AI Art will never hang on a wall

I’ll have just one more essay for the rest of the year. Thought about posting through the holidays, but you probably don’t want to hear from me.

Be with your family and/or friends and enjoy however much time you have off work. I’ll be back in January with some good stuff.

Also, because my son has wanted to listen to How the Grinch Stole Christmas in the car on the way to daycare and every version I’ve found is so terrible, I recorded my own version.

It’s possible this is something I’ll continue to do with children’s books. I may even start doing podcast-esque posts, though these would primarily be me reading an essay and trying to make it sound as if I’m speaking spontaneously rather than through meticulous planning.

The above is AI generated using DALL-E2 from the prompt “wolf howling.” It looks mostly all right if you ignore the wolf’s head! I don’t know what kind of nightmare is happening there, but the fact that this seemingly very simple prompt spat out this monstrosity may be all I really need to say about AI Art.

There are serious labor and intellectual property issues to discuss with relation to this, especially since a major publisher, Tor, has begun using AI art for novels they’re publishing. It should be no surprise that a large, very successful publisher would be among the first to try to find a way to stop paying artists, but it sure is depressing to see it happen so quicky and with such hilariously bad results. John Scalzi wrote a bit about this.

My biggest thought on AI generated art (whether visual or written) is that no one actually wants this shit. Yes, it’s fun to play with and it probably has some commercial utility, which I’ll talk about in footnotes right here¹ and here² and also here³.

Outside of that potentially positive utility which can amount to saying it saved me x amount of hours, I don’t really think this has much application outside of the enjoyment of giving the machine funny prompts and seeing what it spits out. And this is because humans like art, but they only actually want to engage with art made by people.

Watching the machine create strange, surreal pictures for the prompt you give is sort of like watching an ape wearing a tuxedo. It’s diverting, but you’re not going to do much with it. Even the seemingly complex responses this new chatbot gives are amusing in the way it’s amusing to see children say Fuck or watch a bear riding a bicycle.

But even if the art got as good at the technical aspects of drawings/painting to be a master, I think humans will still recoil from it. Same thing goes for a chatbot that manages to dump out a comprehensible and interesting story.

We all know what the Uncanny Valley is with regard to CGI in movies and games, but I think the same principle hits for this other stuff. Even if an AI could create something as beautiful as the Sistine Chapel, we would find it cold and sterile and oddly frightening rather than awe inspiring. And that’s because of the intangibles that we rightly or wrongly assign to art.

A machine coming up with a perfect math formula to describe something true about reality is not as interesting as some big-brained dork at Oxford coming up with the same math formula for the same problem. And that’s because we know what it is to be human. We know how difficult it is to accomplish anything of value, let alone something other people derive value from. And so a human overcoming these difficulties is beautiful.

When a computer does it, we instantly value it less.

It would be like being impressed that a hammer was useful for driving a nail into a wall.

When computers first beat human Chess and Go players, it wasn’t awe but horror that filled those rooms.

(This documentary is amazing, incidentally, and well worth the 90 minutes)

And now that the AI can reliably beat any human player at either game, we don’t now watch games played between the AI. Elite players do for training purposes, but no human wants to watch two computer programs play a game against one another.

Yet, despite every human being worse than the AI, we still watch humans compete in Chess and Go.

The reason is quite simple.

It’s not so dissimilar from the humble spider web. We don’t value the beautiful and intricate designs made by a spider, but if a human made the exact same thing, we might be astounded by the craft displayed.

And, look, I’m not saying that this is rational or the right way for us to be. Obviously a lot of work went into making these AI that can accomplish a monstrous, faceless wolf like the one above, or that can answer essay prompts from the SATs with reasonable enough English to get a decent score.

But we will never value it the way we value a real flesh and blood human making the exact same argument or painting the same faceless wolf.

No one, excepting weird tech freaks, will ever hang AI art on the walls of their homes. Probably no one will even put it on the walls of motels. And no one is going to buy a novel—even a good one!—that a robot threw together.

The novelty is interesting and a lot of fun, but the lack of intent actually does matter to us as people. And, sure, someone on twitter or reddit has probably spoken at length about how they only care about quality and couldn’t give less of a shit if what they’re experiencing was done by a human or AI, but these people also probably invested in crypto and think that the only subjects that should be taught in school are Math, Science, and Logic.

And so maybe, as a writer of words and professional opinion-haver, maybe I should worry about the new chatbot, but I honestly just don’t care. The chatbot will become very good at some things but it will forever be a machine regurgitating. And while I am just a human regurgitating my experiences of life through the prism of all the different artists and people who have impacted me, I think people prefer a human idiot over a robot genius.

And, I mean, I could be wrong, but I won’t be proven wrong for a long time.

I hope I’m dead by then.

1. AI visual art will probably be used by graphic artists as shortcuts. As in, they may use this kind of thing to fill in the background of something they’re working on. Or to fill in some kind of space that takes little skill but requires some amount of time that the artist would prefer to save (think of it the way Michelangelo used his apprentices - if this analogy is baffling to you then I don’t know what to tell you).

2. AI generated text will likely save certain types of writers a lot of time. Like, if the AI can spit out a press release or a quick summary that outlines the services provided by a company on their website, I imagine this will get a lot of use. Then the human can clean it up and adjust what needs to be adjusted without having to spend the time to come up with it from scratch. There also seems to be a decent amount of utility in coding, where you ask the machine for some kind of coding and it spits out what you need: basically a giant archive of formulae.

3. The other biggest use for this is that it makes propaganda effectively free. Pumping out junk science or political screeds of dubious merit and antisocial intent will become so easy to make that it will likely flood every place where conversation through text routinely happens. This is not great.