Can Artificial Intelligence Make Art?

[Cheerful instrumental music starts]

Quinn: Hello and welcome to the first episode of Pictorial on Relay FM. I’m Quinn Rose, I am a person who did not go to art school, but I love art very deeply and I’m excited to be doing this show. Every two weeks we’re going to be bringing a different topic about art history and culture, learning about it, and sharing what we learned with you.

[Music fades out]

Betty: And hi, I’m Betty, I’m also someone who did not go to art school, I don’t have any art training. But I do host a show called “Articulations” on YouTube where I talk about art, architecture, and design, and I have actually been a gallery guide here in Toronto at the Art Gallery of Ontario for almost seven years, so I would say that I’m kind of a self-proclaimed art person with expertise. And I’m also really excited to talk about all kinds of ideas about art.

Quinn: Oh yeah, you’re way more qualified for this than I am for sure—

Betty: No, no.

Quinn: I’m just like “art is so cool!”

Betty: Well that’s who I was when I first went to the AGO to volunteer, they were like “so what are your qualifications?” I’m like “I took two art history classes in school!” and they’re like “good enough!” And then I just slowly built it up from there, I guess.

Quinn: Well we’ll be talking a little bit more in detail of what the kind of topics and what the show is going to look like at the end of the episode. For now I’ll just say, if you’re someone who has no art training but is just interested in art, this is the place for you. If you’re someone who has tons of art training, then also—that’s awesome, listen to this show, please don’t judge us.

Betty: Yes, we are not the establishment of art by any means.

Quinn: We’re the fun fringe podcast art people and we’re here to share cool art stuff with you.

Betty: Yes, exactly.

Quinn: And what are we starting off with?

Betty: Recently, or back in 2018, so this was about a year ago, there was an auction at Christie’s where an artwork that was created by an AI sold for $432,500. It’s a work that was done by a Paris collective. So they created this algorithm called “GAN,” G-A-N, which stands for “Generative Adversarial Network,” and so it’s basically an AI where they fed 15,000 portraits painted between the fourteenth and twentieth century and this generator—then from its learnings of those portraits—generated a new image based on what it learned. And so then they created this art piece, which we can put in the show notes, that basically it sold apparently for 45 times higher than its original estimate. And so yeah, it sold for almost half a million dollars.

Quinn: Yeah I have so many thoughts just about this painting in specific. Like Betty said, there will be links in the show notes, if you can’t look at it right now, just to describe it briefly—it basically looks like a portrait of, just like a man in black clothes, but his face is kind of… blurry, like it’s kind of off. It’s hard to describe exactly, but if you just picture a person and then just make it kind of weird. And honestly what it really reminds me of, is there was that thing a little while back of a priceless portrait in a church that someone decided to clean and then she messed it up, that’s what all of these portraits look like!

Betty: Yeah, it—it was one of like Jesus and then she decided to “restore” it and it looked like a blurry monkey person face thing. It does look like that, yeah.

Quinn: She is the person making all of the AI art. [laughter]

Betty: It’s quite possible. And for this one, for me, it does look like if you went and put in art from the sixteenth to twentieth century and put it together in a jar, swirled it around and dumped it out, I feel like that is what you would get.

Quinn: It’s so interesting, of the whole valuation of it. Because if you look at it, you wouldn’t think this is such an exquisite piece of art. Like it was valued this high because people knew that it was AI art, and so they said “that’s really cool, and I want to be the cool special rich person who has this first AI art to be sold at Christie’s,” which is like a big thing. It’s all about status. Like obviously all of art only has value because of what we decide it does, but I guess my point is I feel like in some of the research I did on this there was some focus on “oh it sold for this much money,” and I’m like, well it’s not like someone decided it was so aesthetically beautiful that it was more valuable. They decided it was more valuable because a computer made it.

Betty: Yeah, like the whole… obviously we could do a whole episode just on the art market and why things are valued a certain way and how that doesn’t necessarily reflect its quality, but it just reflects some sort of… I mean, there’s all kinds of things behind why certain paintings are more expensive than others. But yeah, like a lot of times it’s the actual artist, like the actual background of the piece that’s driving its price. And in this case, it was probably yeah what’s more interesting is the fact that it was made by an AI, not really because of how it looked.

Quinn: Another piece in the background on this sort of specific instance was this was made by, as you said, this French art collective called Obvious. But it was, uh, potentially directly pulled from, or at least very closely inspired from a GAN, a Generative Adversarial Network, that was made by this nineteen—like this eighteen-year-old kid, I think he’s nineteen now, named Robbie Barratt. So that was a thing that came up a lot too. It’s interesting because this kid’s nineteen years old and I saw him quoted a lot, so it seems like he’s almost become a minor authority on AI in the art world because he made this network. He threw all of these, like hundreds of thousands of WikiArt images into it, and you have the other program trying to make a fake portrait until it can fool the other one who’s seen all the real art. I mean that’s a really simplified version but that’s basically what’s happening. And then Obvious was like “we could do that too!” But the ones that they made look so close to the ones that Robbie Barrat had previously made. Which is weird, because neither of them had actually made them, the AI made them. So is it plagiarism? Question mark.

Betty: That’s a really good question. I just wanted to go back actually to a second to what you were saying earlier about the price. Because I was talking to someone at work recently, kind of in preparation for this I brought up the topic of AI art, and someone at work—he actually said “yeah of course computers can make art, but nobody would want to buy that.” But then I brought up, well did you know that one sold for $432000? And he was like “what?” So it was just interesting how this person’s first reaction was “why would anybody buy this?” But I feel like that could be a similar reaction as if somebody saw a really abstract painting that’s just like a red block, and say “well nobody would buy this” and then you say well, some people would. You don’t know what other people’s motivations are.

Quinn: And I think when you see AI art now, like we were talking about earlier, it very clearly doesn’t look right when they’re trying to make realistic images. AI art is better at making more modernist stuff because it’s abstract. And so now it’s definitely, it’s just like a status thing of wanting that cool computer generated art. I do have a personal theory about this, is that it will get better. It’s not—I don’t have the computer expertise to know, to be able to estimate, if anyone can at this time, how many years it’s going to be until it truly is indistinguishable. But I do have a personal theory that for this period it's going to be valued really highly and purchased by rich people as this status symbol of this like really cool interesting ting. And then at some point it’s going to flip, so that AI art, which is going to just necessarily be super cheap to produce and easy and basically no scarcity limits, and that is going to be incredibly accessible to people with less money, and then human-created art will be what only rich people have.

Betty: Hmm, that’s a really interesting theory, yeah because I feel like what you’re describing could actually be a similar parallel to photography. Obviously when it was first invented it was really expensive and you had to be really wealthy in order to access it, but then over time now we all have access to a camera, pretty much. Even people in developing countries have smartphones that can take pictures. And then so, even now, I mean photography still is an art and there are still really expensive photographs being sold. But yeah, it is kind of one of those things that as it gets more accessible and more people have it, it’s less valuable in certain cases.

Quinn: That’s such an interesting comparison. Especially because photography was also said, like people said photography isn’t real art. Because it’s a machine, right?

Betty: Exactly.

Quinn: And it’s a very similar parallel to like, oh AI isn’t real art, it’s a machine.

Betty: Yeah that’s actually, that was one of my points, as soon as you mentioned this topic I just thought of photography and how it wasn’t considered an art. And probably there still are people who don’t consider it an art, because they just think “oh you click a button and it’s done by a machine, it’s not done by you,” even though you set up the shot and do all kinds of things to make the photograph happen. It is along that continuum of, humans have for a long time have used technology and used machines to create art, it’s just now AI is another tool, that’s what some people would argue.

Quinn: Well I want to ask you whether you personally believe that artificial intelligence can create art. And I want to do the caveat that I feel like there is the possibility that we get to the point where we make true artificial intelligence and it has sentience, and I feel like that’s a very different question. With sort of the models that we’re working with where they are like definable computer algorithms where they’re sort of scanning art and creating it just based on what already exists, do you believe that that is art? Do you believe it will get to a place where you consider it art? What are your thoughts?

Betty: Yeah, so my thoughts are—hopefully I don’t make this too convoluted, but I do want to go back to a couple of examples from art history. So this is a piece called “Light Prop for an Electric Stage” by László Moholy-Nagy, and he was a modernist artist who did all kinds of things. He did everything from photography to installations. And this one is an installation he did which is basically this sculpture essentially, but it’s got like circular, rectangular, interlocking parts, and these metal rectangles that jerk around automatically, metal discs that spin around and glass spirals, and there’s also like 130 electric light bulbs that shine through the piece and there’s moving patterns that go around the space it’s displayed in. And then… so there’s him, and then there’s a couple of artists which I won’t go into extreme detail but we will have links as well, there’s another artists I think of whose name is Julio Le Parc, and he does this automatic moving pieces of metal and light that’s spontaneous and has like light and shadow. And then another artist whose name is Rebecca Horn, and she has—I have a video here too—where she has violins that move by itself and drums that play itself and these weird chairs with knives on their leg that dance randomly. And they’re like timed, it’s programmed to do all these things. Anyway, the reason I’m bringing this all up is these are all in a category called kinetic art. But nevertheless, these are artists who have programmed a machine to do something. To either dance around automatically or turn on and off automatically. The artists did initiate the work but the work is kind of performing itself. When it comes to the question of does artificial intelligence create art, I think back to these examples as, these art pieces do perform by themselves, but do do things by themselves, but they’re put in place by a person. Even the art group Obvious and even a lot of these works created by an AI, it’s like but somebody made that AI. Somebody programmed it or initiated it, or somebody put in the input of those 16000 artworks or something and clicked a button. So it’s just that, because obviously Rebecca Horn’s work, she’s attributed as the artist. And some of these AI pieces where you have the computer that’s being attributed as the artist but you then think, well then how much of the artist’s hand has to be in there for them to not be considered someone who made the thing? Or can you just have someone who literally is just someone who pressed a button, and they can be considered the artist? So that's my dilemma, is I think we’re still not at the point where AI independently decided to create a picture, there still has to be a human input somewhere.

Quinn: No I think this is a complicated question. Going back to the copyright thing I mentioned earlier, it’s almost like there’s this sort of dispute over whether, like who actually made this algorithm? Was the algorithm that Obvious made different enough from Robbie’s algorithm that it actually counts? And does it even matter?Bbecause they’re all from a very similar AI framework. And I feel like that kind of complicates that question as well, if it’s… that’s almost a question of if we count the human input as part of the art, but all the human input ends up with the same result, is it still art? Is it just bad art?

Betty: Yeah, well, there was actually an interesting video I watched from PBS Idea Channel, which is kind of asking, it’s also a video called I think like “Do Machines Make Art?” or something. And in there Mike, the host, he said—I like the words that he used, which is like he’s saying the question is not necessarily can machines make art, it’s can machines make compelling art? It’s can machines, and are we going to allow machines to make art independently on its own rather than telling it to make art based on what we want? Because we all agree—like again I asked another friend this question and I seem to get a lot of people saying  “yeah, of course machines have the ability to make art,” like we can program something to output something else. But it’s whether we agree that it’s art or not.

Quinn: So my basic answer to this question is that art is something that is created beyond practical value, basically. If there is an intention behind it to not just be for practical use.

Betty: Yeah so I actually, I really liked your definition because actually especially because of the word “intention.” Because it goes back to asking “can AI have intent?” Like its own intent, not just what we tell it to do. And so, does it matter that the intent was put there by a human creator, or do we want the AI to have its own intent for us to consider it as art. And the answer could be one or the other. And I remember I think my definition was also, I said that art exists between the artist and the people. It’s the viewers who decide whether it’s art or not. But I also said I want there to be a story, a reason, or something that the artist is trying to express. So then it goes back to the question of what is the story, the reason, or the expression the AI is trying to convey? And when you ask that question, sometimes it’s there. Like in the example of this piece that was sold at Christie’s, by… so I think it’s called the portrait of Edmond de Belamy, I’m like—I don’t know what the intent of the computer is. Maybe there isn’t one. Or maybe there’s so many intents, maybe its intent is just to output art. For me that’s not compelling enough, but maybe for some people it is.

Quinn: I think there’s a part of this for me that’s… like the more I peel it back—I feel like my gut reaction is “no, that’s not art.” And then the more I peel it back into like, sort of like working through logic first principles of what I think art is, I’m like “well if there’s human intervention in the intention of it, then I guess technically it could count.” Even if a human didn’t physically do it, is that important enough to count it as not art? I don’t necessarily think so. So… but I think maybe it’s just art I don’t like.
Betty: [laughter] Right, yeah. Did you at all come across the Google AI Deep Dream generator?

Quinn: I saw a little bit about that, but I didn’t really play around with it.

Betty: Yeah no I didn’t either, but maybe I will try to input—I tried to go on, but I think the website crashed or something. But it’s basically, it’s this really… I’m going to try to explain it, again I’m not a computer person, so I’m going to try to explain what this is, but it might be a very clunky explanation. But it’s basically a, like the Google AI team developed this… at first they were trying to develop a system for the computer to detect what a picture is, so they show a picture of a bird and they want the computer to know that it’s a bird, and they want the computer to know what type of bird it is. So it’s like “oh it’s an eagle,” or whatever. And they want it to be able to be able to identify its species, so the same thing, they would show it like a tree or a leaf. So they basically put all this visual input into the computer to teach it how to identify a picture. Show them a picture and have them identify what it is. So what they then did, they did this weird thing where they reversed the algorithm, where they tell the computer “bird” and get the computer to reverse the… like draw something. To basically tell us what you think this thing is. I think I’m just gonna populate, I’m gonna send you the Wikipedia page on it, but basically it ends up being these really really creepy looking, almost like you’re on an acid trip type of pictures. I think if you, on the Wikipedia page if you scroll down a bit there’s like a picture of—I want to say dogs, but it’s a dog layered with eyes and legs and random eyes everywhere. It almost looks like a fractal but more horrifying. So anyway, because like you’re saying it’s art that you don’t like, for me it’s the same thing. It’s… I would not have this on my wall, or anywhere near me, because it’s just really really horrifying. But it’s really interesting, because it really is, because the computer has learned so much about—or has received so many visual inputs, it’s just putting back out what it thinks it’s supposed to say. So anyway, I wonder what you think of it.

Quinn: They are very creepy. Honestly, I do kind of like these better—

Betty: Oh!

Quinn: —than the sort of blurry portraits. Even though they are terrifying, some of them are really cool.

Betty: Some of them are, yeah.

Quinn: I almost like these better. Almost because I feel like it’s very deliberately a digital art style, it’s not trying to look like a painted image the way some of the other AI art is. So I feel like it almost, even though it’s disturbing, it still almost looks like something that a person could have made it because it is supposed to look like digital art.

Betty: Yeah that’s true, I was looking, I came across an article that was basically “oh, see if you can tell if this was done by an AI bot or by a human,” and obviously it was all abstract paintings, and again you’re like yeah if it’s like a geometric or abstract painting, it could have been done by a person or by a computer. To me that’s like whatever, obviously if a person can paint a red square so can a computer. It’s interesting because this deep dream generator is one where there really isn’t an artist, like a human artist behind it, unless you count Google’s AI team as the artist. So this is one where, because before with the examples I showed before, you could always say “oh, Rebecca Horn did this.” Like she made this mechanism. Whereas this one, you can’t really pinpoint a person, you just have to say “well the Deep Dream Generator is the artist who made it.”

Quinn: Ooh, such a fine line.

Betty: Yeah, so—or you can say well no they didn’t, because it always goes back to whoever developed the Deep Dream Generator. Or there’s actually, I just realized, there was another piece I wanted to show you that kind of also blurs this line. So it’s an artist called Roxy Paine, and what she does is she does these, well she calls them auto sculptures. I think the name of the piece I just sent you is “Scumak.” And so she has like a machine that melts plastic, like the one that I’m showing you is red pigments. And it periodically extrudes them onto a conveyor belt. And each sculpture is unique and different, and again it’s all automatically made. So the only thing Roxy Paine did is make this machine that spits out paint. She has another one where she actually has a, another one that spits out white paint that she just has on a computer controlled timer that just sprays white paint onto a board every once in awhile. And it created these really cool looking land—what looks like landscapes and things like that. So anyway, this is another one where obviously Roxy Paine is credited as the artist, but she didn’t make any of these, technically. It was made by the spray nozzle. She didn’t decide this sculpture is going to blob over here, it’s the, whatever, it’s the air currents that happened to have been in the room and gravity that resulted in this artwork. Again like you were going back to, she didn’t make these decisions about how this art was going to end up, she just kind of prompted its initial generation, so did she create the art or did the machine create the art by itself?

Quinn: I guess I think considering all of these examples that we’ve talked about on the different sides of like, the AI art line, I think that—what I’m going to say is my concluding thought on this, is that for me the real difference is do I see—and this is I think different people have different perspectives on this, it’s not a firm line—but do I see this as the artist using a tool to make art in a different way but there is still a person, or a small team of people who are making decisions about it? Or is this just people using a neural network to sort of scan what has come before? Because I think one of the big things about this that really makes the difference is that in the portraits and the abstract art that we’ve seen created as opposed to the sculptures and the kinetic art and stuff like that, is that the sort of neural network stuff is only able to make things based on what’s already come before, instead of being—like AI technology at this point… At this point in time, there is no ability to truly innovate. You’re not going to feed one of these GANs a hundred thousand pictures of portraiture and they’re not going to spit out, for example, Duchamp’s The Fountain. Because they just don’t have that capacity. And so I think that is… even though there is a level of human intention in writing the algorithms, there’s still not the same level of human intention and true human innovation that comes behind even these works that use sort of this quote-unquote “artificial intelligence” or artificial physical creation in the end, there’s still a much closer hand in that.

Betty: Like I think one reason these things like the Deep Dream Generator and the GAN technology is interesting and we do call it artificial intelligence is—if you really think about it, us as humans—like me, if I were to go create art, I would be also outputting something that’s based on things I have learned in my, like growing up as a person. Like I would have seen hundreds of thousands or whatever of portraits and landscapes and abstract art, and say that I decide that I want to be a landscape painter, I’m not necessarily—like on the topic of plagiarism—I’m not necessarily going to just copy another person’s landscape, but what I create could possibly be a composite of the types of art that I’ve seen and I’ve learned as an artist. It just so happens I’m an organic computer who happens to have seen more than what some of the computers, some of the artificial intelligence have. I happened to have seen trees and lights and gone to art museums and have seen movies and stuff. So the culmination of what I know, and what I output onto a piece of paper, you could argue is just what I’ve learned. So in a way you can say a human and a computer is really no different because we’re all just receiving inputs and using that as a background to our creations. There’s also a lot of, I don’t remember the quote exactly, but a lot of artists just say all art is kind of an iteration of what has come before. So in a way you can argue yeah, artificial intelligence is kind of just doing the same thing.

Quinn: As a sociology major, I will say, you’ve got a very good point.

Betty: [laughter]

Quinn: That is so true, and really fundamentally when you break everyone down to their building blocks, everyone is so shaped by everything that has come before. Is there some magical spark in human beings that make us innately able to create in a way that animals and machines never will be able to? I don’t know. Maybe one day we really will be able to make a computer that is as complicated as the human brain and it hits that level. We don’t know. 

Betty: Yeah, and obviously—but I would definitely agree that at this point in time I don’t see computers as creative and as able to generate new ideas as humans. But I do think that’s just the limits of technology, and the limits of well, we just don’t—computers right now just don’t have that capacity to have as much input as like a human brain can. But probably sometime soon in the future they will, and as you said they’ll take over, and then they’ll be making the podcasts.

Quinn: Yeah. Well I think the big kind of takeaway with all of our conversation about this is that there isn’t an easy answer? That’s going to be the theme of the show! Congratulations.

Betty: Exactly, we don’t have an answer to anything.

Quinn: Yeah, well not every episode is going to be so philosophical about what art is, but that definitely is going to be one of the themes, is looking at different things and talking about whether or not they’re art. But we’re also just going to be talking about cool stuff that’s happened in art history, or weird things about contemporary art culture that you wouldn’t think about if you weren’t in working in a museum or wealthy enough to be involved in Christie’s art auctions. [laughter]

Betty: Yeah, or we might even talk about some artists that aren’t very well known, or that you’ve never heard of but who are still really interesting to us, so we might go over that.

Quinn: Yeah! We have a lot of directions with this that we’re excited to dive into. We also want to hear from you about what you think. So there is going to be a form in the show notes where you can suggest a topic, if you have something in particular that you want us to cover. Clearly it’s a very broad field so maybe with time we’ll narrow down our focus, and we’ll let y’all know about that. For now, we’re keeping it open.

Betty: Mmhmm. So anything that you think is at all related to art, feel free to suggest that as a topic.

[Cheerful music starts]
Quinn: Thank you so much for listening to this first episode of Pictorial. If you want to follow us we have a Twitter account and Instagram account with all the images that we talked about today, @pictorialpod on both Twitter and Instagram. And I am at @aspiringrobotfm on Twitter and Instagram.

Betty: And I am @ArticulationsV, so that’s Articulations with a V at the end. Or you can follow me on YouTube at Articulations.

Quinn: And you should! It’s great! Watch Betty’s videos!

Betty: [laughter] Aww, they’re okay.

Quinn: Thanks for listening, art enthusiasts!

[Music fades out]