Will AI Replace Human Artists? One Study Says Not Yet - Terms of Service with Clare Duffy - Podcast on CNN Audio

CNN

4 hours ago

Clare Duffy
00:00:01
'Welcome to another episode of Terms of Service. I'm Clare Duffy. If you've been on the internet at all in the past few years, whether you realized it or not, you've probably come across art that was generated by AI. With programs like Midjourney and DALL-E, which were trained on visual data from across the internet, users can generate images from text prompts in just seconds. Some of these images are beautiful and striking, but a lot of them have contributed to a growing AI slop problem across the internet. And all of this has artists worried about their jobs in a field where it can already be challenging to get viewers' attention. Legal battles have started to emerge over this AI-generated art, too. Earlier this month, Disney and Universal teamed up to sue Midjourney, claiming its AI-generated images of their famous characters violate copyright law. And this trend raises an even bigger question too, what does all this mean for the future of art and creativity? We probably can't answer this in a single episode, but I'm excited to have Sheena Iyengar here to help me think about where to start. Sheena is a professor at Columbia Business School where she and two PhD students recently published a fascinating study about people's perceptions of art. And how that perception changes if the art was generated by AI rather than created by humans. In the study, she showed participants art that was labeled as AI-generated, as well as art labeled as human-generated and had viewers rate them. We'll get into it all, but spoiler alert, people still seem to like human-made art better.
Clare Duffy
00:01:45
Hi, Sheena, thank you for being here.
Sheena Iyengar
00:01:47
Hi, thank for having me.
Clare Duffy
00:01:49
'So it feels like AI-generated art has exploded over the past couple of years with more people using Midjourney and DALL-E and other programs that you describe the image that you want to create and then it pops out a result. When did you first start noticing this phenomenon?
Sheena Iyengar
00:02:08
'It was in 2020 when the world was shut down. I actually ran across this pretty well-known computer, well, he's a mechanical engineering professor but does stuff using generative AI. And so I discovered that he was making a robot, a robot that he called an artist. And so what he did was he fed the code all kinds of things, like leaves, trees, everything. It was over a million different let's call them like pieces. And so then what he would do is he would ask the robot to create a piece of art. Now, he didn't train it on what art to create, but he absolutely fed it pieces of art that had been successful. So...
Clare Duffy
00:03:07
And this is like a physical humanoid robot?
Sheena Iyengar
00:03:10
'Think of it as a printer. Okay. And it would literally in 24 hours, create a painting. And he actually had examples of these paintings. And he was training it to make impressionist art. And we actually showed pictures of these impressionist paintings that were created by this artist, AI artist. And people could not tell that it wasn't made by a human. So that was my first introduction. I saw this and I was like, hmm. This is interesting. What does this mean for creativity? What does it mean for the difference between human made versus non-human made?
Clare Duffy
00:03:56
Why is it that this technology and the outputs of it have improved so much in the past few years? Like it feels like back in 2020, 2021, a lot of the AI generated images you would get would be pretty cartoonish. It was obvious that they were not real, but now you are more and more seeing very photorealistic images coming out of these AI generators. Why is that?
Sheena Iyengar
00:04:21
I mean, you could ask the same question for lots of technologies. It doesn't have to just be AI art, right? I mean why did the, initially when you had the camera that came out in the 19th century, you know, artists didn't think it was beginning to become competition, but you know pretty fast it became better and better and better. Of course we all get better and better because what's it doing? It's iterating. Just like a human being, we're iterating, you know? You see it and he's, ah, that's wrong. Let me fix that part. Then you iterate again, and you fix that. Now, did the camera actually kill art? Well, no. Did it kill portraits? Yeah, right? Because it's cheaper, you could be more precise, and you can be faster to just take a photo of someone than have someone actually make a portrait of you. Just like with the camera, the average individual Suddenly has access to something they didn't used to have you know every single one of us can now take a selfie. That's kind of And every single individual can now also take an idea that they're trying to explain to their friends or they're just trying to be therapeutic with themselves and be artistic and they can now take this app and it sort of levels the playing field. Are we all gonna be equal? No. Are some of us going to be better at using the generative AI to make art? Yeah. And we're gonna be more creative about it and that'll become its own art form. And that's what you're seeing.
Clare Duffy
00:05:58
'So now that we know a bit more about how AI-generated art is made, what do humans make of the art that's created by computers? That's after the break.
00:06:11
'Okay, so Sheena, you recently worked with two Columbia Business School doctoral students to research how people react to AI-generated art compared to how they perceive human-made art. What question were you trying to answer exactly?
Sheena Iyengar
00:06:27
'So we looked at two things. First, can people actually tell the difference between art that's been made by the human versus the non-human? It turns out we really can't. Even though people swear they can tell them apart, they actually don't know if it's human or non-human made. But if they it's been made by a human they will price it higher. So that means they value it more than if they think it was been made by AI.
Clare Duffy
00:07:02
And how did you go about this research? What kinds of art did you show the study participants?
Sheena Iyengar
00:07:08
'Well, we couldn't show them art that's famous, right, because then they know. Although you'd be surprised because there's times when I show people the initial self-portrait that Picasso made of himself and people now increasingly tell me, oh, that must have been made by AI. But we, you know, this is a few years ago, so people weren't quite as savvy or at least as convinced that everything is made by AI. So, you know we just showed them art that wasn't famous. Uh, and in some cases we had them tell us, do you think this was made by human or AI? Uh, in some case, as we told them it was made by a human or AI, and, you know, obviously we didn't tell them the truth. And then we asked them to price it or how much they would sell it for. So that was the basic design.
Clare Duffy
00:08:02
'And so what did you find? It sounds like people do value human-made art above AI-made art, at least for now, if they can tell the difference.
Sheena Iyengar
00:08:14
If they believe it was made by the human, so it could have actually been made by AI, but if they thought and labeled it themselves as being made by a human, then they priced it higher.
Clare Duffy
00:08:26
And how much higher are we talking, like hundreds of dollars more or?
Sheena Iyengar
00:08:31
Oh, I wouldn't focus on the exact price because we didn't tell them this was fancy art.
Clare Duffy
00:08:38
What surprised you most about your findings?
Sheena Iyengar
00:08:42
'I was mildly surprised that they couldn't tell them apart. I didn't expect that. I suppose after we observed that, I was no longer surprised. We actually even tried it for fun. We actually put it in the Columbia Business School on the main TV screens as you walked in, and we had random people, like MBA students, as they're walking in vote. They couldn't tell either. So I was a little surprised by that. I go back and forth as to whether I'm surprised by the fact that people value something they believed was human made as higher. So of course, if I know that this is let's say a handmade rug versus a machine-made rug, then of course we take it as a given. We would pay more for the handmade thing. So in that sense, it's not surprising.
Clare Duffy
00:09:44
'Did any of the survey participants sort of elaborate on that piece, why they felt like the human-made art is more valuable? Is it just because artists put a lot of heart and soul into their works, whereas AI doesn't have heart or soul?
Sheena Iyengar
00:09:59
'Yeah. So people think that if it's made by a human, there's more intentionality behind it, that there's a story. I mean, the reason why you made this guy in the way you did is because there's something related to, I don't know, your personal story or, you know, something you saw that you were trying to communicate to the rest of us to see a kind of meaning. Whereas if it is generated by a non-human, you know, it's just random. It doesn't feel meaningful.
Clare Duffy
00:10:30
'Has this study changed how you think about AI-generated art?
Sheena Iyengar
00:10:36
'Yes, so my Ph.D. student and collaborator, Carl Blaine Horton, I remember he first brought me this idea. I was like, yeah, why would anyone want this AI thing? I mean, it's just weird. You can't call that a real artist. And then he started to send me poems. Because, you know, before AI art, they were doing all this, like, poems and they would be tested by some touring prize or whatever. And he's like sending me these poems and I'm like, no, no you can't like come on a poem has meaning to it. And he said to me so what if seeing something even if it was randomly generated by a non-human gave you ideas that you otherwise wouldn't have had. Would you consider that meaningless? I said well I guess you're right. It is, it could be used for that. I don't know if you know this, but I am blind, and I love to go to art galleries and learn about the new art. And I will tell you that one of the displays which really fascinated me was we went in and they had like a million new colors that were generated on a canvas using a computer software. I thought to myself, hmm, that's very interesting. People might actually start to create more colors that they otherwise might not have thought of and they might not have thought for two big reasons. A, we're so used to seeing certain color patterns in our everyday life that we're naturally going to put certain colors together or assume certain variations in color just because of habit. Imagine if there was another entity that didn't have to worry about your predisposed biases and could just be egalitarian and create whatever it wants to without thought to the past. That could be very interesting. And so that was one thing that struck me. And then the second thing that struck me is even if it has no intentionality, which clearly it wouldn't, it could still teach me something. So I believe that the camera taught people something about their world, that they didn't see as easily or as clearly or as precisely as they learn to see through the use of the camera. And I think the same thing applies here.
Clare Duffy
00:13:47
'That's so interesting. Yeah, I hadn't thought about that idea that, like, the creativity of it, the sort of randomness of it could potentially create something new in a positive way. I also think when we talk about the reaction to AI-generated art, we also have to talk about the reaction from human artists who have raised concerns that...For example, their works have been used to train AI without their consent. How big a concern is that in your mind?
Sheena Iyengar
00:14:19
So I do think they deserve IP, you know, they are using their art for which they would have normally gotten paid. So I think, you know, I think the same principles that we normally apply to patents or trademarks should apply here. I think the laws just have to catch up to that.
Clare Duffy
00:14:40
Yeah. I'm wondering if, based on your research, you think the worry that AI could put artists out of work is a valid one? Like, should artists be worried about their livelihoods at this juncture?
Sheena Iyengar
00:14:56
'I think so, I'm sure. So if you were the artist that only made portraits and you're not willing to change, it'll be a problem. So one of the examples that I use is the case of Picasso. He had two self-portraits. There's one self-portrait he did in 1901 and another self-portrait that he did 1907. There's a huge difference between those. And that's because he, during those ensuing years, is dabbling and trying to form a new style that will now have greater value given the advent of the camera. Now he gets influenced by impressionists, and then he mixes that with his with the influences of African art. And that's what leads him to eventually, starting in 1906 with La Demoiselle d'Avignon, he starts to create this new art form called Cubism. So sure, if he stayed with his style in 1901, he wasn't gonna go anywhere.
Clare Duffy
00:16:18
'This was something I was going to ask, and now I'm almost even more curious what you're going to answer based on our conversation thus far, but it was a question for me going into this, like should we even consider AI-generated art to be art in the first place or if there's something else that we should call it? What is your thought on that?
Sheena Iyengar
00:16:36
So, a human made it, in the sense that a human decided what it would be. And in the end, it's not AI that gets to decide if you like it or if you found it interesting or meaningful. That judgment still resides with humans. So I think, look, if you're willing to say that this thing that a camera took, you know, held in a human hand, is art because of the way in which the photographer took it. Same thing, I don't see why the principle is different.
Clare Duffy
00:17:26
'In your research, it sounds like at least in some of the cases, you labeled the pieces that you were showing, the study participants, human-made art or AI-made art. But of course, in the wild, it can be really difficult, increasingly so, to tell if a piece of art online has been AI-generated. Do you think that that labeling is important It's like going forward?
Sheena Iyengar
00:17:54
Absolutely.
Clare Duffy
00:17:54
Why?
Sheena Iyengar
00:17:54
Because the same reason why other things, you know, like masterpieces in the old days, you want to know whether it was like, it was just a replica or whether it was the original. Same thing with, you know jewelry or as we were talking about rugs, I don't know why I'm so fixated on rugs. I recently got a handmade rug. So it's about took three years to hand make this silk rock.
Clare Duffy
00:18:23
Wow.
Sheena Iyengar
00:18:25
And it absolutely, you can, I mean, I'm blind, but everyone tells me you can immediately tell that it's different from a machine made thing. Now, even if you couldn't tell, I think most people would feel cheated if they didn't know that it was handmade versus machine made. It just, it does have some greater value.
Clare Duffy
00:18:54
Well, and the idea that somebody put time, like time and life into something feels meaningful in some way at least. I'm curious too, I mean we've talked a lot about art that people will see on a screen, but I wonder if, given your findings, you think there is going to be a movement towards more people going to see art in person in museums or galleries.
Sheena Iyengar
00:19:20
I think you're already beginning to see that, right? They are, even if you go to the galleries, there's much more emphasis on tactile experiences, auditory experiences. I mean, even recently, if you went to the Biennale, right, there's a lot more of really trying to address other senses other than visual.
Clare Duffy
00:19:45
On that note, I'm curious, I mean, because you mentioned that you're blind, do you see a world where AI could make art more accessible to blind folks?
Sheena Iyengar
00:19:59
Yeah, I mean, you know, we we often think blind people don't have visuals in their mind. But actually, a large part of your visual life has very little to do with your sight. So I constantly live in a visual life, I am constantly trying to describe things in a way that other people will see it in my head. Make all kinds of visuals. So sure, nowadays you could have a blind person interact with AI and start...So here's a way you could make a collaboration between human and AI something very interesting, right? You could take a blind person and put AI in their hands and now suddenly that blind person could show the sighted what they see.
Clare Duffy
00:20:57
Hmm yeah...
Sheena Iyengar
00:20:58
That would create value.
Clare Duffy
00:20:59
Have you tried that?
Sheena Iyengar
00:21:01
No, but I just got the idea as I was talking to you.
Clare Duffy
00:21:05
I'm like, I want to do it now. That sounds really interesting.
Sheena Iyengar
00:21:11
Yeah, and you could even imagine not just any kind of person, like a deaf person.
Clare Duffy
00:21:16
Yeah, I mean, really, I can imagine anybody. We all have sort of inner lives and inner worlds that we think about in different ways. What do you think that we, as consumers of art, should consider next time we come across a piece of AI generated art.
Sheena Iyengar
00:21:34
Well, I think you should always ask yourself, what does this enable me to discover or see or learn that I didn't think about before? And that's true no matter what. I mean, you know, if you think of like Andy Warhol and, you know, Campbell Soup Cans, right? He gave us a new way to see those. Unto itself, you wouldn't say it was particularly, you know hard to do.
Clare Duffy
00:22:08
Right, that perspective.
Sheena Iyengar
00:22:09
Yeah, he just gave us a new to think about our lives, the world, Campbell Soup cans.
Clare Duffy
00:22:16
That's valuable. Yeah, that's interesting. Is there anything I didn't ask you that you think is important to mention about this?
Sheena Iyengar
00:22:24
I think there's always a new technology and when that new technology takes over our lives, we get worried. And that's not just true of art, it's true of anything, I mean the car, you know, now we forget how much disruption at the time the car created, or the train. But we as humans are so good at adapting such that we now give ourselves new jobs to do. And I think we shouldn't underappreciate that.
Clare Duffy
00:23:07
Yeah, it's so interesting. A big part of the promise of AI, broadly, is like it will take up some of the sort of busy work tasks that people don't want to do and leave more time for things like creativity and human expression.
Sheena Iyengar
00:23:24
I doubt that'll be true. I think it'll just mean that you're going to do a lot more in the same amount of time. I mean, did the car really, like how did the car change our lives? Well, you now could own a house because you could live further away. Nice. Did it actually reduce the amount of time you work? No.
Clare Duffy
00:23:47
Yeah.
Sheena Iyengar
00:23:50
Not a bigger house.
Clare Duffy
00:23:52
The expectation is that you can go farther for work.
Sheena Iyengar
00:23:53
Same thing happened with Zoom.
Clare Duffy
00:23:55
Yeah.
Sheena Iyengar
00:23:56
You're still working. If anything, you might even be working longer hours. They made it easier for you to get out of bed and get to work.
Clare Duffy
00:24:03
Yeah, it all turns in that direction, unfortunately. Well, Sheena, thank you so much for doing this. Really appreciate it.
Sheena Iyengar
00:24:11
Thank you.
Clare Duffy
00:24:14
'So if you're concerned about how AI-generated art is going to impact human artists or human expression, here are a few things to keep in mind based on my conversation with Sheena. According to Sheena's study, people still tend to value art that's fully made by humans more than AI-generated art, even though it can sometimes be hard to tell the two apart. So it seems like human creativity still carries that extra spark. We'll link to Sheena and her team's research in the show notes. Remember that even AI-generated art typically has some kind of human intention behind it, even if it's something as simple as a text prompt. As far as honoring artists' work and making it clear when an AI tool is drawing from existing art, that's an ongoing conversation in boardrooms and courtrooms right now. And if you want to support human artists, consider buying their art or going to see it in person, in a gallery or museum. Thanks for tuning in to today's episode of Terms of Service. I'm Clare Duffy, catch you next week.