David Stonehouse Features

The robots are coming

Within five years, the boundary between humans and artificial creatures will begin to blur

By David Stonehouse

Jose was the media darling of the University of British Columbia's computational intelligence lab. He was a snazzy dresser too -- outfitted neatly in a black tuxedo and white gloves, he had a distinguished, refined air.

But then Jose would speak.

"You are blocking my path," he'd bark. "Please move out of my way."

Or he might, in his snarky monotone, accuse you of stealing, or moan about his day. "I am so tired of going around obstacles." Oh, the hard life of a robotic waiter.

Jose's smart-aleck attitude -- and success at a robot-waiter competition staged as part of an artificial intelligence conference in Seattle last summer -- made him a novel catch on the media circuit.

A natural for television appearances, he soon became a 'bot in demand, and it was not long before he was criss-crossing Canada, dropping in to one publicity event after another to demonstrate the hors d'oeuvre-serving skills that helped him win the robotics competition.

His career was short-lived, though. Jose is retired now, and has become both yesterday's robot and a symbol of the future -- his creators are deprogramming him and tinkering with upgrades. Jose the robot waiter is being turned into Homer, a Human-Oriented Messenger Robot -- an early prototype of a looming robot revolution.

The transformation of Jose is well underway inside Room 108 at UBC's Centre for Integrated Computer Systems building. For a few months now, doctoral student Pantelis Elinas and others have been working on his evolution, outfitting him with an LCD screen and setting him up with a new two-camera vision system. Unlike most other university robotics labs that use laser or sonar to allow their robots to navigate, the UBC lab
uses cameras that send visual cues to its robots -- a system that lab researchers say is the most advanced in the world.

Where Jose functioned with only one on-board computer, Homer will have four computers. The humans in the lab are designing and tweaking a whole suite of software that will make him tick, enabling him to understand voice commands, recognize certain humans and their facial expressions, follow people and navigate around the lab.

The hope is that Homer will be able to roam the three rooms of the lab until someone calls or beckons him over. Homer will then analyse the person's face to determine who it is. He'll listen for the message while watching face and head movements for clues about things such as the urgency of the message.

Homer will then seek out the person who the message is to be delivered to, aided by his own knowledge of where that person usually works or takes breaks. Once he has passed the message, he will return to tell the sender the message was delivered. He could be up and functioning within a year.

Elinas, who's working on a thesis on human-robot interaction, says a messaging robot is better suited to his studies than a robot waiter. " When was the last time you bothered to converse with a server at a reception?" asks Elinas. "[Creating a robot waiter] was just our response to a challenge that we thought was worth conquering."

With researchers at UBC and around the world conquering the challenges and mysteries of robotics, a whole new revolution is underway. The robots, it seems, are on their way. Finally.

There have been bold predictions before of a world sliced from the futuristic cartoon family, the Jetsons -- a world where robots whirred around the house taking care of the mundane, the tedious and the monotonous. This was a world we were supposed to be living in by now.

But the breakthrough never came. Sure, robots have worked diligently away in factories and helped us explore space, but office help -- like Homer -- and household help, like the Jetsons' faithful mechanical maid, Rosey, has been longer in coming.

Now, however, advances in artificial intelligence and the galloping speed at which computer power is becoming simultaneously faster and cheaper are helping to make it happen.

The United Nations Economic Commission for Europe is predicting something close to an invasion of domestic robots. In statistics released last year, it forecasted that there will be as many as 290,000 household 'bots purchased around the world by 2003 -- nearly 10 times the number found in homes in 1999.

Rodney Brooks, director of the famed artificial intelligence lab at MIT, is predicting the revolution is about to "burst over us."

In a new book released this spring, he talks of an explosion of creation that has seen companies around the world -- including household names like Sony and Honda -- race to create domestic, humanoid robots.

"Mankind's centuries-long quest to build artificial creatures is bearing fruit," he says in Flesh and Machines: How Robots Will Change Us. "The coming robotics revolution will change the fundamental nature of our society."

The robots we know best are those fashioned by the imagination of science fiction -- humanoids like C-3PO in Star Wars or Star Trek's Commander Data. Brooks believes the world of fantasy they inhabit is about to become real.

In five years, he says, the boundary will begin to blur. And in 20 years, it will vanish altogether and robots will be a pervasive presence in our lives.

Pretty soon we will stop bothering to count the robots in our homes, Brooks says. "They will be a new class of entity, moving about under their own free will, doing their tasks as they decide they need to be done."

He envisions a small army of robots around the home, each doing specialized tasks. Teams of tiny robots -- which he dubs " almost-life-forms" -- could live on ledges, for instance, and be entrusted with keeping our windows clean.

"We might soon see cleaning robots for the bathtub and for the shower recess. We might soon see a robot that can clean toilets," he says.

And then there is the robot nearest and dearest to the heart of sports fans -- the one that'll fetch a beer. In his book, Brooks says this robot is not far away. Since he wrote those words, an engineering student in Florida has made some important strides toward this lofty goal. Jean-Philippe Clerc created Abor -- the Autonomous Beer-Opening
Robot -- which can manoeuvre along a bar, steady a long neck in front of it, pop the cap and back away.

Abor hasn't yet reached the market, but the first generation of robots is already here.

Robot vacuum cleaners are already on sale in Sweden and will likely be introduced in the United States in the near future. Several companies have already showcased them. Eureka boasts that its Robo-Vac -- a disc that whirs around the floor on its own, gobbling up dust and dirt -- could just be "the advent of effortless cleaning."

Robotic lawnmowers are already available in the U.S. for those who are fed up with sweating away their Sundays trimming the yard. Toro's iMow, for instance, zigzags randomly, chewing away at any patch of grass in its path. All for an initial outlay of less than $500 US.

Robots that can help care for your elderly parents are coming. The Fraunhofer Institute for Production Technology and Automatization in Stuttgart has already demonstrated a prototype of its Care-O-bot, a short cylinder of a robot on wheels that can help out by serving breakfast, bringing over the TV guide or, with optional attachments, helping his owner walk.

And just this week, Honda took out two-page full-colour newspaper ads to herald the innovation of its humanoid robot, Asimo. The 1.2-metre-tall machine looks like an astronaut clad in a white space suit. It can walk, climb stairs, dance, remember phrases and talk.

But don't look to pick one up at Wal-Mart any time soon. For now, Asimo is only available for lease -- for the princely sum of $150,000 US. IBM in Japan uses one as a receptionist.

Rodney Brooks, as you might expect, is a man with a robo-mower. He bought one two years ago, and it proved to be more an aggravation than anything else.

It mulched away under its own power for two hours or so before emitting a beep-beep-beep warning that its battery was running low, which meant recharging it overnight before setting it loose again the next day. Even after a second round, there were still plate-sized patches here and there that the robot had not covered. Brooks figures he spent more time tending to the robot than he would have doing the lawn himself. To make matters worse, his wife Janet hated the mish-mash pattern the mower left in the grass.

He figures it will be 20 or 30 years before cheap household robots like that come with navigation technology good enough for the machines to know where they have been, where they are going and how to get to the closest recharging station before they power out. But the teams of cleaning robots could come sooner, since each robot is responsible for cleaning only a limited, well-defined space.

One of the innovations Brooks thinks will help kick-start the revolution -- the "killer app" that will be quickly and widely embraced -- is to use robots for remote work. Robotic co-workers could, in the not-too-distant future, work alongside humans, operated by humans living in other countries, or even other continents.

Nations plagued by labour shortages will be able to supplement their workforce in this way, Brooks suggests. Japan, with its low birth rate, aging workforce and near-zero foreign worker population, is heavily involved in remote work research.

"Remote presence provides a way out of the Japanese conundrum. Foreign work, yes, foreign workers, no. And the United States and Western Europe will not be far behind," Brooks says in his book. "They too may well turn to the importing of physical work by renting the minds of workers in foreign countries, minds that become the supervisory controllers of robots across the globe."

The advantages of overseas control will help keep us closer to home too. This kind of innovation will also make it possible to do things around the house even when you aren't there. Say you can't remember if you have left the stove on before you rushed out the door to your office. Well, Brooks says, just fire up the office computer, connect to your in-home robot via the Internet and tell it to go over to the stove. Through video imaging, you see what the robot sees -- and discover you did turn off the burner after all.

This, he says, is possible today, and "early-adopters" -- keeners who rush to buy the latest technological innovation -- are already in line to snap up the first generation of remote-presence robots.

When Brook develops a robot, he's aiming for something more than a walking, talking menial labourer -- he's looking to create a robot that behaves like a human. At his MIT lab, researchers work on a cast of robots with names like Coco, Cog and Macaco. The most famous of the bunch is Kismet, a robot programmed to seek out human attention and react with human-like emotion.

With its large bulbous eyes, droopy eyelids, protruding ears and lips that look like strands of licorice, Kismet's antics tend to charm visitors to the ninth-floor lab. The mechanical head looks crude, but hides an array of technology from wide-angle cameras to microphones to actuators that manoeuvre the head. A network of 15 computers serves as Kismet's brain.

Kismet can speak limited English, keep eye contact during a conversation and react to movement and other social cues, making humans believe he is paying attention to them.

"Kismet can interact with people like a human," Brooks says. "Kismet acts like it's alive."

Kismet is programmed to seek out things that move, have saturated colours or skin tone. When there are several in his field of view, he chooses based on his "mood." If he is lonely -- a state that occurs if he hasn't encountered any of these human-like elements over an extended period of time -- he is much more likely to zero in on the thing with
skin colour. If bored, he leans toward interacting with colourful objects. If something moves, he locks his eyes on it and follows it. Kismet can also hear and detect intonations in voice that can also influence his mood, much like our own moods are affected by the
emotional states of people around us. Kismet's "emotions" are determined by a set of internal drives that change according to what catches his attention. His emotional state then dictates his mechanical expressions -- he might raise an eyebrow if surprised, or pull his lips into a frown if unhappy -- and change his inflections of voice.

"What Kismet cannot do is actually understand what is said to it," Brooks explains in his book. "Nor can it say anything meaningful ... Kismet hears only that people are speaking and the prosody in their voices."

What Kismet says in return is usually nonsense, but he does know his manners -- he waits for his turn to speak, shifts his gaze at appropriate times and fills in awkward silences in conversation.

The work with Kismet is proving that it is possible for robots to exhibit emotion and behave like a human. It is not real human behaviour, not real emotion, but that is just semantics to someone like Brooks.

What is important, he argues, is that people think it is.

"The question of real or not real doesn't make sense any more. If robots' emotions are not real then other peoples' emotions are not real either," he says in a telephone interview from his home outside Boston. His voice is tinged with a drawl from his native Australia.

Brooks is no computer geek caught up in his own little world deluding himself with his own futuristic fantasies. But he is a firm believer in the changed world to come. He is helping to make it happen, after all.

It could have been destiny. The son of a defence department electrical technician, Brooks has been toying with technology since he was a child. By the time he was 12, he'd built a computer that could play tic-tac-toe. Later he read a book by W. Grey Walter, a 1940s pioneer in robotics, and became inspired to create his own robot. It worked -- Norman moved along the floor and responded to light.

Brooks' continued deep-rooted fascination with machines remains something of a mystery, even to him. "Why? I have no explanation. My brothers and sisters are just ordinary people back in Adelaide, Australia, enjoying their lives, going to the beach. I can't explain it," he says, though he goes on to share some insight to the appeal.

Creating robots means coming to grips with what makes humans work. That's part of it.

"The other is I like to make these things actually work," he says and laughs. "I really enjoy it, immensely."

Still, the early robots just now hitting the market are crude when compared to Hollywood robo-stars like C-3PO or R2-D2. Both the lawnmower and vacuum robots are far from intelligent -- they whir around randomly with no clue where they are.

Getting robots to understand where they are, where to go next and how to successfully operate in the world around them have been the toughest problems for researchers to solve, and they are the reasons why the revolution has not taken hold long before now.

"The real world is so rich," says Martin Buehler, director of the ambulatory robotics lab at McGill University's Centre for Intelligent Machines.

"For example, it is exceedingly difficult to get a robotic vision system to do anything other than the most trivial tasks in a real-world environment -- where the lighting is not controlled, where you can see all sorts of things that you have to understand what they are," Buehler says.

"Go out and walk on the street -- in an environment where it can be raining, it can be overcast, it can be dark. That is so much more difficult. And that has not been solved at all. Many things that we take for granted are just much harder than researchers have anticipated."

For the last 10 years, Buehler has been working on making robots mobile and has created machine creatures that can climb stairs, swim, run and turn over. But he is not ready to crack a guess when robots will be ready to become a part of our everyday lives.

"Until everybody has a robotic vacuum cleaner or until everybody has a biped-type robot that washes your dishes?" he asks.

"It is pretty much impossible to predict. Right now nobody really has a handle on what it is that will compel people to go out and buy a physical, mobile device that will serve that need. And it is completely dependent on that. Without that insight you could say two years or 20 years."

He shies away from talk of a revolution.

"It's more an evolution," he says. "It's more a slow and steady trend."

If we can build robots to do as we do, then it isn't much of a stretch before we become part machine ourselves. Of course, this is already happening in limited ways with sophisticated artificial limbs, cochlear implants that help deaf people hear and experiments with retinal chip implants.

Brooks imagines a day when our brains will be wired into computers, boosting our thinking power. Computer users, he suggests, will be able to operate the Internet without a mouse, using only their mind. Students may someday be required to take ISATs -- the Internet-aided aptitude test.

He is not alone in this kind of thinking. There is a debate in artificial intelligence circles over whether robots could end up a superior race, outwitting and outperforming humans, thanks to the ever-sophisticated advancements in technology we gave them.

Brooks, though, dismisses this notion. He believes we will stay a step ahead of the robots by becoming part machines ourselves -- a race of robot people.

"There is no need to worry about mere robots taking over from us," he says in his book. "We will be taking over from ourselves with manipulatable body plans and capabilities easily able to match that of any robot."

David Stonehouse last wrote for Mix on e-mail overload.