Yes, AI Does Make Art
My response to The Medium Newsletter’s snap poll, “Can AI make art?”
Yes AI can create art. Or at least let me put it this way: If a person believes that ‘random’ paint splashes like Jackson Pollock paintings, or vacant work like the blank white canvas painting, or purely stupid things like a banana taped to a wall can be considered art, then that person must acknowledge that what generative AI produces can be considered art. Note, this isn’t symmetric: I believe the above examples are not art, but AI can be.
What makes something art?
There are several positions about what makes something art, and I don’t think any of them are new to the debate thanks to AI. Every single opinion about what art is falls into at least one of the following categories:
- Art is in the eye of the beholder.
- Art is any form of creative expression.
- Art is an original form of creative expression.
- Art is something that brings beauty into the world.
Each position is more restrictive than the last, but they are really all up for interpretation at some level. If you take the first position, then you are aligned with people like Marcel Duchamp and Andy Warhol — basically you believe that art can be something mass produced or that ‘ordinary’ objects can be perceived as art. If you take the second position, then you are aligned with Jackson Pollock that the artist can fully remove his presence from the art without it losing the status of ‘art.’ In other words, something purely random with no driving creative soul behind it can be artistic.
If you take either of the first two positions, or if you agree with both, then you’d say a banana taped to a wall, a blank canvas, or random splatters can be art. These are the sky is the limit positions, and it’s pretty much the stance taken by the art critic industry today. Position 1 is patently stupid, making art or any classification meaningless, and position 2 is almost designed to get people arguing about feces.
Now what about positions 3 and 4?
If you take position 3, then you must believe that an artist is essential in the creation of art AND that derivative work is not art. So you have to disagree with Andy Warhol and Marcel Duchamp in this case. Art can’t be mass produced, and “ordinary” objects aren’t art. But you can believe that the prototype of a mass-produced work is art. Why? Because there was a person crafting it and every aspect of that piece involved a creative decision. So in this case you would appreciate art even including Brutalist and intentionally ugly work (actually, the mere fact that something was meant to be ugly would compell you to give it some artistic credit).
Finally, if you take position 4 then you are putting a higher bar on position 3 — not only does art need to have creative intent, it needs to be beautiful. This is where most people would agree on whether something is art. Old Masters, oil paintings, sculptures and all the things you see in classical museums or English palaces. Stuff like that. But even here, there’s room for disagreement because standards of beauty vary from person and place.
It’s about you, not AI.
When you properly consider these 4 positions, the real question isn’t so much “Can AI make art?” It becomes, “Can you accept or deny that AI makes art without reevaluating your existing opinions about art altogether?” It’s a more intricate question, but here’s a simple example.
Medium highlighted a position taken in the New Yorker denying that AI can create art:
For Deans and many writers, it’s the generative part of AI that’s disqualifying for art, a position really well articulated last week by writer Ted Chiang in the New Yorker, under the headline “Why AI isn’t going to make art.” Chiang’s argument centers on how the choices that a writer (or painter, or musician) makes are central to what we call art, and simply aren’t a part of what happens when AI tools get involved.
That’s a fine position to take. It falls squarely into position 3. But then you have to look at a person who says this and ask them whether a Jackson Pollock painting is art. If they say yes, then they are a hypocrite. A person cannot take positions 1 or 2 and deny that AI makes art because they cannot hold that originality or the presence of an artist is important. I think this is a difficult fact for many artists or critics to digest.
Only people who hold positions 3 or 4 have any room to deny that AI makes art. Someone holding position 3 would be fair in saying that the generative aspect of AI art is derivative by definition. Position 4 might say the same thing but oddly enough, someone holding position 4 might weigh the beauty of AI images at a higher value. So the most strict definition of art might be the fairest judge of AI out of all of them.
Hardly anyone has the conviction to take position 4 after the modern art movement’s dominance for the whole of the 20th century. Relatively few people truly hold to position 3 either. They hold position 3 when criticizing the things they dislike but shift to a looser position 1 or 2 when they happen to like something that most might not call art. Therefore, the same kind of critics denying generative art its due will go to MoMA, marvel at a banana on a wall, admire a blank canvas, and then write articles about the depth of those artistic works in turn. That’s a shame.
If you can’t look at obvious B.S. like a banana on a wall and say “that’s not art” then you can’t look at a Midjourney render and say “that’s not art” either. So if we’re going to ask lofty questions like “Can AI make art?” we ought to get some standards because right now they are so low that the question is basically pointless.