ROBOT: Beginning to think like people

From: J. R. Molloy (
Date: Wed Mar 21 2001 - 16:15:44 MST

They're beginning to walk, talk, and, yes, think like people.
Is the age of the robo sapien just around the corner?
Your Next Doctor May Not Be Human
Chart: The Robot Population Is Growing

It's 1996. Japan's economy is lurching toward recession. Instead of
investing in a new generation of expensive industrial robots, Japan is
shipping more production offshore to Southeast Asia and other cheap-labor
locales. It's a tough time for Tatsuzo Ishida, Sony Corp.'s (SNE ) ace
robotics engineer. He has spent the past 15 years developing steel-collar
workers to assemble millions of Walkmen, handycams, and game consoles in
Sony's Japanese factories. With Sony joining the exodus, the future of
factory robots looks dim.

Being a robot maniac, Ishida refuses to throw in the towel. Instead, he
hatches an outrageous plan. Together with comrade-in-arms Yoshihiro
Kuroki, he proposes that Sony engineer a whole new species of humanoid
entertainment bots, along the lines of C-3PO, the golden chatterbox in
Star Wars. The technical challenges are tremendous. Nobody has yet built a
biped that can stroll blithely through a house, maneuvering around
furniture and dodging people--unaided by a human controller. And if the
robotmeisters actually achieve their goal, Sony could face major liability
risks. How long before one of the creatures trips and topples onto a
toddler, or wanders into the path of a car on its way to the supermarket?

To Ishida's amazement, Sony goes for the idea, despite all the potential
snags. The new business development chief, Shingo Tamura, turns out to be
as much of a robot nut as Ishida and Kuroki. All three studied humanoids
at Waseda University's famed robotics lab in Tokyo. "Anyone else would
have dismissed us as lunatics," admits Ishida. "But we shared this dream
of building personal robots."

Last November, Ishida's dream took its first bow in public. At a time when
the world was going gaga over Aibo, Sony's insanely cute robotic puppy,
Ishida and Kuroki put a half-dozen prototype humanoids through their paces
in Yokohama at Robodex, a new expo for personal robots. The crowd was
spellbound as the pint-size acrobats, just 50 cm tall, jumped, danced, and
kicked mini soccer balls. No one minded that the bots were fed cues over a
wireless computer network. The acrobots were real. Now the future had a
symbol of hope called SDR, or Sony Dream Robot (view a Sony promotional
video for SDR).

SDR and other robots at the expo caused a sensation in Japan. Years from
now, Robodex may be remembered as a watershed event. While Aibo and its
ilk now hog the headlines, scientists and engineers in laboratories across
Japan, the U.S., and Europe are struggling to build far more sophisticated
bots in the image of humans. The timetable is fuzzy, but so-called robo
sapiens could be strolling sidewalks by decade's end. Whenever it happens,
the new humanoids promise huge benefits to society--and financial bonanzas
to their creators.

SOCIAL SKILLS. Many of the inventors are inspired by the sheer scope of
the technical challenges. The complexity of building supercomputers,
erecting skyscrapers, or even designing whole cities pales beside the task
of imbuing machines with humanlike motor skills, synthetic sight, smell,
hearing, and touch, plus something that approximates human intelligence.
But the robo-maniacs feel driven to make it happen. "We have an image of
robots as partners, and a machine similar to humans can evoke shared
feelings," says Masato Hirose, 45, a Honda Motor Corp. designer who has
built a line of stair-climbing robots. Atsuo Takanishi, 44, a professor of
robotics at Waseda, says visions of helper bots took root in his
childhood, as he "grew up watching cartoons of Atom Boy," a sort of
robotic Mighty Mouse. At the Massachusetts Institute of Technology Media
Lab, researcher Cynthia Breazeal sounds the same chords. For robots to be
meaningful companions, "they must be socially savvy," says the 33-year-old
researcher. So her creation, named Kismet, is learning to recognize human
emotions--and has a nimble face to express its own moods. "If the robot
wasn't anthropomorphic, it wouldn't work," she says.

Beyond psychology, humans could make good use of mechanical devices that
walk, talk, and think like us but vastly exceed our capabilities in
memory, computational skills, and raw physical strength. Heavy industry
wants such creatures for labor in hazardous environments. The military
wants them to shoulder a soldier's burden on the battlefield. The Japanese
government wants them to help care for Japan's swelling ranks of elderly
people. And Sony figures the best place to launch this revolution is
entertainment. Since dancing and singing robots like SDR don't do anything
essential, says Kuroki, "it's O.K. if they make a mistake sometimes." That
would be only human, after all.

None of Japan's humanoids can brew coffee or do the laundry--yet. But the
next-generation technologies that robo sapiens will need are being refined
today--often in the shapes of other animals. Chances seem good that Aibo
and other robo pets will evolve into Japan's next great export wave. And
following in the toys' wake could be all manner of household robots, from
steel guard dogs to autonomous vacuum cleaners and kitchen "Cuisinbots."
Already in Japan, toylike robots are reminding senior citizens to take
medicine and keep appointments. And future plans call for bundling more
advanced computers and mechanisms into roving maids and butlers. "Robots
are just on the verge of becoming intelligent machines," maintains Honda's

Humanlike robots were stirring imaginations long before the word "robot"
entered the English language in R.U.R. (Rossum's Universal Robots), a 1921
play by Czech writer Karel Capek. In ancient Greece, Aristotle wished for
mechanical slaves that would obey people. In the Middle Ages, people
marveled at mechanical crowing roosters and clockwork automatons. But
those mechanisms were anchored down. For real versatility, robots must be
mobile--and robot companions must walk on two legs. That's a major
engineering challenge.

Bipedal movement is essentially a continuous, controlled fall that's
avoided only by precisely timing each step. People do it without thinking,
so it seems simple. But the computational horsepower needed to prevent a
biped bot from falling on its face is considerable. Early mobile robots
had to drag a cable connecting them to a big computer.

RECHARGING. Now that the underpinnings for smart robots are falling into
place, turning Atom Boy into superbot has become something of a national
obsession in Japan. Researchers in dozens of corporate and academic labs
are racing to develop working models that within a few years may become
our cohabitants and co-workers. The leaders--companies such as Matsushita
Electric Industrial (MC ), NEC (NIPNY ), Sony, and Omron (OMTEY )--are
investing tens of millions of dollars annually in the development of
personal robots. Toyota (TM ) now wants to join this clique. And Honda
(HMC ) has spent roughly $100 million since work on its first experimental
unit, EO, started in 1986.

After dropping behind in information technology, Japan needs a new
industry to recharge economic growth. It's one reason why the Ministry of
Economy, Trade & Industry (METI, formerly MITI) is backing research in
the field. In 1998, METI coughed up $50 million in seed money to nurture
development of a humanoid by 2003. Japan's leaders figure home-care robots
for the elderly will soon be a social necessity: By 2005, 25% of Japan's
population will be over age 65.

Personal robots won't even approach their full potential in this decade,
but robot makers have already fielded many commercial previews. Although
sometimes clumsy and unpredictable, service robots are functioning as
guards in warehouses, delivering hospital food trays, and carrying
documents from one office to another. The Japan Robot Assn. (Jara)
estimates that by 2002, some 11,000 service robots will be deployed, 65%
of them in hospitals and nursing homes. By 2005, give or take a year, Jara
projects that health-care robots will be a $250 million market, then hit
$1 billion by 2010--approaching 10% of the total market that year. As for
personal bots, a panel of industry and academic experts last year
predicted they will be as common as PCs and cell phones within 10 or 15

One of the first robo sapiens on the market will be Honda's Asimo, a
1.2-meter-tall android that resembles a child astronaut. It strides
confidently, climbs stairs, and negotiates corners. It can turn out the
lights, and to show off, walk a figure eight or compete with Sony's SDR
bots on the dance floor. The hitch is that Asimo currently is blind, deaf,
and dumb--and must be remotely controlled. Over the next few months, says
Hirose, it will be equipped with programs and artificial senses that will
render it autonomous. This fall, for an undisclosed fee, Honda will start
renting Asimo to companies and museums for use as a visitor's guide.

PERSONABLE. Another robot about to hatch is Pino, an infantlike android
built by Hiroaki Kitano, an artificial-intelligence (AI) expert who heads
his own government-funded research group, Kitano Symbiotic Systems
Project, in Tokyo. Pino is outfitted with neural-network circuits that
will one day mimic the human brain, thus enabling it to learn to walk and
ape humans.

Sony has already proved that mechanical companions are a promising market.
Now, sales of entertainment robots, it believes, are primed to explode.
These include electronic critters like the $1,500 Aibo, which can learn
tricks and respond to voice commands. The latest model looks like a lion
cub. Over the next five years, an expanding menagerie of mechanical pets
from Sony and others could whet consumer demand. And then, around
mid-decade, Sony's acrobots should hit the market--probably at prices
comparable to a high-end PC. Come the 2010s, predicts Toshi T. Doi, head
of Sony's Digital Creatures Lab, each of Japan's 46 million households
will have two or three robots, including a humanoid.

Researchers everywhere are equally fascinated by the potential of home
robots and cyber-companions. Ironically, some European and North American
labs are counting on biological approaches to help produce mechanical
beings--a tack called evolutionary robotics. The basic idea is to raise a
machine like a child, letting it learn from its own experiences and
sensory impressions, rather than feeding it canned software written by
humans. Hothouses of this research include the University of Sussex in
Britain and Switzerland's University of Lausanne--and MIT and Michigan
State University. At Brandeis University, the Golem Project has even
produced robots that design and build other robots.

MIT's Cog and Kismet are probably the most famous of the self-educated
bots. Cog is the brainchild of Rodney A. Brooks, head of MIT's AI Lab.
This humanoid torso has been learning to interact with its surroundings
and with people since its "birth" in 1993. Cog has mentally progressed to
the crawling-infant stage--although its legs have yet to be attached.
Cog's face is less expressive than Kismet's, but even so, it can be
engaging. Its camera eyes track moving people, and it establishes eye
contact with people facing it. Almost invariably, says Brooks, visitors
and researchers quickly treat Cog as a person, despite its skinless metal

No matter how personable Cog may seem, fulfilling Brooks's dream of a
robot with human-level intelligence remains on the distant horizon. Still,
many experts believe truly smart robots are inevitable, given the
ever-growing power of computer chips. Computing power increases a
thousandfold every 15 years, notes Sony's Doi, so it's only a matter of
time before even little metal pets like Aibo become intelligent. "Robots
will be able to have normal conversations and even gossip with humans," he

At the other end of the "evo-bot" scale are the creeping creations of Mark
W. Tilden, a researcher at Los Alamos National Laboratory. Most of his
buglike robots don't even have a microprocessor brain, just some sensors.
They function on the basis of primitive survival instincts. Yet some
fairly complex behaviors can evolve. For example, a solar-powered
mechanical insect has one purpose: to soak up energy from the sun's rays.
But give it legs, and it gradually learns to walk so it can follow the sun
on its path across the sky. Give it suction-cup feet with attached
brushes, and it will keep a window clean--following the rising sun up the
window, scrubbing as it tries to crawl closer to the light.

Tilden has harnessed similar mindless creatures to clip grass, clean up
dust, and draw draperies. A decade ago, he founded a movement called BEAM
(for biology, electronics, aesthetics, and mechanics) to popularize simple
robots. There's now a worldwide army of BEAM aficionados, including
schoolkids. They stage contests and swap tales of how many robots parts
were scavenged from trash cans and junkyards. "You can build a BEAM robot
for next to nothing, and you don't have to write an elaborate program,"
says Tilden. "But more important, it shows you the power of biology as a

FRUSTRATION. There are a lots of BEAM proponents in Europe--and many
researchers there also share Japan's view of the near-term need for
personal robots. For example, Germany's Fraunhofer Institute has developed
Care-O-Bot, which looks a bit like R2D2, to help elderly and infirm people
maintain independent lifestyles. It can guide and support people who are
unsteady on their feet, run errands around the house, and operate home

In the U.S., though, getting the funds to turn robot research into a going
business has been a problem, laments Joseph F. Engelberger. Widely hailed
as the father of the industrial robot, Engelberger co-founded Unimation
Inc. 40 years ago and created the factory-robot industry from scratch.
After cashing out in 1983, he founded HelpMate Robotics Inc. (HELP ) to
build service robots. His flagship: wheeled cabinets that scurry around
hospitals, distributing medicines and patient records. But Engelberger
always had his eyes on domestic robots because the market potential is
clearly far bigger.

In 1997, HelpMate built a two-armed, wheeled prototype for NASA to
evaluate as a helpmate in space. Engelberger intended to adapt it for home
use, since its touch-sensitive hands and two arms could make beds and
prepare meals for seniors. But it was not to be. He couldn't raise the $5
million he needed to bring it to market, and financial problems forced him
to sell the company in 1999. This lack of interest in robotics puzzles the
Japanese. Many reckon it's because Hollywood often depicts robots as
monsters, whereas the Japanese view them as helpers.

As a result, Japan has amassed a formidable stockpile of robotics knowhow.
It has long been the world's largest manufacturer of industrial robots,
producing twice as many as the rest of the world combined. Last year,
robot shipments, including exports, came to $5.7 billion. Export sales
alone--$3 billion--outclassed total production in every other country.

Furthermore, Japan has an army of young, savvy robotics experts. Close to
half of its 4,500 registered robot engineers are focused on AI or related
disciplines aimed at enhancing robot intelligence. "The level of AI
research in Japan and the U.S. is now the same," asserts Takanori Shibata,
34, who studied under Brooks at MIT. Shibata is a senior scientist with
the governmental Mechanical Engineering Lab in Tsukuba, north of Tokyo.
His creations are plumbing interactions between people and robots. Paro
the baby seal responds to human touch and is a big hit with children at a
university hospital ward, where the cuddly crawler is being tested as a
therapy tool. And Shibata's cat robot will be marketed this year as a
companion for the elderly by Omron Corp.

SONG AND DANCE. Shoestring budgets at Waseda University and other academic
labs nourished Japan's early robot dreams. Now, well-heeled Japanese
corporations are transforming research into commercial products. For
example, Sony cut a deal with Waseda, in effect purchasing its expertise
and hiring Takanishi as an adviser to its SDR program.

Backed by ample resources, corporate engineers are now busily refining
sensors and rewriting algorithms to create more sophisticated machines.
Sony's Ishida and Kuroki developed a new kind of compact, lightweight
motor for their robot's 24 joints. The motors are one reason why SDR can
prance along at 15 meters a minute, keep its balance on a seesaw, do the
twist, and even stand up after falling down.

Honda engineer Hirose turned to biology for help with Asimo. After his son
was born in 1988, he studied his progress from a crawler to a toddler. "I
recorded a lot of videotapes of my son's first years, but they show only
his legs," he quips. The tapes served an important purpose, though. Hirose
realized that nerves in the sole of the human foot are vital for walking.
So he attached sensors to the robot's feet to help Asimo locate the edge
of stair steps and maintain its center of gravity on sloping ground.

As margins dwindle on consumer electronics and industrial equipment,
Japan's manufacturers are desperately seeking new genres of products.
Smart robots that combine a virtuoso entertainer, flawless social
secretary, compliant maid-butler, patient listener, and multimedia
communications system could be just the ticket. Such companions could
record the passing years and regale people with intricately detailed
multimedia stories of family triumphs and personal exploits. Anything that
owners forget, their cyber-companions would remember.

The chief blemish on Japan's bullish outlook is that buyers might expect
too much from the first crop of bots. Early androids may not have enough
finesse to poke around inside packed refrigerators, and people might be
wise not to trust them to iron shirts. "Just because robots can sing and
dance doesn't mean they'll soon do everything," cautions Shigeo Hirose, a
robotics professor at Tokyo Institute of Technology. Aibo the robo pet has
proved they don't have to. Robots can ease their way into our lives, step
by mechanical step.

By Irene M. Kunii in Tokyo and Otis Port in New York

Stay hungry,

--J. R.

Useless hypotheses:
 consciousness, phlogiston, philosophy, vitalism, mind, free will

This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:42 MDT