The next artistic masterpiece may not emerge from the hand of a human, but from the electrical circuitry of a computer. But can we ever really connect with such artificially-inspired creativity? And should we even want to? Kevin Berger explores these questions in this curated piece, originally published by Nautilus.
- Like artists, machines see the world in a different way.
- It’s already possible to teach computers to be creative.
- You can say that art basically amounts to representations with concepts. AI art is capable of that.
- One day, AI-fueled art will surpass the wildest dreams of human artists, according to Arthur I. Miller, a University College London emeritus professor and specialist on the connections between science and art.
I’m trying to explain to Arthur I. Miller why artworks generated by computers don’t quite do it for me. There’s no human being behind them. The works aren’t a portal into another person’s mind, where you can wander in a warren of intention, emotion, and perception, feeling life being shaped into form. What’s more, it often seems, people just ain’t no good, so it’s transcendent to be reminded they can be. Art is one of the few human creations that can do that. Machine art never can because it’s not, well, human. No matter how engaging the songs or poems that a computer generates may be, they ultimately feel empty. They lack the electricity of the human body, the hum of human consciousness, the connection with another person. Miller, a longtime professor, a gentleman intellect, dressed in casual black, is listening patiently, letting me have my say. But I can tell he’s thinking, “This guy’s living in the past.”
Miller is sitting at a simple table in a dim and sparsely furnished apartment on New York City’s Lower East Side. It’s an Airbnb place that’s keeping him housed while he gives talks in bookstores and colleges in the city about his latest book, The Artist in the Machine: The World of AI-Powered Creativity. Miller is the Virgil of art and science writing, a guide through the underworld of artists employing scientific practices like gene-splicing, brain imaging, and computer-code writing to create works that ask viewers to reflect on how science and technology are changing our views of the world and everything in it, including us. His previous book, Colliding Worlds, features artists like Austrian sculptor Julian Voss-Andreae, who studied quantum physics. One Voss-Andreae work, Quantum Man, stands eight feet tall and is constructed of more than 100 vertical steel sheets. The sculpture looks like a man when seen from the front, but as viewers move around it, the figure vanishes, representing how experiments in quantum physics, depending how they’re set up, detect electrons as either waves or particles: “how you look at it, that’s what it is,” writes Miller.
Miller has a Ph.D. in physics from MIT and is an emeritus professor in history and the philosophy of science at University College London. The Artist in the Machine profiles an array of fascinating artists and engineers who write computer programs to generate music, paintings, and literature. (Miller’s profile of Ross Goodwin, who’s written algorithms to generate screenplays and a novel, is featured on Nautilus.) Miller argues that AI-fueled art gains independence from its algorithmic parents and takes flight in works that bear the hallmarks of creativity and genius and will one day exceed human artists’ wildest imaginative dreams. Miller says he sympathizes with what I’m saying about the power of art coming from the connection with a human artist, plumbing their emotions and consciousness. But I’m being premature. Just wait, he says, computers will one day produce art as transcendent as the works of Beethoven and Picasso were in their times.
“Machines in art can have intent, and they can have a bit of free will, too”
So you’re saying we’ll one day connect with machine art as profoundly as we do now with human art?
Yes. The machine sees the world in a different way than we see the world. Just like an artist does. That gives you an inkling that machines will have a different physiology. In time, they will evolve emotions. Just from scanning the web now, they could imitate our emotions. They’ll say, “Oh, thirst, that’s cool. I think I’ll be thirsty,” and they can convince you they’re thirsty. “Love, that sounds cool too, I just had this nice discussion with a machine down the street, and it seems like love.” They’ll hone their notion of love by reading novels, and soon they will evolve emotions and consciousness. That will be the point of artificial general intelligence. Then it’s just a hop, skip, and a jump to artificial superintelligence, where they go beyond us in intelligence, emotions, and consciousness.
Does machine superintelligence worry you?
No. I think if we teach machines to be creative, then they’ll be beneficent toward us, rather than just keeping us around as household pets.
That’s good to hear. But back to art.
I think we should keep your question in mind: Can we appreciate art that we know has been produced by machines?
“We can have hopes, dreams, and aspirations for humankind. But we are machines, just like computers are machines”
We could learn to, yes. We may develop a preference for it. Just as we may eventually have a preference for the prose produced by sophisticated machines, which at first may be nonsense to us. Now machines generate prose with word play that we’re not used to. That shows us machines can change the landscape of language. It’s not unlike what happened in the 1840s, when the camera was invented, which freed artists from a too literal interpretation of nature, and opened the doors for the impressionists.
Still, I don’t see how we can feel the agency of machine art as we do with human art.
Machines in art can have intent, and they can have a bit of free will, too. You can take a painting-robot out of a studio, out on the road, with a webcam attached to it, and it will look around and see something that strikes its fancy as a nice pattern, and say, “I think I’ll paint that.” So that’s intent. That’s a bit of free will, too. So machines can be out there in the world, and have some sort of physical experience.
Are you saying we don’t need the human element to appreciate art?
Look what’s going on today with people sitting in front of their screens. Where’s the human element? People are walking down the street looking at their phones. They’re getting hit by cars.
I’ve interviewed neuroscientists who’ve said human connection activates brain networks that move us profoundly, perhaps more than practically anything else.
Well, you have to leave your mind open. That may well change.
Has researching the connections between science and art given you a more mechanical view of humans?
Yes, it has. In doing this work, I have developed a reductionist view of nature. There is a definite relationship between us and machines. We’re like machines; machines are like us. We are basically biological machines.
What does it mean to be a biological machine?
It means we’re made up of atoms and molecules. They obey laws of nature and evolved to produce us. They are the power behind our engines. We’re just big chemical reactions. Each part of a machine is made by means of the good old laws of Newtonian physics, which were deterministic. But when you put this conglomeration together, it’s capable of unpredictable or chaotic behavior. Unpredictability is one of the hallmarks of creativity. So right from the word go, machines can be creative. Since we are machines, we can just go out of existence like a machine. We can have consciousness, but when the electrons stop flowing, consciousness disappears and that’s it. But that doesn’t take away from our creativity. We can still be creative in an inspiring way. We can have hopes, dreams, and aspirations for humankind. But we are machines, just like computers are machines. When one says that to people, some get very upset. They ask me, Is there a soul in a machine? I say, you have to believe in a soul, and I don’t believe in a soul. It isn’t necessary.
Does it make you queasy to think of yourself as a machine?
No, not at all.
Why should it?
Because being a machine implies we’re designed.
We are designed. Just not by a designer. We’re a big mistake. It just so happens a number of billions of years ago that there was an isotope of carbon in the sun that was spit out, landed on the Earth, and so we’re carbon-based. This accident could occur, probably did occur, somewhere else in the universe.
Right. But the biology that makes us what and who we are has evolved over billions of years. That evolution makes us very different from machines.
Well, machines are evolving too, and much, much faster than us.
But they were engendered by humans.
We were engendered by nature. Yes, a human made the program, and a human put the machine in operation. But that’s like saying Leopold Mozart taught Wolfgang the rules of music. But we don’t attribute Wolfgang’s creations to his father. Similarly, you don’t attribute AlphaGo’s success at Go to the AlphaGo team. Or if you teach your 4-year-old daughter how to draw, she will draw like you at first. But 20 years down the line, when she’s at art school, she’ll be drawing much differently.
“We should not judge products of AI on the basis of whether they can be distinguished from products made by us. Because what’s the point? They may produce something that we can’t imagine right now”
It sounds like you’re saying there’s basically no difference between machine intelligence and human intelligence.
The bottom line is there is no difference. We are computable and machines are of course computable, so there’s no reason why we’re not like machines. Consciousness is computational. It’s computed from 100 billion neurons in our mind. Since machines use computation, they’re computational, and there’s no reason why they cannot have consciousness.
You said you’ve become a reductionist. What got reduced in your mind?
What got reduced in my mind is our body, us. You could say to a reductionist that there are things that science can’t explain. But science can and will explain everything we see around us.
Let’s step back. What is art?
Art is representations with concepts. That’s what it is to me. AI art does that. AI art has concepts because it’s generated by scientific means. It goes beyond the science, but it has a scientific basis to it, and so it has concepts. I have my own definition of aesthetics, too.
Lay it on me.
It makes art historians’ hair stand on end. Aesthetics equals the image in a work of art—and the image need not be a visual image—but an image from our five senses, plus the apparatus that generates it. For example, at CERN, I asked a physicist for his definition of aesthetics, and he said, “My notion of aesthetics is nicely laid out wires.” I saw them. There are these units of parallel wires, nothing crossed over, and the wires are color coded. It was beautiful.
I like what Simon Colton, a professor of digital game technologies, tells you. One of his projects, The Painting Fool, is great. He fed Guardian news articles to a computer and had software analyze words and phrases for moods like happy or sad. The computer then created portraits of people, which varied, according to its mood. Colton says comparing AI to human art is irrelevant. We should “be loud and proud about the AI processes leading to their generation,” he says. “People can then enjoy that these have been created by a computer.” I agree that’s the way to appreciate AI art, as its own beast.
Me too. We should not judge products of AI on the basis of whether they can be distinguished from products made by us. Because what’s the point? They may produce something that we can’t imagine right now, that could look like nonsense, but may then turn out to be better than what we could produce.
Why is the intersection of art and science, what you call a third culture, important?
Because it investigates the world as it is on a much better basis than somebody standing in front of an easel with oil paints. We’re looking at the world of the future. We’re looking at a world where we are merging with machines. A world where machines will be of importance. We will eventually be collaborating with them. They will produce works for us to enjoy and for their brethren to enjoy. So this is the way the world is evolving, and this is why we should be aware of AI art, and try to understand. Try to understand that it’s not stealing things from us.
Finally, what has your research into this third culture taught you about people today?
That they should widen their point of view. Too many artists and too many scientists have a narrow point of view. Scientists think art has nothing to do with science and artists think science has little to do with art. It is easier for a scientist to understand art than an artist to understand science. But, in a conceptual sense, art is an attempt to understand the concepts of science. I think my research has changed my view of humanity in the sense that people should leave their minds open to what goes on in fields other than their own. It could widen and even change their beliefs.
Kevin Berger is the editor of Nautilus.
This article first appeared on Nautilus on 5 December, 2019.
Nautilus is a different kind of science magazine. Online, in print, and in the classroom, Nautilus leverages deep, undiluted, narrative storytelling to bring science into the largest and most important conversations we are having today. After all, that is where modern science – which is so personal, pervasive, spiritual, and transformative – deserves to be.