“Usually, when people think of an artist as an important creator, what they’re saying is that he is capable of continuously surprising us with new, good work,” explains Pablo Gervás.
Does that same expectation hold when the artist is a series of algorithms? Gervás is a researcher at the Universidad Complutense de Madrid, where he takes a rather unconventional approach to studying short stories and poetry: he programs computers to write them.
Computational creativity is an area of artificial intelligence based on developing creative machines. These systems can produce paintings, compose music, make scientific discoveries, and write short stories and poetry. In fact, they’re actually pretty good at it.
But developing software to create art raises a good deal of questions. Are programmers modelling the same kind of creativity that humans demonstrate? Should we consider the products of these machine creators art? If that’s the case, who is the artist — programmer, or programmed?
Berys Gaut, a professor of philosophy and aesthetics at the University of St. Andrews, proposes that artificial intelligence is simply a new step in the evolution of artistic practice. “I think that computers are some of the newest tools that artists invented to make art. They’re the next step beyond the paintbrush, in many cases. But you wouldn’t want to say a paintbrush is creative. Rather, it’s the person who is creative, who uses the paintbrush.”
Enrica Piccardo, a professor at the Ontario Institute for Studies in Education, echoes these thoughts. “The technical part is bigger, but the mind behind them is still human.”
Yet as technology continues to progress, the line between programmer and programmed is becoming increasingly blurred. One approach to creative systems is a technique called genetic algorithms. Inspired by biological processes, these programs are allowed to “mutate” on their own, in order to evolve toward better solutions to a problem.
James Moor, a philosopher of artificial intelligence at Dartmouth College and editor of the journal Minds and Machines, explains the process. “You have something you’re trying to create: you have a standard about what it would be like to have that. Then you have a bunch of programs that try to figure out how to do it. Some of them do it well, and some don’t do it so well. And then you have a process of natural selection, and you may allow the programs to mutate, to change slightly, and then you pick out the best programs and let them do it some more.
“Eventually in some cases, through this natural selection among programs, you end up with something that’s quite novel and non-obvious — and valuable.”
The issue of value is critical to defining both creativity and art. In computational literature, the most common way of evaluating whether a system is creative is to look at what it produces: is the product original and worthwhile?
However, acknowledging the products of machine creators as valuable leads us into trickier territory; it brings us to the heart of how we actually define art.
David Moos, the curator of Contemporary Art at the Art Gallery of Ontario, provides some context to the question of what constitutes art and how it relates to the growing field of computational creativity. The definition of art has undergone a steep evolution over the past century. According to Moos, Marcel Duchamp first began the discourse of breaking down traditional definitions in the 1910s with works that he called “assisted readymades.” These were ordinary objects which he found and raised to the status of artwork simply by branding them as art.
With these initial barriers broken at the beginning of the twentieth century, artists have had the space to use an increasing range of tools and technology to produce their artwork. In 1990, German contemporary artist Rosemary Trockel created a painting machine — a mechanical device with fifty-six brushes that make different kinds of marks on a roll of paper.
Technology in art is even more obvious in the digital realm. According to Moos, Andreas Gursky is the current master of digital large-format photography. However, Gursky’s massive photographs of landscapes and architecture would not always have passed for art. “By a 1960s definition of art, that could not be art,” says Moos. “That would be Madison Avenue advertising. But it’s definitely art today.”
Nowadays, the machines at an artist’s disposal are becoming more and more complex — at times, even lifelike. It seems the next plausible step in the evolution of art might involve technology that is increasingly independent of the humans who use it. There’s something inherently thrilling in using a system that seems to have a mind of its own. However, once we cross that threshold, it becomes difficult to distinguish where the human begins and where the machine takes over.
This is the issue of creative agency, explains Jon McCormack, who doubles as an electronic media artist and researcher in computational creativity at Monash University in Australia. “Lots of people can use computers to make art, but there’s no doubt that the creative agency behind the task is a person,” says McCormack. “But in some cases, people have tried to shift that agency more and more to the computer. So the creative responsibility is more to the computer than to the person. That’s where I think it starts to get quite complex, quite ambiguous as to where the art is actually coming from, and who is actually responsible for producing the artwork.”
Some feel uncomfortable attributing creativity to something that isn’t human. Calling what a machine does “creative” doesn’t seem to do justice to the mystery of the human creative mind. “It takes the alchemy out of it,” says Toronto-based musician Emilie Mover. “I grew up with bebop jazz. To me, the idea is that people get together in a room, and strum or blow into an instrument in their hands, close their eyes, and they don’t have to think about it anymore.
Whatever comes out of their spirit goes into the atmosphere.”
Subrata Dasgupta, a cognitive scientist at the University of Louisiana, Lafayette, suggests, “I think people are uncomfortable [with computational art] for the various reasons that, over history, people have been uncomfortable when the uniqueness of the human being has been challenged.”
The fact is, we still see art as something that’s fundamentally human. Moos explains, “Our understanding of that word, that idea, that concept, goes back to the beginning of the history of visual art — which is, of course, in caves. The markings on the caves in Europe indicate humans using images to narrate a certain essence of their lives. And so deep down, fundamentally, that’s what art is.”
Once you start giving the human a smaller role in the art-creating equation, our definitions once again begin to break down. One thing to consider, however, is that technology is not only changing our definition of art; it is now also changing what it means to be human. As machines have become increasingly lifelike, humans in turn are incorporating more and more machinery into our biology. Cyborg culture is no longer a sphere reserved for science fiction. As Moos explains, even medical advances such as artificial organs and prosthetics challenge the idea that humans are autonomous. If our own bodies are no longer purely human, why, then, should we require art to be so? “We’re slowly getting to a point where the programs can compete with humans,” says Gervás.
But while machines might be able to create the same kind of product as a human, their brand of creativity is still very much distinct from ours. McCormack suggests, “I think that certainly what’s currently possible is a long way from the kind of creative ability that you see in humans. There’s no computer program currently that can display even a modicum of the kind of creativity that we see in human beings.”
Part of that problem lies in scientists’ inability to hone in on what human creativity actually means. Moos explains, “The trouble with the creative process is that we know so little about how we do it. So it’s very difficult to evaluate it. It’s all kind of a mystery.”
Dasgupta suggests that a large part of creativity relies on processes at work in the unconscious state. Original ideas don’t come out of nowhere: we are influenced by past events, seemingly insignificant information, and things we aren’t even aware of. “My own work has been mostly on doing historical and field studies of creativity — that is, studying people in the real world, as they’re actually being creative. It’s astounding to see what kinds of situations arise that lead to creativity: their past histories, their cultural background, chance remarks passed by someone, chance observations of things. These kinds of issues, to my knowledge, are hardly ever replicated in laboratory studies.”
In fact, the notion of creativity itself is a relatively recent one. According to Gervás, “Creativity, as a word, as a concept, exists only from the nineteenth century or so. It’s not one of those things that you can trace back to the Greeks or Romans, because it’s quite a complex abstraction. There’s this idea that there is something in common between the process that people apply to reach goals in the arts, in music, in literature, in scientific discovery, in engineering, in design — which is what we call ‘creativity.’”
It seems that many of the roadblocks faced in creativity research come from attempting to define creativity itself. “You have to just recognise that all creative processes and creative products have characteristics that make them creative — but not necessarily always the same characteristics,” explains computer scientist Tony Veale. “It’s about trying to model something on a computer that people will agree is creative, without having to agree on what the definition of creativity is.”
Gervás continues, “At one stage, people realized that it’s not so important what creativity is — which is what had happened with artificial intelligence a long time ago. It’s not that common anymore to have people trying to define what intelligence is: nobody really knows. But people do build programs that everybody agrees are artificial intelligence.
“Computational creativity is reaching that stage now. Let’s not bother so much about what creativity is. Let’s just go and do it. Let’s get programs that can do things that, if people did them, we would consider them to be creative.”
According to Veale, computer science is the perfect discipline for studying creativity. “Because if you don’t have a definition, in most disciplines, you don’t know what you’re studying. But in computer science, you have the ability to build things. I mean, we still argue about definitions and concepts. But the main thrust these days is in building systems, and trying to agree or disagree why they show some form of creativity, or the seeds of creativity at the current stage at least.”
It’s unrealistic to think that machines will ever replace human artists, and that’s not what computational creativity is aiming for anyway. “I don’t think that a computer of any sort will ever be able to sing like Al Green,” says Mover. “And I think that still, no matter what happens in society, if you play a robot-record and an Al Green record, I think the majority of people statistically would choose to listen to the Al Green record for the rest of their life, if they had to choose.”
Instead, the value of creative machines will come from what they can do that humans can’t do on their own. We hope to go beyond the human.
“Let’s build something that can help a human artist do work,” proposes Gervás. “You need some kind of augmentation to be happening. You need the human artist to be able to produce something he wouldn’t have produced without it.”
Veale concludes, “Of course, that shows why computers will be useful. […] We want them to help us think in novel ways, and not just reinforce our thought patterns — we can do that for ourselves — but show us the possibilities that we didn’t recognise.
“What we really want is computers that will surprise us.”