Sebastian Thrun builds cars that drive themselves. In fact, the Stanford University scientist may do it better than anyone in the world: In 2005, his fully automated Volkswagen Touareg drove itself 212 kilometres across the Nevada desert in less than seven hours to claim a $2-million prize in an international robot off-road race.
He predicts that a fully robotic car for consumers is less than 20 years away. But, as Dr. Thrun understands with a personal, still-sharp grief, the aim here is higher than putting an ultra-cool set of wheels in wealthy people's driveways: This is technology that could help save an aging population.
Last May, when his elderly father, who lives in Germany, caused a car accident, Dr. Thrun and his brother decided that they had to take away his driver's licence. Their father was furious. Without his freedom, he simply faded away. In November, his sons moved him into a nursing home. By Christmas, he was gone.
"It was a very sad episode," says Dr. Thrun, who still blames himself. "I caused it by deciding it would be unethical for my dad to drive." Losing your independence is the nightmare of getting old: "It's a human disaster when it happens."
In Canada and across the industrialized world, growing numbers of people are staring at the fate of Sebastian Thrun's father. According to the 2006 Census, released this week, one out of seven Canadians is now over the age of 65 - double the proportion half a century ago. Four million others are less than 10 years away from senior citizenship.
The resulting burden on a limited group of caregivers will be immense. Young adults will be too busy keeping the economy running to tend to ailing parents. And those parents - the wealthy, educated, super-mobile baby boomers - will expect to live in their own homes and drive their own cars until they are dragged into nursing care.
Their best hope may be the advent of robots who can serve as companions, nurses, drivers, household help and safety monitors for aging Canadians. And, luckily, robots are evolving just as fast as people are aging.
Bill Gates has said that the robotics industry today is on a similar threshold to where computers stood 30 years ago. By that standard, we might expect dishwashing, babysitting androids by mid-century.
Sebastian Thrun predicts that within 10 years - the time it will likely take for Canadian seniors to outnumber children - there will be cars that can motor down the highway and change lanes; the human at the wheel theoretically could take a nap.
Researchers around the world are fine-tuning a virtual ecosystem of robots, which may or may not look like science-fiction fantasies - computers watching us through cameras, intelligent wheelchairs, machines with perky grins and cone-shaped bodies - but can remember prescriptions, give directions, order takeout or just have a chat.
"Baby boomers don't want to retire like their grandparents," says Dan Kara, president of Robotic Trends, an industry consulting company. "They have the money. They have the need. The technology is there. It's the perfect storm."
And the winds are picking up: The two million personal robots in use as of 2005 - mostly toys and a few flying-saucer-shaped robots that vacuum carpets or mow lawns - make up a tiny portion of the robot industry, which largely serves companies that want cheap, precise factory workers. But by next year, the International Federation of Robotics predicts 5.6 million more personal robots will be sold.
A new Japanese cone-shaped robot on wheels called RI-MAN can see, hear and smell (by chemical detection), and the company website shows it effortlessly lifting a young woman. Some day soon, its creator has boasted, RI-MAN will be able to tell if a human is sick by smelling his or her breath.
Last week, in Nagoya, Japan, a temp agency (ironically named People Staff) announced that local businesses can now hire Wakamuru, a three-foot-tall, rolling yellow robot that recognizes faces, gives visitors friendly directions to destinations - and works for human wages, roughly $25,000 (all figures U.S.) a year.
By September, the flush narcissists among us will be able to pay $125,000 for a conversational companion robot fashioned in our own images from Hanson Robotics in Texas.
The question is: Are we ready for them?
Many scientists say it's no longer clunky software keeping robots in computer labs. It is the ethical and legal debate over what human beings want and expect of artificially intelligent machines once they head out into the world. Designing a car that can drive while you snooze at the wheel may be easier than figuring out who is liable if the car crashes.
"It's very clear that the technology is up to the task," Manuela Veloso, a computer scientist at Pittsburgh's Carnegie Mellon University, says by phone while waiting for a flight at LaGuardia Airport in New York City.
"I am here at an airport thinking, 'Why aren't there any robots here helping people?' There is nothing that forbids it from a technology point of view. This is a commercial problem now."
Game boys
To get an idea of just how far robotics has come in a decade, consider the RoboCup, a soccer tournament for smart machines.
In 1992, Alan Mackworth, a computer scientist at the University of British Columbia, says he was discouraged by the static approach to artificial intelligence research and wanted a way to test robot behaviour in an environment that approximated the real world. He and his students wired up a few monster trucks and model Porsches they bought at Canadian Tire and set them loose with a beach ball inside a field fenced in with 2-by-10 sheets of plywood - the first robot soccer match.
In a game context, he reasoned, robots would have to react and make assumptions about their environment: "In soccer, hardly anything works out the way you expect it to."
Five years later, shortly after IBM's Deep Blue computer defeated the world champion in chess, Nagoya played host to the first international RoboCup, with 40 robot teams competing. Dr. Veloso, who was there, remembers ungainly hunks of metal jerking nearly aimlessly around the field - nothing like the robots that gathered two weeks ago to play soccer in Atlanta.
Today, they run, see the ball in real time and talk to each other. When they fall, they pick themselves up. Among the 1,000 robots that gathered at Georgia Tech were some that stood on two legs and others too small to be seen without a microscope.
This year, robots played soccer against a team of human beings for the first time - the ultimate goal of RoboCup is to beat the FIFA World Cup Champions by 2050. So far, humans still have the edge: Our team won 2-0, with the robots struggling to take possession of the ball. One weakness, Dr. Veloso says, is figuring out how to get robots to learn new strategies as the match progresses.
Playing soccer, she says, uses the same skills robots will need for search-and-rescue or for navigating traffic, algorithms that will let them alert the forgetful senior to a burning saucepan left on the stove or help her to the washroom.
That said, there are no Star Wars C-3POs in the near future: The smart robots of the next few decades will most likely be hidden in the walls of our homes, controlled by our cellphones or rolling on wheels.
Humanoid robots, while the subject of cultural fascination, are expensive to build and not entirely practical. After all, Dr. Thrun points out, a dishwasher doesn't look anything like a human being, but it still does a better job at the dishes.
Thanks to work done at UBC, robots now see the world much as we do, by mapping out the shape of an environment rather than blindly using infrared and sensors to determine distance from objects.
At a competition that is part of the annual conference for the Association for the Advancement of Artificial Intelligence in Vancouver next week, teams of robots will compete in a new scavenger hunt, which requires them to match physical objects though an Internet search, without human assistance. That skill eventually could help companion robots find lost keys or determine that their owners have fallen (and are not sleeping) by their position on their ground.
This Monday at the conference, a computer from the University of Alberta will bluff through four rounds of Texas Hold 'Em against two professional human players in the first Man vs. Machine Poker Challenge, for stakes totalling $20,000. To win, the software, called Polaris, must learn the strategy of its competitor, says its creator, Jonathan Schaeffer, a team leader on the U of A project.
Faster computers and smarter sensors, along with military-developed applications such as global positioning systems, suddenly have made it possible for robots to process huge swaths of data relatively cheaply and quickly.
A robot that once may have tripped over a rock tossed in its path can now adjust and go around it. This ability to synchronize information from many different sources in real time is what makes it possible for Dr. Thrun to send a car on its own into the desert.
When the Grand Challenge race, which is hosted by the Defence Advanced Research Project Agency (DARPA) in the United States, was first held in 2004, nobody finished and the prize money went unclaimed. The best entry travelled less than 12 kilometres. The next year, five vehicles crossed the finish line.
That's amazing progress. But Dr. Thrun, who heads Stanford's AI laboratory and its DARPA team, says robot technology will probably find its way only piecemeal into passenger cars - like the Lexus sedan that now can steer itself into parallel parking.
"For example, it is completely within reach to build cars in the next decade that can stay on a highway with absolute super precision," Dr. Thrun says. "And when you want to change lanes, you press a button."
In this way, robot cars also will reduce traffic accidents, cut pollution and make highway travel more efficient.
But navigating a busy city street is a complex endeavour, even for humans, and robot cars aren't there yet. In November, Dr. Thrun and his team will try to repeat their victory at the DARPA Urban Challenge. This time, the cars will race through a city-style course, complete with simulated pedestrians.
"Cars will drive themselves on public roads entirely by themselves by 2025," he predicts. That seems very soon, but he says: "Just go back 10 years and ask yourself how the world was different."
Circuits of compassion
When Alex Mihailidis was a young grad student, he encountered a senior fellow engineer caring for a wife with Alzheimer's, who now needed her husband's help in the washroom, an embarrassment for both of them.
"In passing," Dr. Mihailidis recalls, "he said, 'Well, wouldn't it be neat if a computer could do these things, and take care of people at home?' That kind of sparked something inside of me."
Now a researcher at the University of Toronto, Dr. Mihailidis is testing a third prototype of his invention in a Toronto long-term care facility - an artificially intelligent caregiver to help Alzheimer's patients wash their hands properly.
Using a camera mounted in the bathroom, the computer system can tell when a person forgets, say, to wash the soap off. If the person doesn't need help, the computer stays quiet.
This kind of technology, Dr. Mihailidis points out, at once gives independence and privacy to the ill person and makes things easier on the caregiver.
Eventually, he expects homes to have a seamless system of assisting computers with voice capabilities, to help people use the toilet, to remind them to take their pills, to monitor blood pressure or to call a neighbour or ambulance in emergencies.
"It is a more realistic vision to have these systems embedded into our environments, as opposed to having this little thing that follows us around and does things for us."
He is also working with Dr. Mackworth and other scientists at UBC to build a motorized wheelchair for patients with cognitive problems: It would stop itself from crashing, and give directions - it might remind a passenger about lunch time, for instance, and guide her to the cafeteria.
Already, in Japan, big-eyed, human-shaped robots with cheerful high-pitched voices are being marketed as company for lonely seniors.
"There are a lot of elderly people who spend their days and evening alone," says David Hanson, whose company uses a trademarked skin called Frubber to make highly realistic humanoid robots.
Dr. Hanson sees his creations as companions, which might deliver messages from family members, read the news or even record a person's memoirs. His website describes his life's mission as being able to "advance the human race by creating benevolent robotic applications which will serve for the greater good of all mankind." For now, he concedes, his robots are mainly for entertainment.
Once we do get "benevolent" robots serving our needs, though, who takes responsibility when they make a mistake? Who decides when a person with dementia has to have a machine watching them use the bathroom? Even more fundamentally, as robots start to look more human, what impact will that have on how humans treat them?
These questions have become the subject of much debate among scientists and ethicists. This year, South Korea - which has set the ambitious goal of a robot in every home by 2013 - announced plans to write a charter of robot ethics. The U.S. Congress announced last month that it was creating a bipartisan caucus to study the social impact of robotics.
The question of legal liability creates a major hurdle for companies and it has made it hard to lift the technology out of the domain of toys - such as next Christmas's two must-have robots, an adaptive baby dinosaur named Pleo and a spybot named Spyke that can "guard" the house or send video footage of kids to office-bound parents.
Robot toys create a safe and profitable testing ground for robotics with more serious applications, Mr. Kara of Robotic Trends says.
But as robots become more adaptive, predicting their behaviour in every scenario gets more complex. Normally, Dr. Mackworth says, they are built by trial and error, debugging them until they work most of the time - a standard that doesn't hold up for general public use.
"There's a huge issue of trust for robots," he says. "If they are in a factory and you've got a safety curtain around it, you don't have to worry too much if it destroys itself. But as soon as you move it into a home, where there are humans, you have to deal with human interactions and safety issues."
Expect, then, that Sebastian Thrun and his peers will be able get a car to drive itself reliably long before consumers can haggle over one at the local dealership. Car manufacturers have spent long years keeping responsibility for crashes on the operator's side, not their own. But can a sleeping driver in an automated car be held liable? "We have to come to grips with this," Dr. Thrun says. "It will delay things."
For now, consumers may have to settle for the Roomba, the robot vacuum cleaner, and put their name on the waiting list for a Pleo, due out in the fall. Although, at Hanson Robotics, a new market for the companion character robot has emerged among the aging. The company recently took an inquiry from a husband seeking a robot version of his wife to have by his side after she passed away.
"It would be a very bold form of memorial," says Dr. Hanson. "And we would do a very conscientious and considerate job of portraying them."
Erin Anderssen is a feature writer for The Globe and Mail.
Erin Anderssen