advertisement
advertisement

A glimpse into the uncanny world of robot-made fine art

Here’s why art–and paintings in particular–are a surprisingly potent way to gauge the future of machine intelligence.

“I don’t know if you watch Westworld,” Andrew Conru says, “but the artistic process may be a very simple algorithm.”

advertisement
advertisement

Conru is the founder of Robotart, a global contest for artwork created by AI and robots. He’s referring to the big reveal [SPOILER ALERT] in the second season of the HBO series: After looking for a way to simulate human consciousness, AI researchers realize that under humanity’s apparent complexity lies a very simple set of rules coded in our genetics. This set of rules, or algorithms, control our basic behavior from birth, and therefore they can be easily simulated with code. We are extremely predictable and simple in our apparent complexity, not only according to Westworld writers but also according to actual scientists.

So it’s reasonable to argue that the artistic process is also a set of pretty simple algorithms that can have complex outcomes. The way artists look at and reinterpret the world, along with the techniques they work with, result in a particular aesthetic–impressionism, say, or abstraction. “Of course, part of it is the artist’s physiology,” Conru says, “the way the hand, the arm, the eyes, and the brain interact to introduce an element of unique randomness” in each brushstroke. But beneath it all, an artist’s algorithms–their methods, their processes–are not that complicated, though they may cause powerful emotions in viewers.

Could machines ever create art that moves us? Conru–who holds a PhD from Stanford in mechanical engineering and, incidentally, founded FriendFinder Networks, home of Adult FriendFinder–thinks so. Robotart is a benchmark for gauging how close we are to that goal.

First-place winner, CloudPainter. [Image: courtesy Andrew Conru/Robotart]

Robotart, now in its third year, is judged both by a public voting system and a panel of art critics, including New York magazine’s Jerry Saltz. The critics weigh three criteria: originality and aesthetics; painting techniques like layering, energy, subtleness, or blending of brushstrokes; and the art’s technical contribution to the field.

Some of the 100 artworks submitted by 19 teams this year were extremely convincing. As Conru points out, some human artists are taking advantage of robot “apprentices” to aid their output, using the robot to do the heavy lifting before finishing it the same way many artists use studio assistants for these tasks today. But perhaps more importantly, the entries suggest that we’re getting much closer to the moment when humans won’t be able to tell the difference between machine-generated art and human-generated art.

Take this year’s winner, CloudPainter. CloudPainter is the project of generative artist Pindar Van Arman. He describes his role as “designing creative algorithms,” which control an industrial robotic arm equipped with brushes and powerful artificial intelligence. The results can fool a human: In a recent interview, Saltz pointed out at one of his works and said “that doesn’t look like a computer made it.”

advertisement

But Saltz followed that apparent compliment with a major caveat: “That doesn’t make it any good.”

That’s the crucial next step, Conru says–the point when machine art can move not only Saltz, but any viewer, the same way that the Blacks Paintings by Goya or Monet’s waterlilies can move us; a “sort of Turing test for art,” as he puts it. When will an AI-composed symphony move us in the same way Beethoven’s Ninth does? Conru believes that we’re close to fooling humans to the point of moving them, for instance by reproducing the styles of canonical painters. But that kind of technology still relies on imitating what people already do, emulating human sensibilities acquired through deep learning. We’re light-years away from the sci-fi vision of AI that has the kind of consciousness upon which great art is made, and Conru thinks that no matter how intelligent machines become, human viewers will always respect human artists more than machine ones, the same way human grand masters elicit more admiration than computer grand masters.

Still, there’s room for plenty of debate–and imagination–when it comes to the future of AI, and art is a way to discuss those issues. If machines ever gain sentience, would there really be any difference between the art humans create and the art those new life forms create? Perhaps these new beings could push our human practice of art, which is so heavily based on iteration, into an entirely original new phase and let us see the world through eyes that are not human, interpreted by brains that are not human. Regardless of the quality of the art, I would love to experience that.

advertisement
advertisement

About the author

Jesus Diaz founded the new Sploid for Gawker Media after seven years working at Gizmodo, where he helmed the lost-in-a-bar iPhone 4 story. He's a creative director, screenwriter, and producer at The Magic Sauce and a contributing writer at Fast Company.

More