http://www.nytimes.com/library/tech/99/04/biztech/articles/05mit.html
April 5, 1999
Project Aims to Unhitch Computing From PC Harness
By JOHN MARKOFF
CAMBRIDGE, Mass. -- For David Clark, an MIT computer scientist,
research is like "an expedition into the future."
With the faculty and students at the Laboratory of Computer Science
at the Massachusetts Institute of Technology, Clark, one of the lab's
top managers, is about to embark on just such a journey -- one meant
to liberate computing from the PC-centric world it has occupied for
the last two decades.
On April 12, scientists here plan to introduce Oxygen, an ambitious
$40 million, five-year research project that is being financed by the
Defense Advanced Research Projects Agency, or DARPA, the Pentagon's
research arm. The objective is as broad as it is audacious: to
reinvent all facets of information technology, from chips and
software to computers and networks. And in the Oxygenated future, a
desktop computer would be largely beside the point.
In this digital new world, a person could know if a colleague working
elsewhere was reachable at her desk, because sensory-based computers
in the wall would know. A handheld point-and-click pointing device
could be used to remotely command every household appliance to do its
bidding. And instead of typing on keyboards or stabbing at a screen
with a stylus, a person would simply tell any computer what to do.
Or at least such concepts are the vision. In reality, the Oxygen
project, Clark said, must walk the exceedingly thin line between an
unrealistically futuristic "Star Trek" notion of computing and an
overly timid short-armed reach for the easily attainable.
"You have to quantum tunnel from 'It's too soon' to 'It's too late,"'
he said. "You've got to have guts."
It might be easy to dismiss such talk, were it not for the Laboratory
of Computer Science's track record over the last three and a half
decades. True, it has simmered at a lower publicity temperature than
its bubbly academic sister, the MIT Media Laboratory. But since its
creation in 1964, the LCS, as it is known, has been associated with
many of the most important developments in computing.
The lab's many historic fruits include word-processing and
spreadsheet software. And though it was invented elsewhere by Robert
Metcalfe, Ethernet, the standard architecture of today's PC networks,
was subsequently refined during the 1970s as Metcalfe wrote his Ph.D.
dissertation at the lab.
It may be an MIT effort predating the lab by a year -- Project MAC,
set in motion by a $2 million DARPA grant in 1963 -- that offers the
best model for a project like Oxygen and the impact it could
conceivably have on the next generation of computing.
It was Project MAC, after all, that redefined the computing world by
making it possible for many users to simultaneously share a single
computer. From Project MAC sprang innovations like an early e-mail
system and the concept of shared information utilities, the
forerunners of today's online communities and even the World Wide Web.
Project Oxygen, which will be formally kicked off at the laboratory's
35th anniversary next week, is the brainchild of the lab's director,
Michael Dertouzous, a computer scientist whose own career traces to
his days working on Project MAC as an MIT student. Intent on putting
people first, Dertouzous is determined to chart an information
technology future in which computers recede into the background.
"We want to liberate ourselves from the tyranny of 'going to the
computer,'" he said.
Oxygen, seen as a proof-of-concept testbed for technologies that
others might turn into actual products, will be based on several
components that the lab is already developing.
The first will be called Handy 21 and will be a portable device with
a small screen, a video camera, a Global Positioning System receiver
and a powerful computer. Combining the functions of a cellular
telephone, two-way data radio, television set, beeper, handheld
computer and intelligent remote-control pointing device, the Handy 21
will rely heavily on advanced voice-recognition technologies that
have been developed at the lab.
Speech recognition -- the ability for machines to understand spoken
instructions -- is a vital part of the Oxygen project. A group led by
the MIT speech researcher Victor Zue has developed an approach
significantly different from that now being pursued by most others in
the speech-recognition field.
Zue's Spoken Language Systems group has developed systems with
powerful speech-recognition capabilities -- in large part because
they focus on relatively narrow subject areas, like the weather,
airline reservations or local traffic.
"I have to confess that I don't know how to build HAL," Zue said,
referring to the computer in "2001: A Space Odyssey" with the
complete conversational repertory of an eerily dispassionate human
being.
The aim of the Spoken Language Systems research group is to achieve a
limited, though seemingly conversational, voice input system by
stitching together many different recognition systems with
vocabularies based on narrowly defined subjects.
Zue cites three factors making voice recognition more important as a
computer interface: Computers are becoming increasingly mobile;
people love to talk, and the shrinking size of computers is making
the conventional keyboard a large and ungainly component.
Much of the design effort will revolve around the idea of delegating
tasks to machines. For example, Zue said that instead of merely
inquiring of the Oxygen system when a plane is scheduled to arrive at
the airport, the user, planning to meet the flight, might also give
this command: "Call me half an hour before Flight 116 lands."
Besides the Handy 21, the Oxygen system will be based on another
building-block computer that the researchers call Enviro 21. These
devices, which would be powerful versions of the Handy 21 system,
would be embedded in office walls, car trunks or basements at home.
The systems would have extensive sensor networks meant to enable them
to monitor the state of the environment -- knowing whether an office
door is open, for example, or if the car trunk is unlocked or the
basement sump pump is operating.
And besides merely sensing a situation, the Enviro 21 would have the
intelligence to instruct mechanical systems to close the door, lock
the trunk or reset the pump. Or by the same token, if an office door
is open, a distant colleague might infer that the occupant would not
mind being interrupted with a phone call.
Both the Handy 21 and the Enviro 21 will be based on a processor chip
design developed by Anant Agarwal, a computer architect at the lab.
Working with IBM, Agarwal's group is now designing a processor
architecture known as Raw. Unlike earlier chip designs which are
based on predetermined instruction sets, Raw processors subject even
the most minute, or "raw," circuitry of the processor to
customization by software designers.
Raw is a gamble, because it would be harder to program than
conventional processors. But it might have a huge payoff if the
design proves workable, by offering markedly faster data processing
and the potential for large numbers of the chips to be custom
programmed to work in powerful concert.
"A lot of our experiments will be high risk, and they will fail,"
Dertouzous conceded. "But there is also low-hanging fruit."
Project Oxygen is the clearest example yet in a new direction in
federal financing of information technology research patterned after
previous research projects that upset the dominant computing
assumptions of their day. A notable example from the past is Xerox's
legendary Palo Alto Research Center during the 1970s.
"This is all about investing in long-term, high-risk research," Tom
Kalil, President Clinton's special assistant on economic policy, said
of Oxygen. "What's motivating this is our understanding of the impact
of government investment in information technology research in the
1960s and 1970s. Innovations from that investment are driving the
economy today.
"The administration," he said, "really wants the research community
to invest in the future and swing for the fences."
Besides speech recognition and Handy 21 and Enviro 21, the other
basic technology component of Oxygen would be a computer network
capable of linking the other systems together and also connecting to
the Internet.
With these, as well the underlying software that supports the basic
technologies, Dertouzous hopes that Project Oxygen can lead to
computing systems that bring a meaningful increase in human
productivity.
"Our overarching goal," he said, "is to enable people to do more by
doing less."