by
David G. Messerschmitt
Department of Electrical Engineering
and Computer Sciences
University of California at Berkeley
Invited paper in special issue on "Impact of Information Technology",
Technology in Society, Elsevier Science, Ltd., to appear.
Acrobat version (best for printing). Copyright(c) 1996, Regents of the University of California. All rights reserved.
ABSTRACT
The physical world has many ways of representing information, including
for example a molecular structure (as in DNA), the swinging of a pendulum
(in a clock), or as a pressure wave in a gas (as in human speech). The modern
computer and telecommunications world have agreed on a single elemental
representation of all information: the bit. A bit is an abstract
atom of information that can assume one of two possible states: zero or
one. A collection of bits can represent an arbitrarily large number of states.
When he developed the "information theory" in the 1940's, Claude
Shannon developed the notion that any information can be represented
as a collection of bits, including text, data, music, speech, pictures,
and video. What is remarkable is that these entities do not have a common
representation in the physical world. Text is composed of icons on
a page, music and speech as pressure waves in a gas, and pictures and video
as light patterns. Yet, technologists have arrived at a common representation.
Following Shannon's lead, all information in computing systems and modern
communications systems adopt a "collection of bits" as the common
representation of all information whether it be the numbers in a calculation,
the speech in a telephone conversation, or a television picture. A communication
and storage infrastructure that carries or stores collections of bits becomes
readily applicable to any form of information. This is not to say the collection
of bits is a lingua franca for communications in the technological world,
any more than the human's use of sound as a representation for their thoughts
allows them to communicate without a common language. One must have a knowledge
of the syntax and semantic content of the collection of bits to make use
of them. An audio and a video signal look superficially similar when represented
by bits, but their interpretation is quite distinct.
The modern computer incorporates many inventions, but arguably the most
significant among them is the idea of cleanly separating physical realization
from functionality. This is accomplished through programmability:
the application (intended use) is defined in software, which is a sequence
of instructions represented, naturally, as a collection of bits. The computer
executes this series of instructions. The functionality is thus embodied
in the software at the time it is used, and not in the hardware at the time
of manufacture. This separation of function from physical realization is
not unique to the computer world; for example, a child's profession is not
defined at the time the child is born. However, the computer has great flexibility
in that its functionality can be changed quickly and easily
by software substitution.
Computers were once expensive, and thus few in number and centralized. The
advent of the desktop computer in the early 1980's was significant because
it freed the user from the bureaucracy of the centralized computer center,
allowing him or her to take direct control. From the perspective of the
industry, the desktop computer became a platform for applications.
By "platform", we mean the basic physical realization, plus bundled
software that performs routine and universal tasks (the "operating
system"), that provides a foundation for new software-defined applications.
Once there is a large installed base of these platforms (because many users
have purchased them), entrepreneurial companies, large or small, can develop
new applications and sell them into an economically significant installed
base.The barriers to entry for applications development are dramatically
reduced, and the computer center bureaucracy is eliminated. Thousands of
new companies were started to developed desktop computer applications, and
the rate of innovation was expanded dramatically.
The key technical advance in the convergence of computing and telecommunications
has been the computer network, such as the famous Internet. A computer
network enables any computer to transport an arbitrary collection of bits
to another computer, exploiting the consistent abstract representation of
information used in telecommunications. The implication is that applications
are no longer restricted to a single computer, but rather a single application
can share two or more computers connected by an appropriate network, exploiting
the communications among them. There are many ways to leverage this capability,
but for our purposes the most important is that the computer becomes a useful
tool for telecommunications among people, rather than merely a tool for
processing data.
The telecommunications industry during its century-long existence has relied
on specialized terminals such as the "telephone" or "video
conferencing set" or "radio" or "television set".
In each case, these are specialized boxes dedicated to a single telecommunications
function. The computer industry went through a similar stage of evolution,
with dedicated limited-functionality products such as "calculators"
and "word processors". The significance of the networked computer
is that it becomes a programmable terminal for telecommunications
applications, including many of the most familiar applications like telephony
and television. It brings to telecommunications many of the same benefits
of programmability long enjoyed by computing. As a result, the barriers
to entry for new purveyors of telecommunications applications are reduced,
and the user will have greater choice. Any applications developer can market
directly to users without the intervention of the telecommunications service
provider. The networked computer thus offers the telecommunications application
similar advantages and capabilities as the desktop computer offered the
computing application.
If an application is defined in software, which is represented in the same
way as any other information -- a collection of bits -- that application
can be transported on demand from one computer to another. A networked computing
infrastructure thus represents an important new infrastructure for the distribution
of (software defined) products. The barriers to entry for new applications
are reduced further, since the functions of the manufacturer, wholesaler,
and retailer are eliminated. Reflective of this, there are a remarkable
number of public-domain applications for which the developer does not even
request remuneration.
Communications networks suffer from a problem economists call "network
externality". That is, the network is useful to an individual only
to the extent that there are other entities on the network offering some
reason to communicate. Network externality presents a tremendous obstacle
to commercial exploitation of a network. Who, for example, will be the first
person to buy a new video conferencing product, given that there is no one
else with whom to conference? Distribution of applications over the network
(in the form of software definitions) will neatly bypass this problem, since
a new application can be dynamically loaded into the shared platform consisting
of network and programmable terminals, and thus lead to a vastly greater
rate of innovation in telecommunications applications.
The collection of mutually networked computers clearly represents an important
infrastructure for society, no less significant than our telephone system
or our advanced transportation systems.
The development of writing systems allowed people to communicate accurately
from one point in time to another, as well as one geographic location to
another, by placing thoughts on the semi-permanent paper medium. This allowed
people to accurately communicate their knowledge, ideas, and activities
to future generations, or to their contemporaries geographically far removed.
The printing press led to mass communications, or the ability to communicate
to many people. Writing and printing incur a substantial delay from creation
to receipt of information due to the delay in physically transporting the
paper medium. Writing allows two-way communication (the letter), albeit
with a considerable delay in receiving a response, due again to physical
roundtrip transportation.
Telephony offers complementary strengths, such as instantaneous two-way
communications. However, it has limitations, such as the absence of archival
capability and the inability to handle other media like graphics, images,
and video. More recently some of these capabilities have appeared in limited
form in voiceband data modems and facsimile machines.
In contrast, incorporating the computer as a telecommunications tool allows
complete flexibility in the medium of communication, including text, images,
video, audio, or mixtures thereof. Under the banner of multimedia,
computers can store, display, process and transport audio and video signals
over the network. We can expect computers to routinely have video cameras
as well as microphones. When such computers are networked, they can serve
as the basis of traditional telecommunications applications like telephony
and video conferencing.
More interesting possibilities arise from the integration of distributed
information processing and telecommunications into a single networked computing
platform. Where two people are collaborating on some task via a video conference,
they can also arbitrarily share information, or have shared access to information
databases. For example, two or more people collaborating on a document will
be able to look at it and jointly edit it while they discuss it. Alternatively,
one person can voice- and video-annotate a document, and send it to another
person for his or her editing. Or perhaps the document itself should
be partially represented by voice and video rather than completely by text.
Speech recognition (the conversion of speech to text) and voice response
(conversion from text to speech) will move from laboratory curiosities to
widespread application in human-to-computer interfaces.
Less obvious but no less important, since a computer program is nothing
but information represented by a collection of bits, documents can also
incorporate programs. That is, a document can have executable as well as
passive content. For example, a document may query its viewer as to his
or her interests, and tailor its visible representation of content accordingly.
The history of computing, as well as paper-based information processing
before it, is one of the users accommodating to the medium. Writing allowed
people to communicate across distance or through time, but constrained them
to use text and precluded other media such as speech and video. Computers
used to similarly constrain users to textual representations. Computers
or typewriters could generate paper documents, but the only way to transport
them was by the relatively slow post office. In contrast, the networked
computer infrastructure should adapt itself to the preferred modality of
the user. If a "video document" is preferred to a "text document",
the local computer will allow the user to author it, and the network will
allow the user to send it virtually instantaneously to others. If the user
prefers to input a text document via speech rather than a keyboard, fine.
Overall, the goal should be to make the technology of processing documents
and information more accommodating to people.
The networked computer provides an important consistency of information
in both time and space. Because the transport of information
is quick, even on a global scale, the information available to anyone, anywhere,
at any point in time is identical. When information changes, its latest
rendition is available to anyone, anywhere, simultaneously. These forms
of consistency have a substantial impact on many areas of human endeavor,
such as knowledge generation and commerce. As one example, scientists once
communicated their latest results by letters -- with perhaps weeks to months
of delay -- but can now communicate by telephone or electronic mail with
insignificant delay. Since the process of science is largely one of building
incrementally on the results of other scientists, the delay in communication
alters the rate at which new knowledge is uncovered. As another example,
an auction marketplace depends critically on all participants knowing the
actions of other participants as a guide for their own action. Networked
computing allows such a marketplace to have a global reach.
Strictly speaking, there are some limitations imposed by the speed of light.
Certainly the delay due to the speed of light would be a major issue in
communicating with extraterrestrial civilizations, but even back here on
earth the round-trip delay can be as large as a third of a second. This
turns out to be a minor annoyance for interactive applications (like telephony
or video conferencing), but could be a significant impediment in a global
automated auction marketplace, for example. An exciting workaround to this
problem is intelligent agents, which are executable programs transported
in lieu of messages or queries. To participate in an auction, for example,
I might send such an agent to negotiate on my behalf (by being executed
in the remote computer), avoiding speed-of-light delays associated with
multiple roundtrip transactions.
The old style of publishing based on printing and distributing copies
of a work requires a significant investment. This discourages the publication
of works with a limited audience, or unconventional works that don't attract
the interest of a mainstream publisher. Similar obstacles have faced the
artist, musician, or performer in making their works available to a broad
audience. In the world of networked computing, publishing requires no incremental
investment for the owner of a networked desktop computer. The World-Wide
Web today allows virtually any such user to author a work in any electronic
medium (text, images, audio, video) and make it immediately available to
all networked computers worldwide. In principle, this virtually eliminates
publication and distribution as an obstacle to new artists, performers authors,
and information providers. This is another example of the reach of technology
in eliminating the "middleman".
The dictionary definition of community is "a group of people with
a common characteristic or interest living together within a larger society"
[1]. Observing what is happening within the Internet,
the dictionary should remove the words "living together", if that
implies a geographic locality. Increasingly, global communities are forming
around special interests, whether they be hobbies, professions, specialties,
or politics. While telephone network directories are organized by name and
location, in the Internet the directories are largely organized by interest
or function. This makes is easy to locate people who share a common interest,
regardless of where they reside. Further, it allows any member of the community
to make their ideas or wishes available in a form that is available to the
entire community, and receive responses back from interested members of
the community, similar to the "want ads" of the geographically
based community newspaper.
Traditional telecommunications in the form of telephony and television have
had a significant impact on political structures, for example making it
more difficult to sustain totalitarianism. However, their reach has been
limited by restricted access to broadcast communication mediums like radio
and television. (Indeed, a primary technique for maintaining totalitarian
political structures is precisely such restrictions.) Networked computing
cleanly removes this restriction. It has the potential to significantly
shift the political organizational paradigm from one based on geographic
locality to one based on shared interest or agenda. Clearly this has already
happened in global commercial enterprises: successful large companies are
today presumed to be global in reach, and their country of origin is less
and less identifiable.
The market system of commerce is based on the concept of ownership of
property. The essence of ownership is the owner's ability to control the
use of property. Information (where we include such things as software and
audio and video performances) is an increasingly important commodity in
world commerce, and yet the privileges of information ownership are difficult
to exercise because of the inherent ease with which information can be copied.
When ownership becomes ineffective, the market incentives to generate and
enhance information are negated.
This problem arose first with the photocopier and the tape recorder, but
the accumulation of noise and distortion in the process of copying limits
the potential for mischief. When information is represented by a collection
of bits, it can be copied indefinitely without degradation. All we have
to do is faithfully reproduce the bits. Worse, networked computers allow
us to distribute these copies widely at lightening speed. Attempts to address
this problem thus far have been ad hoc and not very effective [2].
Ownership implies use only with the permission of the owner. There are encryption
techniques to insure that this permission is required, and computer networking
provides the means for users to request and owners to grant this permission.
However, this becomes significantly invasive to the user, and requires a
significant infrastructure for the distribution of secret "keys".
In a market system, exploiting information also requires payment for its
use. "Digital cash" will be available for a payment system, although
this also requires an infrastructure.
On the positive side, networking allows "metered" or "usage-based"
access to information. The most effective model for the pricing of goods
has been based on the utility to the consumer, but this is difficult to
implement in older information technologies. Books and software programs
are typically fixed priced because of the impossibility of monitoring their
use. This discourages the consumer who wants to "try out" a book
or program, and gives the consumer who uses the information resource extensively
a free ride. With networking, information access can be priced based on
actual usage, by coupling that usage with communication to the owner.
Even though we are in the early stages of developing a networked computing
infrastructure, some radically new business models are beginning to emerge.
In a "give away the razor blades in order to sell the razor" paradigm,
it is increasingly common to distribute a World-Wide Web browser or document
viewer application over the network for free, in order to sell and derive
revenue from a compatible server application. It is becoming clear that
networked computing will ultimately have a profound impact on the ownership,
distribution, and commerce of information, as well as the computer and telecommunications
industries themselves.
The ease with which information can be published over the network raises
many issues of trust and authentication. One role of a publisher is to validate
the authenticity of the author and work. In contrast, if anyone can publish
their information to a large audience, what assurance is there that the
information is correct, or the source of the information is who they claim
to be? Forgery or misrepresentation is an age-old problem, but becomes a
serious threat in a networked environment. There are fortunately partial
technological solutions to this problem, enabling for example one to determine
that a document was generated by a certain party, at a certain time, and
has not been subsequently modified.
Improved telecommunications results in a rapidly increasing rate of new
knowledge and information generation and accumulation [2].
Unimpeded publication will result in more information providers, and networked
access makes it easier for the user to access the global information store.
From an individual's perspective, an information overload will likely be
the result.
There are partial technological responses to this problem, but they have
so far fallen short. We can imagine the technological analog of the publisher,
who selects useful or valid information at the source, or the librarian,
who finds useful information on behalf of a user. Today the networked computer
can systematically search for desired information, although the results
are rarely precise. The challenge here -- and it is a severe one -- is that
the system must understand the information to be fully effective
[4]. It is likely that the role of the (human)
publisher and librarian -- increasingly assisted by automation tools --
will be preserved for the foreseeable future, and in fact will become more
critical functions. The particulars of their role will be changed, of course.
For example, the publisher will maintained networked indexes of useful and
authenticated information sources for user. The librarian will cruise the
network rather than the card catalog or the stacks for relevant information
sources. Libraries will shift from paper-based to electronics-based, and
from geographic-based to discipline-based.
The computer industry activity in platform and applications can only
be described as chaotic. This presents a serious challenge in the use of
electronic storage as a medium for the archiving of information over long
periods of time. Reliably storing collections of bits and guaranteeing their
retreivability decades from now is surely technically feasible. However,
these bits are useless without knowledge of how to interpret them, and this
knowledge is likely to be lost over time. Storing programs embedding this
associated language along with the information seems like a solution, but
it is unlikely that such programs would run on the computer systems and
operating systems decades from now. Electronic storage thus appears to be
inadequate to the task of archiving information over long periods of time.
Before relying on electronic storage for our civilization's historical archives,
this problem must be solved.
The increasing globalization of commerce, largely enabled by modern telecommunications
and transportation, dramatically increases the number of people with whom
we may have occasion to communicate. Emerging technologies like personal
communications are a double-edged sword. While they make communications
easier, they also impede personal effectiveness by generating constant interruptions,
and in their extreme have the potential to decrease the quality of life.
To understand this phenomenon better, it is useful to examine our own makeup.
At the risk of over-simplifying an elegant picture, we can divide behaviors
into several categories:
In this context, it is helpful to divide human communications into two
classes:
Effective individuals prioritize and selectively perform tasks. A similar
approach applies to communications. The problem is arising in that an excess
of intrusive demands for immediate communications causes people to rely
too much on deferred communications. For example, if two people rely on
voice mail systems to answer their telephone calls, and only personally
originate calls, they will never manage to participate in an actual phone
conversation, no matter how hard they try! On the other hand, answering
the phone offers no way to prioritize. Everyone hiring a personal assistant
to screen communications is not viable for any but the wealthiest individuals.
The effectiveness of immediate communication is undercut to some extent
by time zones and the sleep and work habits of people. If the person I want
to communicate with halfway around the world is sleeping while I am working,
and vice versa, immediate communication becomes difficult to accomplish.
With the globalization of commerce, these problems become more serious,
and better technological remedies are needed. Technological solutions to
the prioritizing of communications is actually considerably more difficult
than the screening of information, because the priorities and intent of
all participants matters. The intelligent agents mentioned earlier
may be able to manage the details of prioritizing communications by negotiating
with similar agents representing parties desirous of communicating with
us. Similarly, we can imagine such agents scheduling immediate communications
sessions. The goal should be to make such negotiations unobtrusive, and
limit people to the more important task of the communication itself.
Generated with Harlequin
WebMaker