The lowdown: The Colorado State Fair recently hosted a digital arts competition. The first-place winner, Jason Allen, used an artificial intelligence tool called Midjourney. (You may have seen author Chuck Wendig’s ponderings on Midjourney already. If not, here ya go.)
Midjourney is an AI visual art creation “lab.” You go in, type in a collection of words like rabid rabbits under a full moon, and the AI spits back visual renderings. It’s a fun process. The renderings are gaining in detail and variety as more and more people teach the artificial intelligence what’s what.
Allen used Midjourney, entering, revising, and recombining a series of words and generated hundreds of images. His efforts created his series “Théâtre D’opéra Spatial.” It consists of a trio of pieces displaying an operatic sensibility within dramatic alien landscapes.
It’s important to note that Allen states he did not just cut and paste the AI creation and enter it into the contest. He reportedly adjusted the generated content using Photoshop.
He entered the contest and won first place.
This caused quite the uproar. Many visual artists and several commenters are claiming Allen is a “cheat.”
Technically, I don’t think he cheated. In my opinion, it’s more like he used the AI to generate elements he then used to build his own vision. If he had cut and pasted the AI results without any adjustments, that would be cheating in my mind. Since the end product was controlled by a human being with a vision, I would categorize that as an artwork.
I will also say that if you love AI generated pieces, which several websites now offer go for it. Some of it’s really pretty and life’s too short to be surrounded by ugly. Enjoy.
But I firmly come down on the side that AI generated pieces are not art, and I believe AI generated content should be viewed with a healthy dose of caution.
To the first point: Why do I believe AI generated content is not art?
My answer: Because I believe a sentient being needs to be controlling the end creation and intent. Yes, art has intent. It is born of a specific point of view.
Jason Allen’s creation comes very close to the line for me. But, because he managed to find the right set of words to generate the content he needed, but mostly because he altered that content to better match his vision, I’m willing to buy into the idea that this series falls under my own “art” definition.
However, where we’re approaching the line from “art” to “content” is—again, in my opinion—if ten or twenty or two thousand people can enter the same series of words into a prompt box and voila! the exact same images are generated. That’s not art to me. That’s an algorithm. So, I do find it concerning that the artist reportedly hasn’t revealed the series of words which developed the images.
A friend on Facebook asked what the difference was between this artist and someone like Jeff Koons or Chihuly, since those artists would hire people to do the “actual work.”
Again, to be clear, in this case, I think this artist did do actual work…which I’m basing almost entirely on his Photoshop work post-AI generation.
But let’s assume he didn’t do anything with Photoshop. A person entering a series of words and picking out a randomly generated picture they like best is not making art. I can go do the same thing with creative commons on Google images or Flickr right now.
Entering words in a box is not designing a sculpture, as Koons and Chihuly both do/did. Sure, Koons and Chihuly hire builders and hands-on makers but there’s no doubt, when the piece is completed, that it’s a piece which came from the mind of Koons or Chihuly.
The same way architects are building artists. No one is going to mistake a Frank Lloyd Wright design, even though he didn’t construct every piece of his buildings.
To my second point: AI generated content should be viewed with a healthy dose of caution.
I have a visceral, icky reaction to AI content. It makes me want to write all kinds of 1984 style nightmare books.
AI content, like all computer algorithms and programming, is only as good/knowledgeable/safe as the people who feed it. When Microsoft released its chatbot on twitter, it took less than twenty-four hours to turn into a “racist asshole.” I mean, it was bad.
As Wendig points out, Midjourney has some problematic elements already too. For example, if you request a human face, odds are it’ll be a white face the AI generates. (This is being corrected…)
And similar problems are already alive and well throughout our computer-algorithmed world. Programming, which is pretty invisible to most of us dictates some serious parts of life. It affects whether your résumé is read by an actual human being, it affects teacher evaluations, it affects the advertisements you see. For further reading, check out Weapons of Math Destruction by Cathy O’Neil.
AI is just crowd-algorithming (it’s totally a word).
Forgive me if I’m not too excited about the “artwork” it generates. I’m not looking forward to the novels or the formulaic movies it will “create.”
My biggest concern—and this is definitely me wearing a science fiction/dystopian hat here—is AI’s ability to develop effective propaganda. It makes it too easy. Throw in a few words like “Nazi,” “white power,” and “racial purity” and you’ve just thrown in a recipe for neo-Nazi graphic design. Almost no effort required.
(And, yes, I’ve played with Midjourney and Dall-E, and they’re easier than Photoshop…)
Right now, AI creations are in the new, fun stages. Yes. IT IS FUCKING FUN TO PLAY WITH THIS STUFF.
We like playing with new toys and seeing cool, neato things. The birth of new technology is always a fun stage. But the honeymoon stage won’t last forever and the science fiction writer in me can’t quite let that go…