èƵ

But is it (robot) art?

Give a robot a paintbrush and look what happens

But is it (robot) art?
(Image: Napolitos 3)
Max Chandler's robots are designed to take mathematically derived patterns and then use them to produce a painting
Max Chandler’s robots are designed to take mathematically derived patterns and then use them to produce a painting
Max Chandler gets along just fine with his robot helpers when they are behaving properly
Max Chandler gets along just fine with his robot helpers when they are behaving properly
(Image: Chris Hardy/San Francisco Chronicle)

Max Chandler art studio is more chaotic than most. It’s not that he is disorganised: the mess is down to the team of brush-wielding robots that help him paint his pictures. Built from Lego bricks and aluminium, the bots look like miniature trucks and cranes. Some propel themselves on wheels, while others shuffle about on legs. All are capable of exasperating behaviour that sometimes turn peaceful painting sessions into circus-like mayhem.

If the bots are not tracking paint onto the neighbour’s porch, they’re falling clean off the canvas, ruining the flow of the picture. Worse, they always need refills at crucial points in their trajectories, forcing Chandler to scramble to insert a new paint-loaded brush. “Sometimes I think this is the worst way in the world to create a picture,” he chuckles.

“Sometimes I think this is the worst way in the world to create a picture”

So why go to all this trouble? Although the bots make life more difficult, Chandler says working with them is the only way he can create the pictures he wants. Chandler, who splits his time between Scottsdale, Arizona, and San Francisco, California, is part of a burgeoning community of artists and engineers who use robots to produce paintings. “It’s a new form of art,” says Sherry Turkle, a sociologist and psychologist who studies the impact of technology on culture at the Massachusetts Institute of Technology. “It can be partly shaped by human intentionality and partly be an elaborate dance, a negotiation with a robot.”

Robotic art dates back to 1973, when Harold Cohen, a British artist who is now professor emeritus of visual arts at the University of California, San Diego, developed a computer program called AARON that composed portraits of people and objects without human help. In the late 1980s, artist and engineer Ken Goldberg, now at the University of California, Berkeley, used an industrial robot arm to paint pictures; the arm later became part of an exhibit in which it drew maps of the world.

While some artists and researchers use robots to explore the meaning of art and whether a machine can truly be creative, Chandler’s goal is simply to create the look that he wants. “The robot in his case is a tool,” says Goldberg. The fact that Chandler executes his own vision is what matters to his peers, says Deborah Harris, an artist based in Phoenix, Arizona. “Intentions are important. I think art is an opportunity for one human being to create something that other human beings relate to.”

Chandler’s intention has always been to infuse his art with mathematics. An avid painter since childhood, he has harboured a love of maths for almost as long: at nine, he was already writing computer programs. He went on to study maths at MIT and has paid the bills working as a programmer. “I have always had this dual nature of being tied up in science and in art,” he says.

By the mid-1990s, he had begun thinking about how to add mathematical shapes to his paintings. One figure that intrigued him was the logarithmic spiral, which appears naturally in the shapes of nautilus shells and spiral galaxies. In a logarithmic spiral, the curves are coiled tightly at the centre and gradually spread apart at the edges. Because the exact rate at which the curves separate is governed by a mathematical formula, humans have difficulty drawing them freehand (see Diagram).

Chandler was also fascinated by computer programs that could simulate the geometry of plants and other living things using simple mathematical rules. These “cellular automata” could mimic the processes of plant growth and generate unique and complex shapes that, to Chandler at least, reflected the living world in a more striking way than anything a human could draw. Simply rendering these patterns on a computer and printing them out, however, would create boring and repetitive pictures.

When hobbyist robot kits such as Lego Mindstorms appeared in the late 1990s, Chandler had his solution. To get the shapes he wanted, he could program small robots to drag a paint-loaded brush behind them as they moved. Unlike a computer-generated image, the shapes would come alive in brush strokes and would vary because of random fluctuations such as wheels slipping or robotic legs wearing down. “One of the characteristics of programming robots rather than a computer is that they are interacting with the physical world,” he says.

His first art robot was a wheeled Lego-bot called Zeb1 that was capable of holding a pen. Three more wheeled robots followed: these hold paintbrushes in a clamp and drag them as they roll along. The Zeb series creates smooth lines, sometimes using brushes of different thicknesses. However, Chandler found that much of the time the Zebs’ strokes were, well, just too robotic.

To create more expressive strokes, Chandler built a second line of bots called Gimpy, this time with legs rather than wheels. The first of these, built in 1999, has two L-shaped plastic legs that stick out of the sides of the main body, spanning about 30 centimetres. Gimpy1 moves a bit like a wind-up walking toy and drags a brush. As it lifts one leg, the brush rises, so Gimpy1 applies paint more lightly, and as the foot comes down, the brush presses into the canvas. The result is a line that isn’t smooth and uniform, but is composed of individual strokes produced as the legs rise and fall.

Getting the bots to create varied brush strokes, however, was only half the battle. They also needed a sense of direction so they could decide where to go. At first Chandler equipped them with light sensors that allowed them to follow a coloured line or a template. This worked well for some effects and allowed him to weave his own sketches into what the robots were painting. But he also wanted the bots to be able to follow the patterns generated by his cellular-automata and logarithmic-spiral programs.

To paint a cactus, for example, Chandler first runs a cellular-automata program on a computer. This defines the approximate shape and dimensions of the plant, such as the number of pads on each branch, the number of branches and how close each branch is to its neighbours.

He gives this information to the robot’s own computer, which calculates how many steps and turns the robot has to make to trace out the figure. In most cases, the bot is programmed to draw the outline of the cactus using fragments of a logarithmic spiral, giving the representation a unique look (see Diagram).

Chandler knows that his paintings run the risk of looking machine-like. To create spontaneity, he plays tricks on the bots. The software will randomly change how the robot perceives its size or speed. As the bot relies on this information to orient itself, it may think it has reached the point where there should be a bend in a branch, for instance, when in reality it has over- or undershot the mark.

“Most people’s impression of a robot-generated thing is that it is repetitive, mechanical,” he says. “I work very hard to make sure that is not true.” That work seems to be paying off. Chandler used to discard 80 per cent of his robotic drawings, whereas now he throws away only 45 per cent.

“Most people’s impression of robot-generated art is that it is repetitive and mechanical. That need not be true”

A recent innovation came when he painted a sculpted guitar for a festival in Phoenix, Arizona. Though perfect for drawing a cactus on the back of the guitar, which was flat, Gimpy1 and even its nimbler cousin Gimpy2 weren’t up to the job on the front, which was bumpy and sloped. The bots slipped on the surface and tripped on the 10-centimetre-tall knobs. So Chandler built Gimpy3, an insect-like bot with six legs that can crawl up sloping surfaces and over bumps without tripping. “Gimpy3 is really tough,” says Chandler. “Although he tilted crazily on the knobs, he didn’t fall over.”

How did you do that?

You might wonder if Chandler feels the end result is worth all the effort. His answer is a resounding yes. “It wouldn’t be the same if it was just the robot or just me,” he says. “Even if a viewer doesn’t know why it is, the picture will look different.” Whether the bots create something better than a human could is subjective, of course, but the pictures are unique. “People say, ‘Wow, those lines are special – how did you do that?'” he says.

Still, some people might find the idea of robotic art disconcerting, says Douglas Irving Repetto, curator of the annual ArtBots exhibition in New York and an artist at the city’s Columbia University. He thinks it will continue to be accepted by galleries and museums, though. “There are always debates about what is serious art,” he says. “For every place that might have a problem, there is another place that would think it is terrific.”

Chandler is counting on this as he steams ahead with his latest innovation: a camera on stilts that captures paintings in progress, detects when the robots go too far off course, and wirelessly resets their positions. He hopes this will allow him to program the bots with more complex instructions, as well as coordinate the actions of multiple bots. At least it should make them less likely to go wandering off the edge of the canvas, something that Chandler and his fans would surely welcome.

Look, no human

While some artists retain tight control over their robotic helpers, others are interested in precisely the opposite: what happens when you leave a robot to its own devices?

“I am trying to build machines that are really making art by themselves,” says artist Leonel Moura of Lisbon, Portugal. “It changes our idea of what is art and what is culture.” Moura and others see art-making bots as an opportunity to explore the ideas of creativity and authorship, and to make people question whether art can only be created by humans.

Moura’s Robotic Action Painter (RAP), built in collaboration with IdMind, a Portuguese robotics company, will be displayed in action at the American Museum of Natural History in New York this spring. The small bot will work unsupervised by humans, painting pictures round the clock, deciding when they are done and even signing them. “When you see the robot at work, when you see the way he makes the painting, it’s very curious, very attractive,” Moura says.

Given a blank canvas, the RAP starts by making a sketch using six coloured pens: it moves around on wheels, selects colours and chooses whether to make a spot, a long line or a short line according to a random algorithm. It stops the sketch when the concentration of colour is above a certain threshold, which it detects using an on-board light sensor. Then it reinforces the colours, finally finishing when the colour saturation is above another threshold.

Digital-media artist Eva Sutton of the Rhode Island School of Design sees a future in which robots are capable of having intentions, something that most people would consider a crucial characteristic of an artist.

“Perhaps in the not-too-distant future, robots might become so sophisticated that they sense enough of the world to formulate some desire or motivation,” she says. Then robotic art will get really interesting.