It seems like there are a million and one computer-driven visual art fads that have popped up in the last few years, but when Dalle-2 and Midjourney hit things got extremely weird. Typically these art generators come in the form of apps that are filter-based and are based on user-submitted photos, but Dalle-2 and Midjourney are straight-up Artificial Intelligence, creating imagery strictly from text prompts. Now, that’s not to say that these generators aren’t based on existing artwork and photos to some extent as they are trained on the enormous sections of available image file data found online whether it’s crawling google for image searches or utilizing existing databases like Deviant Art or Flickr. But the process for generating “art” has an ego-driven allure that is hard to deny and makes the user feel like they “created” something. This is for sure going to be a huge bone of contention going forward because it appears that these services aren’t going anywhere, and are in fact becoming more robust by the minute (and in perfect Skynet fashion will probably, eventually create living holograms made of pixels and murder.)
When Midjourney first hit, my wife and I were immediately sucked into the process because of the fun and potential that the service offered. For my wife it was all about utilizing the application to help bring to life imagery from fictional stories she was writing. It was helping her to visualize her the environments and sometimes spurring ideas to make scenes richer based on what it was throwing at her. For me it was all about fun. What twisted and weird set of prompts could I feed the machine that would hopefully spit out something hilarious and undeniably cool that I could use to start conversation on twitter. But that quickly escalated when one of the prompts (and the A.I. art results) really struck me as legitimately awesome. I took two things I love, 80s era BMX and Monsters, and discovered that they are in some ways better than the sum of their Frankesnteined parts. My prompts were essentially just the following with slight variations “80s”, “BMX”, “Ghosts”, and “Rad” Here are the first few images that Midjourney spit out…
I mean there’s a lot to unpack here in terms of accuracy, fluidity and perspective, but the color scheme, tone and concepts are so on-point that it hurts not to have actually painted something like this myself.
The way that the A.I. interprets ghosts is like something out of dream logic and I love it. The above imagery evokes the Pilot from Alien, but also the hellish imagery of demons and skull & bone based technology that is disturbing in the best way.
The cartoonish aspect of the ghosts in the above piece, along with the overall symmetry and layout suggests a book cover for the most awesome book that’s never been written specifically for me. That’s heartbreaking.
Though my prompts never quite evolved much past what I did above, a friend on twitter took my ideas and ran with them for a bit coming up with some pieces that felt much more intentional and dynamic…
I instead became obsessed with a feature where you can select an image generated by Midjourney and have it re-run multiple variations of the same general idea. So the book cover from above became this much gnarlier and intense version…
At this point I was seriously into this idea of Ghostly BMX riders cutting up the afterlife and racing in real deal Helltracks (if you’re not aware of that BMX reference, seek out a copy of the 1986 movie Rad like right now.) I started hatching plans for creating A.I. art zines based around these spectral bikers. Maybe I would use the art to inspire short stories or poems, It would be the Leaves of Grass of the 80s nostalgia generation. I started pumping out the imagery trying to find the very best pieces. That is until I realized that you were limited to the number of prompts you could freely request on a monthly basis. That’s when I learned that there was a paid subscription for the service to give you free reign to make art. And then there was a separate tier for high resolution versions, and yet another for the ability to use the art in commercial settings. This was getting complicated. But the wife ended up subscribing for her use of the generator in fleshing out some of her stories, so I piggybacked on that to create a bunch of new “art.”
Every prompt was like getting a fix, taking a hit. I was literally getting a high off of seeing the imagery as it rendered and itching to save the files and to generate more. Every new variation got me so excited and I legitimately felt like though this tool I was creating something meaningful. Even if none of these made perfect sense visually, the tone and concepts were superb. Some nights I’d spend hours just generating endless variations of the same image hoping to refine it to a point where it was perfect.
Very quickly I got that feeling that I was chasing the dragon. No single image that I was generating was useful in an of itself, but I hoped that with enough variations the combined set of imagery would be undeniable. I was generating so many images that the idea of trying to wade through them to find the best became daunting…
After a week of generating almost a thousand pieces of art I started to resent the whole process. On the one hand I felt like I was building up a bunch of amazing imagery to make something creative, a zine about BMX Ghosts. But on the other hand I felt myself becoming a slave to the process, trying again and again to tweak the text prompts to try and break through into something just a little more amazing, a little more coherent, a little more perfect. It started to feel like a job and one that was ultimately meaningless because no matter how much fun I thought the core idea was, I was never going to be happy with putting my name on something that I did not create.
And that is the crux of the issue, and I could plainly see that I was falling on a side of the issue about the merits of A.I. art. No matter how good this stuff gets, no matter how perfectly it can translate a set of ideas into an image, it will never be something we create. It will always belong to the computer, the processor, the lifeless spark of a machine. We can trick ourselves into believing that our input is the creativity, but when you boil it down ideas are a dime a dozen. It’s what we do with those idea that becomes art. And we aren’t the ones executing on these ideas. It gets more nefarious when you consider all of the imagery that has to be fed into the databases in order to create these pieces. Even more so when you consider that all of us utilizing this are just training it even further like the human batteries in the matrix. It’s why I had to stop cold turkey as it was leading nowhere and sucking up way too much time. This is a personal stance, not meant as a judgement of folks using these. Maybe people will be able to find a good use for this stuff, but I just don’t see it as anything other than a quick lark on twitter or as a meme. Both of which are immaterial and lead nowhere as well.
That said, I can’t deny how fun it is at first. I still enjoy scoping these images even though I’m never going to do anything with them besides sharing them in this piece. Heck, maybe one day I will write a story about these and call it BMX-Hex. Or BMHex. Or Bicycle Ghoulies. Or Radighouls. Who knows…