Among the visual people, both the fine arts and design communities, there’s a lot of fuss about the development of Artificial Intelligence image generators. Personally, I care less about the debate around authorship and creative accomplishments of humans using tools like MidJourney or Dalle2, and more about the philosophical implications of it. In particular: when will we consider AI… alive?
I mean, I don’t think it is. But there’s progress.
While the generally accepted version of the threshold of “Alive” seems to be its ability to realize it’s existence and individuality, I tend to look more for a few other aspects of it. Such as it’s ability to procreate (which is a very measurable skill, and thankfully I haven’t heard about AI having babies yet) and it’s ability to create something new — which is something I can speak of with a bit more authority.
For those unfamiliar with the machine learning process, which is the core of what we currently call AI, it’s basically a program that isn’t programmed to execute a task, but to explore possibilities and learn from a system of rewards and punishment. Just like human kids do when they play. Or Lion cubs. Or puppies… young mammals in general.
Now, the computer ones have an advantage. An accelerator of sorts. It’s what they call adversarial systems. They “train” two AI engines with the same knowledge base, then assign them opposite tasks. For example: one tries to draw a face, and the other tries to determine if the face looks like a real one. They both get kudos for their performance, and learn from it. Then do it again, millions of times per minute.
You see? These systems no longer get programmed by geniuses to do what they would, like when Big Blue best Kasparov. They are now learning from scratch. They’re playing videogames like Forza Motorsport (and becoming not o ly very good but also mean about it cause the developers forgot to include punishments for nastiness like pushing competitors off the track). But I still think AI conquer of the Chinese board game of Go is way more iconic.
Go is a game of possibilities. Strategies. Trained by humans alone, it took AI a while to be trained enough to beat a champion. Then, they tried something different. They got two adversarial computers to play against each other, with no training whatsoever. And after a while, tested it against a champion again. Legend says the human grinned when the computer made a silly mistake and proceeded triumphant to the end, just to be surprised later and realize the error was I fact a brilliant move never seen or taught to any human. A move the computers created by themselves, by playing with the game and exploring random possibilities.
I happen to be a professional creative, and, the way I see it, creativity is a thought process that is opposite to logic. Instead of efficiently building a thought from the basics, aiming at the desired goal, we try stuff and see if they work. It’s inefficient, but when you hit it right, it’s magical. Now compare that with what the Go computers did? They created a move.
Cut back to our reality.
Here I am, Playing with images generated by MidJourney, asking it to imagine some pictures that would have made my creative brain hurt. Imagine connected brains in the style of Basquiat, I said. And it gave me the image I used on the previous post. I wonder what SAMO, who never lived to see the internet, would think of the result. Then I dared further: “Imagine Death, in the style of Dr. Seuss.” I grinned, in a preemptive enjoyment of my my little aft of cruelty. But then MidJourney did it. It gave me a few versions of it.
The results left me speechless.
Once I collected my chin back from my lap, I stared at these images, and I could not unsee, in each, a different story. I know, obviously, that part of this process is my creative inference on the result the computer spat out of a simple reward system and a wide database of images from which it learned from. Yet, I can’t unshake it from me. It indeed created something new. It didn’t give me a Grim Reaper with a Grinch look. It imagined creatures in contexts and universes of their own. Images that are original, authentic and clearly have a soul. Even if that soul was partly impregnated by my creativity on both the promt, the selection and the stories I am already making up in my mind.
Maybe that’s how we should look at it. As a Creative interspecies intercourse. A human fucking around with a computer (forgive the language but here, it’s the right term), and the creative babies being born.
I tried one more. Picked a fable told from one of the characters in my book, where a Tigress goes to war against a bee-hive, and asked MidJourney to imagine it in the style of Xu Beihong, an old Chinese painter famous for his vigorous nankin horses and birds. Once again, what it gave me back didn’t seem like a mathematical exercise, an algorithmic attempt of merging query terms. It gave me something meaningful, poetic, and absolutely original. An image I even got tempted to make into my book cover. So I added some typography on top of it and voila. Another interspecies baby was born.
Now I’m confused. Rationally, going back to the original argument, I can’t say the computers are reproducing, for it hasn’t created a baby version of itself. But it’s hard not to feel it created a “baby” of some kind. Maybe that’s just a romantic or even unhinged abstraction of my own mind, but they creation isn’t just mine. It is ours.
I know this isn’t an argument I can close. At least not until Midjourney spits not an image but another image generator itself, and gives it a name. And if it calls it MidJoirney Jr, then, we know: it’s time to run for the mountains.