TIME: Almanac 1993/TIME Almanac 1993.iso/time/092589/09258900.070

Text File | 1992-09-23 | 6.5 KB | 129 lines

                                                                                TECHNOLOGY, Page 79

Time for Some Fuzzy Thinking


An oddball approach to computer science pays off in Japan

By Philip Elmer-Dewitt


    In the pages of Books in Print, listed among works like
Fuzzy Bear and Fuzzy Wuzzy Puppy, are some strange-sounding
titles: Fuzzy Systems, Fuzzy Set Theory and Fuzzy Reasoning &
Its Applications. The bedtime reading of scientists gone soft
in the head? No, these academic tomes are the collected output
of 25 years of mostly American research in fuzzy logic, a branch
of mathematics designed to help computers simulate the various
kinds of vagueness and uncertainty found in everyday life.
Despite a distinguished corps of devoted followers, however,
fuzzy logic has been largely relegated to the back shelves of
computer science -- at least in the U.S.

    But not, it turns out, in Japan. As they have so often in
the past, the Japanese have seized on an American invention and
found practical uses for it. Suddenly the term fuzzy and
products based on principles of fuzzy logic seem to be
everywhere in Japan: in television documentaries, in corporate
magazine ads and in novel electronic gadgets ranging from
computer-controlled air conditioners to golf-swing analyzers.
The concept of fuzziness has struck a cultural chord in a
society whose religions and philosophies are attuned to
ambiguity and contradiction. Says Noboru Wakami, a senior
researcher at Matsushita: "It's like soy sauce and sushi -- a
perfect match."

    What is fuzzy logic? The original concept, developed in the
mid-'60s by Lofti Zadeh, a Russian-born professor of computer
science at the University of California, Berkeley, is that
things in the real world do not fall into the neat, crisp
categories defined by traditional set theory, like the set of
even numbers or the set of left-handed baseball players. In
standard Aristotelian logic, as in computer science, membership
in a class or set is not a matter of degree. Either a number is
even, or it is not. But this on-or-off, black-or-white, 0-or-1
approach falls apart when applied to many everyday
classifications, like the set of beautiful women, the set of
tall men or the set of very cold days.

    To deal with such cases, Zadeh proposed that membership in
a set be measured not as a 0 or a 1, but as a value between 0
and 1. Thus, in the set of tall men, George Bush (6 ft. 2 in.)
might have a membership value of 0.7, while Kareem Abdul-Jabbar
(7 ft. 2 in.) might have a 0.99. Zadeh and his students went on
to elaborate a full fuzzy mathematics, devising precise rules
for combining vague expressions like "somewhat fast," "very hot"
and "usually wrong."

    This mathematics turns out to be surprisingly useful for
controlling robots, machine tools and various electronic
systems. A conventional air conditioner, for example, recognizes
only two basic states: too hot or too cold. When geared for
thermostat control, the cooling system either operates at full
blast or shuts off completely. A fuzzy air conditioner, by
contrast, would recognize that some room temperatures are closer
to the human comfort zone than others. Its cooling system would
begin to slow down gradually as the room temperature approached
the desired setting. Result: a more comfortable room and a
smaller electric bill.

    Fuzzy logic began to find applications in industry in the
early '70s, when it was teamed with another form of advanced
computer science called the expert system. A product of research
into artificial intelligence, expert systems solve complex
problems somewhat like human experts do -- by applying rules of
thumb. (Example: when the oven gets very hot, turn the gas down
a bit.) In 1980 F.L. Smidth & Co. of Copenhagen began marketing
the first commercial fuzzy expert system: a computer program
that controlled the fuel-intake rate and gas flow of a rotating
kiln used to make cement.

    Despite such successes, fuzzy logic was not well received
in the U.S. Scientists pointed out that uncertainty and
vagueness could be represented perfectly well by more
traditional means, like statistics or probability theory. Some
of the criticism bordered on the vituperative, and the tenets
of fuzzy logic were dismissed with terms ranging from "comical"
to "content-free."

    The Japanese, however, showed no such resistance, perhaps
because their culture is not so deeply rooted in scientific
rationalism. Says Bart Kosko, a Zadeh protege and a professor
of electrical engineering at the University of Southern
California: "Fuzziness begins where Western logic ends." In the
early '80s several Japanese firms plunged enthusiastically into
fuzzy research. By 1985 Hitachi had installed the technology's
most celebrated showpiece: a subway system in Sendai, about 200
miles north of Tokyo, that is operated by a fuzzy computer. Not
only does it give an astonishingly smooth ride (passengers do
not need to hang on to straps), but it is also 10% more energy
efficient than systems driven by human conductors.

    Japanese researchers are pursuing more than 100 new
applications for fuzzy logic. Nissan has patented fuzzy auto
transmission and antiskid braking systems. Yamaichi Securities
has introduced a fuzzy stock-market investment program for
signaling shifts in market sentiment. Canon is working on a
fuzzy auto-focus camera. Matsushita has delivered a fuzzy
automobile-traffic controller, and is about to unveil a fuzzy
shower system that adjusts to changes in water temperature to
prevent morning scaldings. And in the strongest endorsement of
the technology to date, the Ministry of International Trade and
Industry opened the Laboratory for International Fuzzy
Engineering Research in Yokohama and called for funding of some
$34 million over the next six years.

    The U.S. is not totally out of the fuzzy picture yet. A
small firm in Irvine, Calif., Togai InfraLogic, has already
achieved several of the goals MITI set for itself, including a
fuzzy computer chip that can perform 28,600 fuzzy-logical
inferences per sec. (FLIPS). And NASA is experimenting with
fuzzy controllers that could help astronauts pilot the shuttle
in earth orbit. The results so far, say NASA officials, are
encouraging, and there is growing interest at such aerospace
firms as Rockwell and Boeing. "The only barrier remaining" to
wider use of fuzzy logic, says Kosko, "is the philosophical
resistance of the West."


-- Norihiko Shirouzu/Tokyo