With rosy cheeks and short-clipped bangs framing her round face, the toddler looks like many other three-year-old girls. But even as a baby, she was clearly different. Her parents first grew concerned when she wouldn't make eye contact as an infant or smile when tickled. Months later, she still wasn't responding when they called her name or babbling like other babies. Just before her second birthday, her worried parents took her to see a medical specialist. After psychological evaluation, the doctor estimated their daughter's mental age at 10 months old. More testing yielded the much feared diagnosis for the little girl who researchers would come to know as M.
She was autistic.
Autism spectrum disorders affect one baby in every 110 born in the United States. Children like M have trouble communicating and coping with new situations. They have difficulty reading others' intentions or knowing their own feelings, often becoming locked inside their minds. Some children benefit from therapy and instruction in specialized schools, such as the one M attends after her diagnosis.
One morning at her school, M discovers a new toy among the puzzles, blocks, and dolls. The toy is actually a robot consisting of two squishy, yellow balls stacked on top of each other, representing a body and head, with a black button nose and googly eyes. It looks like the progeny of a snowman and a marshmallow Peep. The robot's name is Keepon-a combination of the Japanese word kee, meaning yellow, and pon, the sound this strangely adorable creature makes as it bobs up and down.
Keepon took quite a journey to reach M's classroom. About a decade ago, Hideki Kozima, a scientist at Miyagi University in Japan, had developed a sophisticated robot called Infanoid, which was intended for the study of social development in autistic children. Infanoid had a movable torso, arms with working hands, and a head with eyes, eyebrows, and mouth. It could respond to human faces and voices. Yet Infanoid proved to be too realistic for its own good. Children fixated on its individual parts and had difficulty treating Infanoid as any kind of social being.
So Kozima turned his approach inside-out. Rather than developing a robot that looks and behaves like a human, he sought to distill our social behavior to its essence. This minimalist design might, conversely, lead to a more socially engaging robot, he reasoned. The question was how best to do it.
A key part of the answer lay with Marek Michalowski.
Born in Poland, Michalowski became interested in computers and robotics while growing up in New York City, where his father ran a networking business. He went on to study computer science and psychology at Yale University, where he became fascinated with how to design robots with social intelligence. He then chose to pursue his PhD at Carnegie Mellon's Robotics Institute because, he says, of its renown in the field. At the institute, he worked with computer science professor Reid Simmons on several social robots, including the robo-receptionist that greets visitors to Newell-Simon Hall. He also began reading about a phenomenon called rhythmic synchrony.
Many of our bodies' activities, from the firing of our neurons to our walking gaits, are rhythmic. It turns out that so, too, are some our social behaviors. Scientists have observed that the rhythms of our gestures are synchronized with the tempo of our speech, such as talking with our hands. Rhythms of speech and movement are also synchronized in subtle ways between two people communicating-think about how you nod, blink, or gesture during conversation.
Michalowski likens this rhythmic synchrony to a dance. Yet, even the most socially adept robots can't do this social tango, which could account for why robots aren't persuasively lifelike. He wondered whether he could design a robot capable of rhythmic synchrony with humans.
In July 2006, during the third year of his doctoral studies, he landed an internship in Kozima's lab, where the efforts to create a socially engaging robot were underway. Michalowski was impressed with the potential of the lab's prototype. It was capable of smooth, agile movements, which made Michalowski optimistic that the robot, named Keepon, had potential for rhythmic synchrony.
With its rubber body and cartoonish head, Keepon is the simplest imaginable social creature. A remote operator can see through the robot's eyes, which are cameras. The nose is a microphone. Four motors tug on wires running up its hollow body like an inverted marionette, allowing it to express emotions through bodily movements-rocking to show pleasure, bobbing to display excitement, or vibrating in fear. It can direct attention around a room by panning its head and nodding.
On Michalowski's first day as an intern, Kozima set Keepon on his desk for him to code. The unmoving robot stared blankly forward but seemed to quiver with its desire to bounce and shake. Could Michalowski teach Keepon how to dance? He worked day and night to program the robot to respond to audio and visual signals with synchronized movement. And after several months, Keepon could dance to hip-hop, disco, or any other musical style; and it could dance in synch with a human partner, too, grooving like no other robot in history.
In March 2007, Michalowski uploaded a video to YouTube of Keepon dancing to a song by the rock band, Spoon. It garnered more than 2 million hits. Wired magazine then invited Keepon to star in another Spoon video. The second video also went viral, unexpectedly transforming the Carnegie Mellon student's dissertation research into a pop culture phenomenon. "This dancing can seem like it's for entertainment purposes," he acknowledges, "but I was developing it to answer scientific questions."
Hundreds of children have played with Keepon since the first prototype was built in 2003, and it has demonstrated an ability to charm its way into the lives of just about anyone. Severely autistic children have responded to the five-inch-tall robot with emotion they rarely show to their own parents, and their responses have been even more positive since Michalowski imbued Keepon with rhythmic intelligence.
This research has called into question longstanding theories about the disorder. People with autism can seem to lack the motivation to relate to others. But their desire to interact with Keepon suggests that motivation is actually intact; perhaps it is just difficult to act upon when confronted by complicated perceptual stimuli in a human social encounter.
M is an example. In her initial sessions with Keepon, she won't get close to the robot or look it directly in its eyes. During the seventh session, though, something changes. M approaches the robot's side, keenly gazing at it in profile.
It's this kind of social connection that helped elicit worldwide attention. Keepon took top prize at the 2007 Robots at Play international festival in Denmark, worth 10,000 euros ($15,000). It was featured on NBC's Today, the History Channel, and NPR. And requests started pouring in-from families with autistic children to scientists who wanted the robot for research to parents of healthy children who simply wanted to buy their kids a new toy. Michalowski and Kozima saw an untapped market. They decided to use their prize winnings to form a company. In November 2007, Pittsburgh-based BeatBots LLC began to sell Keepon to research institutions for $30,000 apiece. "The reason for the high cost is that manufacturing the robot requires a guy in a machine shop in Japan milling individual pieces of aluminum by hand," Michalowski explains. "Very little of that is profit for us."
Michalowski had come to Carnegie Mellon with dreams of a life in academia. He never imagined himself as an entrepreneur. Suddenly, he was in charge of the day-to-day operations of BeatBots, with Kozima as a full partner assuming an advisory role. He had to worry about accounting and bookkeeping and figure out things like registering intellectual property and drafting an operating agreement.
Meanwhile, Lenore Blum, a professor of computer science, had heard about BeatBots. She asked Michalowski to speak at an event hosted by Project Olympus, an initiative founded three years ago by Blum to help move to market early-stage innovations generated at Carnegie Mellon. Blum believes if a student or professor wants to build the next Facebook or Google, the university should wholeheartedly nurture those ambitions with resources and support. "This is the way we are going to keep attracting the most talented students and top young faculty," she reasons. "They want to be part of the entrepreneurial culture."
Under the guidance of a Project Olympus business advisor, an embedded entrepreneur, and a network of economic development partners, faculty and students can explore the commercial potential of their ideas by working in PROBEs, or Problem-Oriented Business Explorations. So far, the PROBEs have launched more than 40 companies and generated over $8.7 million in follow-up funding.
Regular events bring together these project teams with investment, business, and civic leaders. In addition, Blum and her colleagues strive to connect their PROBEs to the University's Center for Technology Transfer and Enterprise Creation, which provides support to faculty seeking to license their technologies. The mission of Project Olympus also aligns well with that of the QoLT Foundry, established in 2008 to accelerate commercialization of research at Carnegie Mellon's Quality of Life Technology Center.
All of these organizations comprise Carnegie Mellon's "innovation ecosystem," explains Art Boni, executive director of the Donald H. Jones Center for Entrepreneurship at the Tepper School of Business, where students develop the necessary skills to become entrepreneurs. "You really need a village to get a company started," he says. "And that's what we are trying to create."
The village helped Michalowski when BeatBots became a Project Olympus PROBE in 2008. Senior business adviser Kit Needham assisted him with his immediate business strategy. Michalowski was focused on three areas:
- Manufacture an inexpensive version of his robotic platform for smaller clinics or home-based autism therapy;
- Make Keepon into a mass-market consumer game or toy;
- Partner, through an existing offer, with an entertainment company seeking to experiment with Keepon in its research labs.
Too broad a focus, Needham advised. "He was trying to do all of them," she explains, "and that just wasn't practical."
After a brainstorming session with Needham, Michalowski realized that the entertainment company deal should be a top priority. Kozima agreed. The financial offer would provide working capital to keep BeatBots afloat while the company explored the two other paths. Understanding the moment was imperative, says Needham: "That offer [from the entertainment company] wasn't going to be there forever. And if he didn't act, he would've lost a lot."
Project Olympus also connected Michalowski with an attorney to guide him through the protracted and tricky contract negotiations and helped with reviewing the contract. Lastly, Needham coached him on how to figure out what to charge for his consulting services, another revenue stream.
Michalowski earned his PhD in robotics last year for his research on rhythmic synchrony in human-robot interaction; his position as president of BeatBots immediately became his full-time job. "That's better than what a lot of people can say for what they do with their thesis research after they graduate," he jokes.
From the entertainment company deal, consulting, and Keepon sales (at least 20 now exist worldwide), the company is generating enough revenue to stay in the black. "I'm not rich, but I'm not starving, either," says Michalowski. Keepon continues to receive accolades, too-the little yellow guy was recognized as one of the top 10 robots of the decade by GoRobotics.net.
As for the friendship of Keepon and M, during their 11th session together, the youngster taps Keepon's head with a xylophone stick and feels her friend's belly gently with her hand, as if checking for fever. She then waves her hand at Keepon, listens to its pon sound, and in response makes her own guttural noises. Three sessions later, M looks Keepon in the eyes and lovingly pecks the robot on its head with her lips. Her therapist laughs and claps at this fleeting kiss, as M's mother watches from across the room, weeping quietly.
Jennifer Bails is an award-winning freelance writer. She is a regular contributor to this magazine.