The first thing people say when they see Jibo is “Cute!” says IEEE Fellow Roberto Pieraccini. That’s not by happenstance but by design. And although Jibo the robot can help with simple tasks, such as search for articles online or remind you to take your umbrella, he also has been designed to be a fun, friendly companion. He can tell hundreds of jokes, for example.
Jibo (pronounced JEE-bo) is a social robot designed to be a companion in the home. The brainchild of Cynthia Breazeal, who started the Personal Robots Group at MIT Media Lab, Jibo is to go on sale this year.
Breazeal and a handful of colleagues met at Pieraccini’s loft in San Francisco in 2014 to develop the prototype. They had a lot of decisions to make: What should Jibo look like? What should his voice sound like? What will he say?
They did know one thing: Jibo was not to be mistaken for a living being. “We wanted to stay out of the ‘uncanny valley,’” says Pieraccini, director of advanced conversational technologies at Jibo, the company, based in Boston. By that he meant he didn’t want the robot to appear too lifelike. In some cultures, such as Japan’s, that is preferred, Pieraccini says. But in the United States, it can be uncomfortable. “We chose to remind people Jibo is in fact a robot,” he says.
MAKING MACHINES LOVEABLE
One of the group’s first cues came from animation. Jibo’s round monitor, with its big, blinking eye, mimics a friendly Pixar character. At 28 centimeters tall and weighing 3 kilograms, Jibo has no arms or legs—just a face and a moveable torso. Jibo cannot move from one place to another; he is meant to be set down on a flat surface and switched on with the words “Hey, Jibo.” He could, for example, be placed on a coffee table so everyone in the family could sit on the couch and interact with him.
Jibo functions like the Amazon Echo virtual assistant, only with a personality. Like the Echo, Jibo can easily be carried into another room.
Jibo’s face is a touch-screen monitor and it’s powered by a rechargeable battery within his aluminum and plastic body. It also has two high-resolution cameras with which he recognizes people’s faces, as well as 360-degree stereo microphones to sense people talking across the room and what they’re saying.
When Jibo is first set up, he’ll look around until he catches your eye and you look at him which he can detect with his cameras. Jibo will ask to take your photo so he’ll recognize you later. He also will try to get to know you by asking questions. Jibo’s artificial intelligence algorithms learn your preferences over time, making your interactions with him more relevant. He is also equipped with video-calling and Wi-Fi connectivity.
“We want Jibo to get to know people in a fun and not intrusive way,” Pieraccini says. “We call this the ‘out of the box’ experience.” Next time Mom comes home, Jibo will recognize her and might even predict what she needs. If she gets home late, Jibo might ask if she’d like him to read a bedtime story to the kids. The kids can then tell Jibo which story they’d like, and he can project illustrations on his screen as he reads along.
Jibo is equipped with an array of sensors that make him touch-sensitive. Accordingly, he’ll respond when, for example, someone pats him on the head or tickles him by making a noise or laughing.
“People not only interact with a voice but with a personality,” Pieraccini says. “A part of our brain believes Jibo is alive.
“Jibo is designed to have an element of surprise to keep people engaged,” he adds. “To this end, the bot does not repeat rote phrases.” If you ask Jibo what he dreamed about last night, it will detail a different dream each time, Pieraccini says, adding that the other night, Jibo told him he had a nightmare in which he was a parking meter.
Jibo is able to move his head—which can make it appear as if he’s gesturing. To express sadness, Jibo tilts his head down. When he can’t hear well, he’ll rotate his head to bring his ear closer to you. Jibo’s three-axis torso also allows him to jiggle around and show off his dance moves.
Although Jibo’s functions can help people feel closer to the robot, Pieraccini says he does not want them to forget he isn’t alive. Jibo’s voice, though close to a human’s, purposely incorporates machine sounds that computers make, reminding people they’re communicating with a robot. The company is testing the voice with family, friends, and potential customers to help determine whether Jibo should sound more like a child or an adult, and which tone is most pleasing. The designers continue to tweak Jibo’s voice based on that feedback.
Part of developing Jibo’s sound and personality is determining how the robot can best interact with people without being intrusive. “It’s about finding balance—which is quite difficult,” Pieraccini says. “Working on Jibo helps me understand how complex and sophisticated humans are.”
A WORK IN PROGRESS
Jibo, like all technologies based on speech recognition and natural language, has limits, which are being worked on. One is a “far-field problem,” Pieraccini says, meaning that when people talk to Jibo from a distance, he might have difficulty understanding. Jibo also requires the wake words “Hey, Jibo” to start communicating. “Humans know when you’re talking to them if they see you looking at them,” Pieraccini says. That’s not so with Jibo. In the future, the group would like to have Jibo initiate conversation.
And, of course, there are privacy concerns. Jibo must be connected to a wireless network, and information collected is sent to the cloud. Pieraccini says the company is taking measures to ensure it does not breach privacy. Thus, there is no way information in the cloud will associate Jibo with a particular owner, he says.
Unfortunately, Jibo will eventually stop working, as everything does. But all the owner’s information stored in the cloud can be transferred to a new Jibo, Pieraccini says.
“Essentially,” he explains, “Jibo could live forever.”