Computing Goes Everywhere
The dream of "ubiquitous computing" has been around for a while. Now
it's serious enough that a company like IBM is willing to throw $500
million at it.
By Robert Buderi
Illustration by John Craig
Not far from the modest office where, 30-odd years ago, Douglas
Engelbart invented the mouse, multiple-window screens and other
mainstays of personal computing, an SRI International computer
scientist
approaches a mock-up of a white convertible-representing the car of
the
future. He plugs a notepad-sized computer into the dash, and at once
the
vehicle's 1,400-odd computerized systems become accessible through a
simple user interface. Using voice commands, he demonstrates how he
can
request a CD track, link wirelessly to his office to check voice
mail or
have his e-mail read aloud by a speech synthesizer. One message is
from
his refrigerator asking whether he'd like to pick up orange juice on
his
way home. "Show me the grocery stores," he orders the car. The
vehicle
quickly accesses the Internet and relays directions to the nearest
supermarkets.
Shopping done, our motorist arrives at his apartment, where the
Collaborative Home e-Fridge (CHeF) is waiting for the OJ it
requested.
The juice is duly logged in, but when lemonade is removed, the
fridge
announces it's now out of lemonade-and asks whether the item should
be
added to the shopping list. CHeF even knows the pantry contents. So
when
asked to suggest something for dinner, it flashes the recipe for a
chicken dish on its screen: in-stock ingredients are highlighted in
green, those missing appear in red, while absent items already on
the
shopping list are rendered in blue.
Ah, the future of computing. Whether it's with refrigerators, in
cars,
around the office or on the high seas, powerful new systems that you
can
access through words and maybe even gestures-and which will then
scurry
to invisibly do your bidding-are promising to friendly-up the world.
The
dream is called "ubiquitous" or "pervasive" computing-and it's fast
becoming the hottest thing in computer science. The ultimate aim is
to
seamlessly blend the analog human world with all things digital.
That
way, either by carrying computing and communications power with you,
or
by accessing it through an infrastructure as widespread as electric
power is today, you will tap into this world on your terms and in
your
language, not a machine's.
Less than a decade ago, such dreams were confined to far-out future
factories such as SRI, Xerox Corporation's Palo Alto Research Center
(PARC) and MIT's Media Lab. But recent advances in computing power,
storage, speech recognition and especially wired and wireless
networking, coupled with the rise of the World Wide Web, are
bringing
the dream within grasp of the real world. That essential truth
explains
why Microsoft and Intel, which built their fortunes on the
stand-alone
personal computer, are shifting gears toward this new, mobile,
networked
world. IBM has just committed nearly $500 million over the next five
years to study pervasive computing and create the hardware and
software
infrastructure to support it. Other players include Sony, Sun
Microsystems, AT&T, Hewlett-Packard (HP) and just about every
corporate
or university computer lab worldwide.
Uncertainties abound. Fights are under way over competing
technologies
and standards; and no one even knows how many computing devices
people
will want to carry in the future, let alone what type. Still, the
field
is maturing rapidly. Researchers agree more uniformly than ever on
where
technology is headed-or at least on which main paths it's likely to
take. This allows what was previously a hodgepodge of visions and
predictions about the future to now be classified into three broad
technological frameworks: 24/7/360; who, what, when, where; and the
digital companion.
While these categories-signifying the importance of pervasiveness,
awareness and personalization-don't capture every aspect of
ubiquitous
computing, they do describe its essence. And just by walking into
computer labs these days, you get the strong sense that the progress
made in addressing these challenges has computer scientists
convinced a
major breakthrough is within their grasp. "Ubiquitous computing is
viable-and will soon be commercially practical," asserts William
Mark,
SRI's vice president of Information and Computing Sciences. "The
revolution is about to happen."
24/7/360
The widely acknowledged father of ubiquitous computing was the late
PARC
computer scientist Mark Weiser, who coined the term in 1988. Weiser
described a world where each person would share thousands of highly
distributed but interconnected computers. This computing power, he
argued, should blend into the background, hidden from people's
senses
and attention.
In the early '90s, PARC researchers created ParcTab, a handheld
display
that connected via infrared signals to a network computer so
researchers
could access files without being tied to their desktops. Other
trailblazing work took place at the Olivetti Research Laboratory in
Cambridge, England (now AT&T Laboratories Cambridge), which
pioneered
the Active Badge. The badge transmitted an infrared signal that
allowed
people to be tracked throughout a building via wall-mounted
sensors-among other things, enabling phone calls to be forwarded
automatically to their location. And then there was the ultimate
popularizer-MIT's Media Lab. Researchers at this largely
industry-funded
lab spread the word about concepts such as news-gathering software
agents that would tailor each morning's electronic newspaper to an
individual's tastes.
These early steps have now loosed a flood of innovation and promise
at
computer labs worldwide. Today, it is a fundamental tenet of
ubiquitous
computing that computational power and services will be available
whenever they're needed-that's the 24/7 part. And not just
throughout a
building, but everywhere-that's the 360, as in degrees around the
globe.
Under the 24/7/360 umbrella, however, lie two radically different
approaches. One continues the drive to push computational power into
objects with ever smaller footprints-via souped-up laptops,
handhelds
and wearables. The other holds that tomorrow's computing resources
will
not be carried on specific devices. Instead, they will live on
networks.
In this view, much as people tap electric power by plugging into any
outlet, so should applications and files be reachable from any
display
or information appliance-be it in a car, hotel or office. The
network,
to paraphrase the folks at Sun, becomes the computer.
This utility-like model of computing is catching fire at companies
that
build the backbone for the Internet and for enterprise computing
networks-the communications, applications, storage and services
associated with corporate computer systems. Indeed, of IBM's recent
$500
million commitment to pervasive computing, $300 million will go
toward
building an "intelligent infrastructure" of chips, mainframes,
servers,
databases and protocols for supporting the data-rich, mobile future.
Sun's take on this idea is evidenced in its four-year-old Public
Utility
Computing (PUC) project. The aim is to create dynamic virtual
networks,
or supernets. Each supernet would be assigned a public Web address
that
its members contact. After authenticating themselves through a
password
or smart card, users would receive the encryption keys and addresses
for
entering the private supernet-where they could securely retrieve
files
and collaborate in real time. With PUC, there is "no distinguishable
difference between being in HP's conference room or in my office, or
at
home, or at the beach, or in New York," asserts senior manager Glenn
Scott.
PUC technology could also allow organizations to store and retrieve
data
and access sophisticated computational services, such as database
software that analyzes customer trends. Only instead of purchasing
these
expensive systems, companies would pay solely for what they used.
This
might be ideal for small businesses, argues Scott. Imagine a
10-person
operation that wants to tap big accounting software requiring a
high-powered machine that the outfit can't afford. Under the PUC
concept, he says, the firm could simply "rent" the application as
needed, perhaps once a week for 10 minutes. Since PUC works at the
network level rather than inside the software, any application can
be
easily brought into the supernet. This, says Scott, makes it far
more
powerful than the pay-as-you-go systems offered by today's
applications
service providers.
The catch comes in making everything secure. Scott says field trials
last year validated the concept for communications and storage,
which
are mainly concerned with encryption of the data-both when it is
being
transmitted and once it is stored. But providing secure
computation-assuring users their data isn't inadvertently copied,
for
instance-is more dicey. Any solution will likely involve securing
both
hardware and software-a tricky combination Sun is only just
exploring.
Still, Scott believes PUC is the way of the future; and Sun has
filed 13
patents around the technology.
This utility concept looks years ahead-but others are taking more
immediate aim at a scaled-back form of 24/7/360. Since 1998, what is
now
AT&T Laboratories Cambridge has made its Virtual Network Computing
software available free for download. VNC turns any Web browser into
a
remote display for a desktop computer, allowing people to access
files
and applications from just about any device-laptop to PC, Mac to
Palm.
What's more, it works on standard telephone lines and cell
phones-lightening the data stream by transmitting only the bits or
pixels that change from second to second.
It's the same principle as PUC-on a more personal level. The reason
people carry bulky laptops is not to have all their data at hand,
argues
AT&T researcher Quentin Stafford-Fraser. "What you really want to
carry
around with you when you're going somewhere is your environment," he
says. That means your sets of preferences, dates, desktop and so on.
With VNC, he notes, "I can pretty much go anywhere in the world and
be
connected through to my machine that is sitting on the desk here."
The system isn't secure, and it doesn't offer the file-sharing
capabilities of PUC. Still, its cross-platform capability is
compelling-as AT&T researchers found when one corporate user's
network
server crashed while its systems administrator was off camping.
Reached
on his cell phone, the technician was told to return 250 kilometers
to
the office. Instead, he whipped out his Palm Pilot, called up his
VNC-enabled desktop and fixed the problem-all without leaving his
tent.
Stafford-Fraser reports there are as many as 10,000 VNC downloads a
day-with about a million machines running the software. But that's a
blip on the screen compared with what AT&T and others believe might
be
the prime player in 24/7/360 for years to come: the already
ubiquitous
telephone. This idea is embodied in AT&T's VoiceTone project, which
seeks to replace a normal dial tone with an automated version of
yesteryear's know-everything switchboard. "AT&T, how may I help
you?"
the voice tone might inquire. Thanks to speech recognition, speedy
processing, the Web presence of just about everything, and
technologies
such as text-to-speech synthesis, callers can ask for messages and
traffic reports, check the weather and sports scores, or make
restaurant
reservations-all in normal language and without logging on in the
conventional way.
AT&T is developing some of these services itself. However, many will
be
provided through voice services concerns such as Tellme Networks of
Mountain View, Calif., in which AT&T has invested $60 million.
Tellme
and competitors such as Santa Clara-based BeVocal seek to turn
ordinary
telephones into gateways to the Web. At Tellme, for example, callers
dial an 800 number, then navigate the system with spoken commands
such
as "Restaurants," "Boston, Massachusetts," "Chinese." They then get
a
list of candidates-and can even hear Zagat reviews. If they wish to
make
a reservation, they're connected to the restaurant free of charge.
Tellme co-founders Angus Davis and Mike McCue left Netscape to
pursue
the vision of telephone-as-computer-interface. "We were these
browser
guys, and we thought it was cool that there were 150 million Web
browsers," explains Davis, Tellme's director of production. "But we
thought, wouldn't it be really cool if we could build a user
interface
to the Internet that reached two billion people? And that's what
made
the phone exciting."
Who, What, When, Where?
Computing by the billions may be too much to hope for in the near
future. Still, it's already clear that more and more computing power
and
services will reside in networks, and that these services will be
increasingly accessible-through wires and wireless networks, and via
myriad devices. Emerging software technologies such as Sun's Jini
and
Microsoft's Universal Plug and Play promise to allow systems and
services to be accessed no matter what operating system or
programming
language they employ. On the hardware front, Dallas market research
firm
Parks Associates estimates that 18.1 million information
appliances-things like handheld computers and Internet-connected
TVs,
mobile phones, car navigation systems and game consoles-shipped last
year. Nascent wireless standards, such as Bluetooth for short-range
radio communications, will add more flexibility for linking between
devices and networks.
But before even a few folks have the benefit of truly ubiquitous
computing, great strides must be made toward creating technology
that
serves people rather than the other way around. That means objects
and
services must sense and respond to what is going on around them, so
that
they can automatically do the right thing-hold a routine call if
you're
busy, let you know if your flight's delayed, or inform you of a
traffic
jam and suggest a better route. Such feats are increasingly known as
context-aware computing. However, to do this job to the utmost,
networks
must know something about the people using them, often including
their
identity and location. This will force a choice: do people want to
periodically cede privacy in exchange for better service?
A lot of the effort to track people and devices-and coordinate their
interaction-dates back to Olivetti's (now AT&T's) Active Badge
program.
The latest twist is called "sentient computing," which replaces the
infrared-emitting active badges with ultrasound transmitters, dubbed
"bats." Since ultrasound provides far more precise positioning data
than
does infrared, bats make it possible to construct a computer model
that
follows people, objects and their relation to each other. The
computer,
explains researcher Pete Steggles, creates a "circle around me
that's
about a foot in radius-and there's another little circle around this
device. And when the one is contained in the other, then I'm in a
sense
the owner of that device, and appropriate things happen" (see
sidebar,
"Sentient Computing").
In one application, the system is integrated with Virtual Network
Computing. By pressing a button on their bat, users can have their
desktop teleported to the nearest workstation screen. A more
multifaceted use of the technology lies in the lab's context-aware
filing system, which automatically stores the data people
create-whether
they're working on a computer, snapping a digital picture, dictating
a
memo or just talking to colleagues-in a personal time line, from
which
it's easy to integrate information with traditional files and
applications.
For instance, Steggles relates, "The other day I took a bunch of
photos
for a presentation-and the shots were just there in my time line, so
I
could just drag and drop them into my PowerPoint presentation." What
would have taken hours to transfer to his desktop, he adds, took
mere
minutes. Similarly, say you couldn't recall a reference a colleague
gave
when you were talking a few days earlier, and the person wasn't
around
to ask. Because the system also captures the context of what people
do-including who else is in the room with them-the time line makes
it
far easier to retrieve such information as well.
Another way to track objects is through radio-frequency
identification
tags, like those used to monitor livestock. These "e-tags" range in
size
from a grain of rice to a quarter and so can conceivably be embedded
in
everyday objects. Most rely on inductive coupling, like that used in
the
bulkier tags placed on clothes to deter shoplifting. Unlike bats,
e-tags
have no internal power source that needs periodic replacement.
Instead,
a signal from a tag reader induces a current in the implant, which
consists of a coil attached to a silicon chip. Energy captured by
the
coil is stored in a capacitor that powers the chip and causes it to
transmit a unique identifier to the reader. From there, the data is
relayed wirelessly to the Internet or company intranet-summoning
more
information relating to the tagged item.
Last year, PARC researchers e-tagged everything from paper to books
to
copier machines around the lab. That way, anyone carrying a tablet
computer equipped with a reader could access additional information
and
services associated with the tagged item. Say, for example, a person
approached a flyer announcing a lecture. By positioning the computer
near the title, he or she could call up the talk abstract. Holding
it
near the date and time announcement, where a separate tag was
embedded,
would schedule the event in an electronic calendar. Even better,
many
tagged items activated services associated with their physical form.
In
one demonstration, bringing a tagged French dictionary near a
computer
summoned a French version of the English document then on the
screen.
Roy Want, who led the project but has since left Xerox for Intel,
describes e-tags as "an evolution of the bar code. I think in the
future
almost anything that is manufactured and traded will contain an
electronic tag." Such tags, he adds, will link to the Internet to
provide information about the item's origin, history and ownership.
Although a world populated by bats and e-tags promises to extend
computing to almost anything, it does not address one of the biggest
hopes for ubiquitous computing-that sensors, effectors and actuators
can
also be incorporated into devices, making systems capable of both
processing information and responding to it. Former PARC director
John
Seely Brown, for example, foresees a world where millions of
networked
sensors are placed in roadways, using information about traffic to
ease
congestion and thereby "harmonize human activity with the
environment"
(see "Where Have All the Computers Gone?" TR January/February 2000).
The Digital Companion
Illustration by John Craig
While promising to add great utility to people's lives, most
context-aware technologies depend on direct communication between
humans
and a known device or application. In reality, whether at home or on
the
road, people will also need help tapping services unknown to
them-and
with which they won't ever want to interact directly.
Enter a third major aspect of ubiquitous computing: software agents,
or
bots, that root around behind the scenes to find services and
generally
get things done without bothering humans with the details. Many bots
are
already on the market, cataloging the Web for Internet portals or
tracking customer preferences for e-tailers. But a new generation is
at
hand. Some bots are specific to individual devices or applications.
Others are more like executive assistants-looking for bargains,
negotiating deals and rounding up dozens of services into larger,
coordinated actions.
Among the first bots to hit the market could be context-aware
applications that seek to prevent information overload by filtering
e-mail, phone calls and news alerts. Many firms are tackling this
problem. At Microsoft, software agents-under-development make these
decisions based on such factors as message content, the kinds of
communiqués users read first or delete without opening, and the
message
writer's relationship with the reader or position in a company
organization chart. Agents can then determine whether to interrupt
or
not by correlating that information-with the help of desktop sensors
such as microphones and cameras-with whether the person is on the
phone,
busy at the keyboard or meeting with someone. If the person is out,
the
agents can even decide whether to track him or her down via pager or
cell phone.
What puts the technology into the futuristic agent class, however,
is
that it employs procedures based on statistical reasoning techniques
in
order to draw inferences from users' behavior and make its
judgments.
The same techniques enable the system to learn from past experiences
to
get better at its job. Eric Horvitz, leader of Microsoft Research's
Adaptive Systems & Interaction group, says he knows firsthand the
power
of the system, which he has been testing personally-and relying
on-for
months. Last fall, after leaving work early one Friday afternoon to
attend a circus with his family, his cell phone buzzed quietly in
his
pocket with word of an important e-mail message. "The circus started
at
4," he relates, "and at 3:55, I got an urgent note from Bill G.
seeking
feedback on a memo he was putting together." Horvitz was able to
quickly
schedule time to review Bill Gates' memo immediately after the
event-rather than waiting until Monday.
And that's only an opening salvo in the coming agent wars, as
indicated
by SRI's car-to-refrigerator demonstration. Explains researcher
David
Martin, the demo employs a software technology that acts as a
superagent-or facilitator-to orchestrate the services of a multitude
of
other agents. Under this Open Agent Architecture (OAA) framework,
humans
take no direct hand in controlling the fleet of servants sent
scurrying
to do their bidding. They merely express their desires to the OAA
through a microphone or keyboard, by drawing on a display screen or
even
speaking over the telephone-and things get done (see sidebar, "Open
Agent Architecture").
Even this, though, is merely an appetizer for an idea, still without
concrete embodiment, which SRI calls the "digital companion." Much
like
Microsoft's statistically based filters, it envisions agents that
adapt
to human needs-only on a much larger scale, as the OAA facilitator
idea
is extended to include personalized agents that will stay with
people
for years or even decades. Just as a good secretary learns a boss's
preferences and even comes to anticipate his or her needs, so will a
digital companion serve its human masters.
"Think of it as a PDA (personal digital assistant) on steroids,"
relates
SRI's Mark. "It is your assistant, it is your broker to this set of
services and devices available in the network." Your companion, he
says,
will authenticate your identity and pay your bills. It will make
travel
arrangements based on your preferences-and will even see to it that
the
rental car's radio is set to your desires. Can't remember the wine
you
drank at a restaurant last week? Just ask your companion: It will
reference your bills and maybe the restaurant's wine list to find
out.
In short, says Mark, a digital companion will be a person's
"universal
remote for the world."
The ubiquitous-computing vision remains in many senses just that: a
vision. Beyond the immense technological challenges of building a
public
utility infrastructure and creating digital companions loom
mind-staggering issues that run from programming for the networked
world
to real fears of Big Brother-like invasions of privacy. Jeffrey
Kephart,
who heads the Agents and Emerging Phenomena group at IBM's Thomas J.
Watson Research Center in Hawthorne, N.Y., even foresees the
billions of
agents that will soon be out there setting prices, bidding and
making
purchasing decisions as an economic wild card with potentially
immense
ramifications. "What we're talking about is the introduction into
the
economy of a new economic species," he says. "Heretofore we've only
had
humans." He's working to model and study the dynamics of such a
system-and divine ways to avoid price wars and generally help
prevent
things from getting out of control.
No one yet knows the solution to such puzzles-nor are the answers
even
evident in today's mishmash of efforts. All of which means that
truly
ubiquitous computing could still be decades off.
Steadily, though, the major pieces seem to be coming together,
giving
rise to a view among some in the industry that the new day is at
hand.
SRI's Mark is one such optimist. So, too, is Jim Waldo, chief
engineer
of Sun's Jini effort, which, by removing many of the barriers that
exist
between systems based on different operating systems and languages,
marks a big step toward the dream.
"My feeling about the whole ubiquitous computing thing is it's
getting
to the point of almost being a supersaturated solution-and at some
point, the crystal's going to form. And when it does, it's going to
happen really fast," Waldo asserts. "There's going to be lots of
this
base work. It's going to be going nowhere-and all of a sudden it's
just
going to be there."
Robert Buderi is Technology Review's editor-at-large.
Just an email away ......
Justin
---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.228 / Virus Database: 109 - Release Date: 1/17/01
|