.
This section looks at “artist” first. It seems like our debates have been more focused about what we should think about making instant art than making instant artists.