AI Art Will Never be Alive
A self-centered portrait generation app is a slam-dunk use case for AI. For the price of an expensive coffee, Lensa eats up your selfies and creates gorgeous paintings of you (or someone that looks kind of like you). These images are beautiful because they’re trained on a set of beautiful art that was hand-crafted to evoke the deepest reaches of our psyches.
My experience using Lensa was of mild awe. After selecting some images and uploading them to the machine, it thought for about an hour. Then there I was as: a fairy princess, cosmic goddess, Celtic warrioress, ageless cloaked traveler, and anime protagonist. The AI was spot on. It cloaked me in forest green and adorned me with flowers and leaves. The long, flowing hair I’ve tried to grow out was right there in generated brushstrokes, more vibrant than in real life. My face was thinner, exaggeratedly magazine-worthy. I ultimately chose to use one of the images as my profile picture. It’s like me but better. A perfect mask for the internet.
This is the perfect use case for AI, because I wasn’t going to seek out an artist to create these images of me. I barely even shelled out $8 for the AI to create them. Sure, in order to produce the art that the machine was trained on, hundreds of artists put in hundreds of man-hours with brush, pen, and stylus. But none of those artists were going to produce even one image of an idealized me. Although now that I see how much I like my AI-painted portrait, I’m thinking about asking for one created by a conscious, fleshy human.
Some of the images were lifeless — just drawings of a character who looked vaguely like me. But some struck a chord, and not just by playing on my surface-level vanity. OK, there was vanity. That’s the base appeal of the app. Seeing myself as a fairy princess and a cosmic goddess is fun.
But some of the images generated cast me as some Celtic warrioress, and these hit me in the bones. Did my many-great grandmother look like that, before her culture’s destruction? I felt closer to this ideal, seeing the image. The machine probably correlated something in my face with a resonant feature of existing art of someone with similar ancestry. Merging us together did something in me. That’s cool; that couldn’t happen before.
—
These tweets by Jordan Chase-Young crossed my feed one day after I’d been deep in a Midjourney creation flow.
Taking in these simple illustrations caused an epiphany. Jordan was right — this illustrator captured something evocative in a different way than all of the images I’d just produced using AI. The simple illustrations drove home something that is in retrospect very obvious.
I’d spent a few hours iterating on the prompt “Aphrodite vomiting planets and universes by Klimt,” having a blast creating minor variations and marveling at the contortions produced. The inquiry produced dark-haired women in various degrees of wholeness, mostly not vomiting at all, looking model-like while surrounded by stylized swirling golden balls for the most part. Some of them had a torturous depth. I wondered who were the women underneath the layers of interpretation; first by artist and then by machine, called forth thus altered by my words.
It’s shiny and spectacular that a machine can do this. But none of the images were quite right. I spent hours trying to make the perfect image by varying it with the machine, but it didn’t work. To make what I wanted, I’d have had to jump into Photoshop and alter the images.
The quality of the images reminds me of making a smoothie. Component pieces are processed into an entirely different form. The result is delicious, smooth, a bit unnatural, and impressively convenient.
A smoothie is nothing like a blueberry.
Sitting in presence with a blueberry is miles more satisfying than sitting in presence with a smoothie. They are both delicious and nutritious. But a blueberry is much more alive. Blueberries are pure organic creation, every last one. A smoothie — well, it’s got some qualities of aliveness, depending on the smoothie, and it hits some of the same spots. But it’s not the same domain of experience. I like to have seeds in my smoothie, so I need to put whole raspberries in it. But, personally, I prefer just eating raspberries.
Let’s torture this simile a bit more. When an artist spends time with a prompt or an alteration of the AI image, using it as a tool and not a crutch, it’s a brand-new cyborg berry. A cyberpunk bush made of wires and veins intertwined. (Hey, that sounds like a good visual prompt).
I, for one, celebrate the birth of grotesque new fruiting forms.
Of course, this analogy is flawed. For one, a smoothie is more expensive than a basket of blueberries. The potential to be priced out of their livelihood by a lifeless machine is one of the major pain points for artists. I feel this pain too. In the cold cruel terms of economics, when a new technology enables a shift in the supply side of a market, demand will respond, and some suppliers will be screwed.
This is the price of innovation. This is the pain of a world where we get to experience magnificent shining cyborgs, the likes of which have never been seen before. And… it still hurts.
I have faith that there will still be a market for blueberries. Even if smoothies are cheaper. I certainly don’t want to live in a blended-food world. Do you?
