A joke-telling robot might not sound very useful, but there’s a lot more to an automaton’s standup routine than making people laugh
Author Cristina Rouvalis Illustration Jon Reinfurt
Heather knight smiles at her robot like a proud mother, gently hoisting it up until it stands on its two clunky feet. “They call me Data the robot,” says the 22-inch-tall, white plastic automaton with a wide-eyed, startled expression. “It makes me feel like some kind of superhero, but actually I am just a mediocre robotic comedian.”
If Data’s opening line doesn’t cut you up, just wait. It has more jokes up its carbon-composite sleeve, calibrated to change based on audience reaction. And there is more to all this than generating a few laughs.
Standing atop a table in a Pittsburgh café, the robot says it is learning social skills so it can help people in their homes. “Just remember: We will never steal your beer or liquor,” it says. “One day, we will be your companions.” Knight, a doctoral student at Carnegie Mellon and founder of Marilyn Monrobot, a robotic theater company, says her aim is to “make machines with charisma.” Data’s standup, she explains, is part of the growing field of “human/robot interaction,” which seeks to make machines more personable as they take on tasks ranging from working in hospitals to providing customer service.
Automated wisecracks may seem like a roundabout way of approaching such a future, but—as anyone who has encountered the inspired goofballs at the MIT Media Lab knows—innovation often arises from play. “What seems aimless can be very productive,” says Karen Wilkinson, director of San Francisco’s Tinkering Studio. “It’s a playful attitude about exploring the environment.”
Data works by “listening” to how much laughter a joke generates, then uses an algorithm similar to that used by Netflix to gauge overall responses to things like topic, duration and bawdiness. If Data finds that, say, long political jokes are bombing, it’ll throw in a few silly one-liners.
The ability to “read” people will become increasingly important as machines and humans work more closely together—but the relationship goes both ways. “Most people don’t know how to act around a machine,” says David Robert, a doctoral student at the Media Lab who designs robots for children. “They are guarded, taken aback. Heather is breaking the ice.”
Knight first became interested in robots while studying at MIT—she earned her bachelor’s and master’s degrees there—and afterward took a job at Aldebaran, maker of the Nao robot. Upon leaving the company in 2010, she was supposed to turn in her loaner bot, but found that she just couldn’t give it back. She’d saved some money to buy a car but bought Data instead.
That same year—despite the fact that her showbiz experience amounted to playing a flower in a middle school performance of The Wizard of Oz—Knight debuted what was billed as the first interactive robotic comedy routine during a TEDWomen conference. “Don’t be shy,” she told the audience. “It doesn’t have feelings—yet.”
Today, Knight is working on developing Data’s offstage personality. She recently put the robot to work as a campus guide for prospective Carnegie Mellon students, and watched how they responded to various artificial personalities. “Sensible” drew stifled yawns; “super-peppy” fared better. A third persona, which channeled a sulky teen, had Data muttering, “Oh, man, another stupid robot thing. I don’t want to be here.” Not surprisingly, the kids loved this one, while their parents did not.
The idea is that, as Data’s analytical skills grow more sophisticated, it will be able to judge subtle mood shifts in real time, thus helping it avoid the kinds of faux pas that could ruin a human-machine relationship. After all, a hospital bot would be expected to take a very different approach when talking to, say, an expectant mother, as opposed to a person with a horrible illness.
It’s a significant leap, of course, from tailoring your responses to fit a particular situation to making meaningful connections between robots and people, but Knight is confident that such a leap is possible. “That’s a natural extension,” she says, “catering it to your personality.” This opinion is based largely on Knight’s relationship with Data, which—on her end, at least—is an affectionate one.
Knight travels a lot these days, mostly to give talks on socially adept robots, and Data almost always tags along. Knight enjoys the work, but not so much when her wisecracking sidekick isn’t beside her. “When I have to give a presentation without Data,” she says, “it’s like, ‘Oh, no!’”
Cristina Rouvalis, a Pittsburgh-based writer, very much regrets telling the joke “What’s brown and sticky? A stick!” at a recent formal function.