Decades from now, how will we communicate with our smart devices? Probably the same way we communicate with the real people in our lives today.
USC computer scientist Andrew Gordon predicts human-computer relations in the next few decades could actually bear a striking resemblance to how 1960s-era sci-fi filmmakers imagined them. Remember Star Trek and 2001: A Space Odyssey? “The whole spaceship is the computer in this model,” says Gordon. “You don’t type into it; you talk to it as if it were a person.” (Ideally, it won’t be prone to psychotic malfunctions.)
A research associate professor with USC’s Institute for Creative Technologies, Gordon is an expert in storytelling and machine intelligence. In a recent lecture titled “Mind-Reading for Robots,” he argued that human-computer interaction of the future will look a lot like the human-human interactions of today. He also thinks the dreary desktop interface — the dominant computing metaphor of the past 30 years — is on its last legs.
Where’s my trash can?
In the early days of personal computing, when Xerox rolled out its pioneering Star interface, the desktop metaphor made perfect sense. Designed for office workers, it borrowed familiar objects from the white-collar workplace. For a generation of first-time computer users, icons such as the paper clip and scissors rendered an intimidating new technology instantly accessible.
Twentieth century office workers already knew how to interact with trash cans, folders, printers and calculators (when the desktop interface was introduced).
“Twentieth century office workers already knew how to interact with trash cans, folders, printers and calculators,” Gordon explains.
A generation later, the desktop metaphor is hopelessly antiquated. Absurdly, the desk calendar and Rolodex remain fixtures in computer applications long after they’ve disappeared from real-world offices. “These relics of office work have even invaded our pockets,” Gordon says, holding out his iPhone.
That’s all about to change.
Small, cheap and ubiquitous
As computing technology continues to shrink in size and price, Gordon believes the desktop metaphor “is going to come crashing down.”
In the decades ahead, who knows what form computing power might take? “It might be something you squeeze from a tube like grease,” Gordon speculates. “Maybe we’ll have computers that look like postage stamps: we’ll be ripping them off and sticking them on stuff. There might be more capability in a single stamp than all the computing power on the planet today.”
Computing power is also getting much cheaper — someday soon, Gordon believes, it may be ubiquitous and nearly free. When that happens, office workers won’t be its primary users. It’ll be brickmakers in India, rice farmers in Cambodia, fishermen in Indonesia. “There are 7.2 billion people on the planet right now who will soon be using computer technology,” Gordon says.
The lives of Bubble Guppies
What’s the most natural way for all these different people to interact with complexity?
“The only interface that makes sense for this future era of computing is anthropomorphism,” Gordon says, referring to the trick of attributing human characteristics to a complex system as a way to engage with it.
A driver doesn’t need to understand how the fuel injection system works to operate a car..
He offers an example: “A driver doesn’t need to understand how the fuel injection system works to operate a car. If the car is having problems, he might tell the mechanic: ‘The engine is temperamental, or sluggish. Or mad.’” Mechanics learn to interpret such anthropomorphic statements as code for any number of engine problems. So, too, might a computer learn to interpret references to psychological state as code for a spectrum of possible desired computing tasks.
Using abduction — a form of logical reasoning that assigns the most likely cause to an observed behavior — computers could be taught to decode much of what people say and do in terms of underlying mental states and processes.
It’s a little like interpreting the behavior of Bubble Guppies on the popular Nickelodeon cartoon series, Gordon says. Using a fairly limited palette of motions and facial expressions, animators can convey to their young viewers a wide range of emotions, desires and intentions.
When a Bubble Guppy shivers, laughs or frowns, kids understand what she is feeling. They need no elaborate theory of psychology to figure this out, just some common-sense models of everyday human mental life.
We need to equip computers with the same common-sense theories that we all use to understand each other in social interactions.
“We need to equip computers with the same common-sense theories that we all use to understand each other in social interactions,” Gordon says.
He and co-investigator Jerry Hobbs, a researcher at USC’s Information Sciences Institute, have been working on just that. Expanding on ideas advanced by mid-20th century social psychologist Fritz Heider, their efforts are bearing fruit after 14 years. Last fall, they put the finishing touches on a book that encodes commonsense theories of human psychology as a set of 1,600 logical axioms. It is a first step toward the future of human-computer interaction, says Gordon, “where we treat our computers as if they were people, and they know how to play along.”