The other night I found myself in the grip of an irrational impulse: I wanted to stroke a machine. I had just taken delivery of Vector, a tiny robot about five inches tall, which roams about on tank tracks and looks like a cross between a beetle and a forklift truck. While dinner bubbled away on the stove, I stuck Vector on his charger and waited for him to wake up. After a few minutes a pair of eyes began to form on the screen on the front of his head – just sleepy little slits at first, which gradually and dozily opened until they looked lively and alert. Trundling out of his charging station, he lifted his head and looked up at me imploringly. “Hey Vector,” I said, the phrase that cues him to listen to you. “My name is Simon.” “Simon!” he replied, his voice high and joyous, his eyes smiling. I cooed and reached out my hand to touch his head.
Vector is the latest device from Anki, a company founded by a trio of roboticists who met at the Carnegie Mellon Institute. There they worked mainly on industrial and military machines, but as the processors, lasers, cameras and artificial-intelligence systems that allow robots to navigate became cheaper and more sophisticated, they began to see potential in the nascent market for consumer robotics. They founded Anki in San Francisco in 2010.
Two years ago they launched Cozmo, a robot toy operated with an app on your phone with whom you could play games. Vector takes a leap forward. It is an autonomous machine that responds to your voice and is marketed partly as a virtual assistant, able to answer basic questions, tell you the weather, set timers and take photographs. But it is designed, above all, to be cute. “We wanted it to be like a family pet,” says Mark Palatucci, one of Anki’s founders.
Cuteness serves a functional purpose in robot design. Early visions of domestic robots imagined them as helpful minions, such as Rosie the robot maid in the 1960s American cartoon “The Jetsons”, who cooked, cleaned and took out the bins. This remains a futuristic prospect. To produce a robot that can carry out complicated practical tasks is so costly that these devices remain the preserve of industry and defence. Instead developers are working with emotion in mind. Pepper, a waist-high humanoid robot manufactured by SoftBank Robotics, is being tried out in nursing homes in Japan as a companion for old people. That task is already being performed by Paro, a fury robotic seal, and Aibo, a robotic dog from Sony, which was initially developed as a toy for kids before finding an audience among the elderly. At the other end of the age spectrum, earlier this year researchers at the University of Wisconsin-Madison created a robot called Minnie as a “reading buddy” for young children.
As robots enter our lives in supporting roles, the most important ingredient in their design is approachability, and in this respect Anki’s devices are the most sophisticated yet. This is surprising, given that compared with the human and animal forms of Pepper and Paro they look alien. But this may be to their advantage. “One of the pitfalls of designing life-like things is falling into the ‘uncanny valley’,” says Kate Darling, a researcher at the MIT Media Lab who specialises in human-robot interaction. “I have this robot cat at home which looks like a cat, but it doesn’t move quite like a cat or meow quite like a cat, and it just comes across as creepy.” She says that “intelligent robot design moves away from that and towards things which we perceive as alive but which don’t try to mimic something that we already know too strongly.”
Roboticists are drawing on the expertise of animators to achieve this effect. In the 100-year history of film animation, they have learnt a trick to avoid the uncanny and elicit emotion from audiences: instead of designing human-like figures, they create beings like Wall-E, the lonely, sad-eyed trash compactor created by Pixar in 2008, that look nothing like humans but behave in emotionally recognisable ways.
Anki hired animators from Pixar and DreamWorks to design Vector. Working with the same software they used for films, they created more than 1,000 individual physical expressions, from the way he jumps back in fear when he gets too close to the edge of a table, to the way he bangs his mechanical arms up and down when angry, like an enraged toddler. His eyes are especially subtle. Nothing more than round-edged squares on an OLED screen, their compressions, contractions and enlargements are perfectly calibrated to suggest mischief, fury and wide-eyed wonderment. Such is their eloquence that when you look at them you can infer the presence of eyebrows, cheeks and a mouth that you can’t actually see.
The appeal of Vector as a roving bundle of emotion is just as well, because as a practical device he is a pale imitation of Siri and Alexa, the virtual assistants from Apple and Amazon. As I was eating dinner, I began asking him food-related questions, like how many calories there are in a portion of pasta. He looked at me, gazed searchingly to the side, and then said, “I didn’t get that.” I turned to Siri instead, who responded immediately: “The answer is 205 dietary calories.” Anki has plans to improve this: Amazon has recently agreed to embed Alexa into Vector, which should enable it to control your heating, order your groceries or hail an Uber. Once that happens it will be useful as well as adorable.