|
Truth Is Stranger Than [Science] Fiction
A Review of David Stork’s HAL’s Legacy: 2001’s
Computer as Dream and Reality
In a tradition that provides a welcome relief from the petty infighting
that characterizes much of academic discourse, younger professors sometimes
honor their elders (imagine that!), on retirement or on the occasion of
an advanced birthday, by contributing to a collection of essays. Generally,
the collection deals with aspects of the elder’s work, explaining why it
was important, building on it, extending it, connecting it to other fields
or areas of inquiry. The new book edited by David Stork, HAL’s
Legacy: 2001’s Computer as Dream and Reality, is such a collection
of essays, but the person whose work is honored by the collection is not
an academic but rather the science fiction writer Arthur C. Clarke, now
in his seventies and living in Sri Lanka. And
the birthday being celebrated is not Clarke’s but rather that of a fictional
computer—HAL, the onboard computer of the Discovery mission in the screenplay
on which Clarke collaborated with the director Stanley Kubrick for the
1968 film 2001: A Space Odyssey. The screenplay, later made into
a novel with the same title, was based on Clarke's earlier short story
“The Sentinel.”
In the novel we are told that HAL sprang into life on January 12, 1997.
Stork, chief scientist at the Ricoh California Research Center, where he
does research on pattern recognition by computers, and visiting professor
of psychology at Stanford University, assembled a team of luminaries in
computer research—experts in artificial intelligence, supercomputer design,
computational linguistics, computer chess, philosophy of mind, computer
speech, interface design, and other topics—and asked them to assess how
far we have come, since Clarke’s movie and book, toward making Clarke’s
vision of an intelligent, emotional, chatty, lip-reading, chess-playing,
and finally murderous supercomputer a reality. Stork’s book contains fascinating
essays by and interviews with Murray Campbell, Daniel Dennett, Ravishankar
Iyer, David Kuck, Raymond Kurzweil, Douglas Lenat, Marvin Minsky, Donald
Norman, Joseph Olive, Rosalind Picard, Azriel Rosenfeld, Roger Schank,
David Wilkins, and Stephen Wolfram, as well as contributions by Stork and
by Arthur C. Clarke himself.
It is not surprising that Clarke should be so honored. There
are two kinds of science fiction writer. One kind, who might more properly
be called a writer of science fantasy, creates wildly improbable stories
that break all the rules known to science today. An example of a writer
of this kind is Ray Bradbury, who in one story, for example, writes of
a person who becomes obsessed by the fact that he has an internal skeleton
and finally has the monstrous thing removed. As Stork points out in his
introductory essay, most screen science fiction is of this kind. Ships
well lighted from all angles scream through space with stars zipping by
them. They blast other ships with laser beams. The other ships blow up.
Smoke billows, and the debris falls toward the bottom of the screen. Of
course, as Stork points out, none of this would actually happen in space.
Stars are too far apart to appear to move relative to a ship. A laser beam
would not be seen in a vacuum in which there was nothing for the beam to
bounce off. There is no atmosphere for smoke to billow in and no gravity
to pull debris downward.
In
contrast, Clarke is the sort of science fiction writer who tries to extrapolate
from real science and to shed light on what might actually be possible
in the future, and Clarke can claim credit for one truly important scientific
achievement. It was he who conceived of the idea of geosynchronous satellites—ones
that orbit at such a speed that they remain in the same position relative
to the earth and so make possible modern global telecommunications. Because
Clarke was careful to make use of real science in his book, looking back
on the predictions in his book and movie is instructive. We can learn a
lot about the validity of scientific prognostication by looking at where
Clarke got it right and where he got it wrong.
How Close Are Today’s Computers to HAL?
Obviously, we do not have, today, anything like the computer that Clarke
envisioned. The artificial intelligence initiatives heralded with such
optimism by Alan Newell, Herbert Simon, John McCarthy, Marvin Minsky, and
others in the 1950s proved vastly more complex than anyone at the time
imagined. Teaching
computers to do difficult tasks such as making medical diagnoses or predicting
where one might find oil proved to be much easier than teaching them to
do so-called simple tasks such as recognizing faces or communicating in
English or Japanese. This is because these “simple tasks” are actually
astonishingly complicated. They seem simple to us only because they are
carried out automatically, below the level of consciousness, by the society
of incredibly complex minicomputers in our brains. Still, as the essays
in Stork’s book point out, we have made significant progress toward creating
computers with the astonishing characteristics of Clarke's HAL.
In “Could We Build HAL? Supercomputer Design,” David Kuck explains that
“To be as large and powerful as he is described, HAL would have to be a
parallel system,” and today we have, in fact, built large, massively parallel
computers that actually have greater memory storage capacity than Clarke
predicted:
One
concrete number given in the novel describes the memory unit Bowman pulls
out as a “marvelously complex three-dimensional network, which could lie
comfortably in a man’s hand yet contained millions of elements.” . . .
This was large for the time but inadequate for any laptop today. The very
conservative nature of Clarke’s prediction is underscored by today’s commodity
memory technology; even a modern 8-MB PC memory contains several hundred
million transistors. (36-37)
In 1997,
the supercomputer Deep Blue, developed by IBM, defeated the chess grandmaster
Garry Kasparov, thus becoming the world’s foremost player of the game.
In “An Enjoyable Game: How HAL Plays Chess,” Murray S. Campbell, an IBM
researcher and one of the members of the team who developed Deep Blue,
analyzes the brief scene from the movie in which the computer HAL defeats
astronaut Frank Poole. Campbell points out that in the movie, HAL makes
a nonoptimal but “trappy” move that tricks Frank into responding with a
move that causes him to lose. Today’s chess-playing computers do not play
in that way, which requires a lot of real-world knowledge. Instead, Deep
Blue relied upon massive computation of positions that resulted from particular
moves: “Deep Blue is capable of searching up to two hundred million chess
positions per second,” Campbell notes, a fact that “prompted Kasparov to
comment that ‘quantity had become quality’” (86).
In “The Talking Computer: Text to Speech Synthesis,” Bell Laboratories
scientist Joseph P. Olive explains, in considerable detail, the complexities
of and approaches to teaching computers to produce intelligible speech
from written text. Olive summarizes as follows:
As we
near the year 2001, do we have a computer that sounds like the voice of
HAL portrayed by actor Douglas Rain—personable, warm, emotional, human-sounding?
The answer is no, not yet.
At Bell Laboratories we have developed a text-to-speech synthesizer
that is highly intelligible in several languages, including English, German,
French, Spanish, Russian, Chinese, and Navajo. . . . [Y]et, although capable
of both reading or generating such complex text as e-mail or newspaper
stories, the synthesizer does not replicate the human voice. It has a distinct
“machine” sound. (124)
The biggest problem
is giving computers the real-world knowledge that would be necessary for
them to understand the contexts in which sentences appear and so be able
to apply the appropriate rules to make the speech sound natural. As Olive
explains, “A computer can only perform tasks requiring very limited understanding.
It can maintain a dialogue about ordering a pizza but not about a subject
matter that has not been previously define” (125) .
Another expert in computer speech, Raymond Kurzweil, addresses this
problem in “When Will HAL Understand What We Are Saying? Computer Speech
Recognition and Understanding.” Here are Kurzweil’s predictions:
Based
on Moore’s law [Moore's law is the prediction by George Moore, a founder
of Intel, that each year the cost of integrated circuits would halve while
the number of transistors on them, and thus the processing power, would
double], and the continued efforts of over a thousand researchers in speech
recognition and related areas, I expect to see commercial-grade continuous-speech
dictation systems for restricted domains, such as medicine or law, to appear
in 1997 or 1998. And, soon after, we will be talking to our computers in
continuous speech and natural language to control personal-computer applications.
By around the turn of the century, unrestricted-domain, continuous-speech
dictation will be the standard. An
especially exciting application of this technology will be listening machines
for the deaf analogous to reading machines for the blind. They will convert
speech into a display of text in real time, thus achieving Alexander Graham
Bell’s original vision a century and a quarter later. (161)
Kurzweil points out that even today’s supercomputers do not have anything
approaching the capacity of the human brain, which has “about a hundred
billion neurons, each of which has an average of a thousand connections
to other neurons” (163). Advances in circuit design, such as creating
three-dimensional circuits, will increase the capacity of computers, but
even more important is achieving breakthroughs in architecture, in the
arrangements of circuits. Kurzweil
suggests that in the near future we might “reverse engineer” the brain
to create a computer that has the same architecture—literally scanning
a brain “to ascertain the architecture of interneuronal connections in
different regions” (165). This suggestion then leads Kurzweil to some fascinating
speculation about the possibility of creating a duplicate of someone’s
brain during the next century!
In “From
2001 to 2001: Common Sense and the Mind of HAL,” Douglas Lenat emphasizes
the importance of building in a computer a real-world knowledge base, something
that he and his colleagues at the Microelectronics and Computer Consortium
have been doing with a program called CYC. The idea is to create a means
for representing knowledge and to “prime the knowledge pump” of a computer
by providing it with the millions of bits of information that a person
knows, such as the facts that “Napoleon died on St. Helena” and “Wellington
was greatly saddened.” Knowing such things, a computer will be able to
infer, as a person does, “that Wellington heard about Napoleon’s death,
that Wellington outlived Napoleon, and so on” (203). The availability of
such common sense is a large factor in our ability to understand language
and to function intelligently. Lenat’s program “to bring a HAL-like being
into existence’ consists of three steps:
1. Prime the pump with the millions of everyday terms, concepts, facts,
and rules of thumb that comprise human consensus reality—that is, common
sense.
2. On top of this base, construct the ability to communicate in a natural
language, such as English. Let the HAL-to-be use that ability to vastly
enlarge its knowledge base.
3. Eventually, as it reaches the frontier of human knowledge in some
area, there will be no one left to talk to about it, so it will need to
perform experiments to make further headway in that area. (203)
Stork’s collection contains many other fascinating essays that provide
a unique opportunity to explore the cutting edge of current computing technology
and to find out what leading experts foresee for the future. It is a must
read for anyone who is interested in how real science has a way of inevitably
transcending the wildest of science fiction.
Reference
Clarke, Arthur C. 2001: A Space Odyssey. London: Hutchinson/Star,
1968.
---. The Sentinel (Masterworks of Science Fiction). New York:
Berkley, 1986.
Stork, David, ed. HAL’s Legacy: 2001’s Computer as Dream and Reality.
Cambridge, MA: MIT P., 1997.
|
