Computers as saviors
Pioneer foresees technology creating `alien intelligence' to satisfy our needs
Brad Lemley
NEW YORK TIMES SERVICE
http://www.thestar.com/NASApp/cs/ContentServer?pagename=thestar/Layout/Article
_Type1&c=Article&cid=998863510285&call_page=TS_@Biz&call_pageid=971794782442&c
all_pagepath=Business/@Biz&col=971886476975
In 1977, Elvis died, Gerald Ford left the White House and Bill Gates still
banged out business correspondence on a typewriter. Most Americans had never
seen a personal computer, much less a laptop.
Meanwhile, James Martin looked into the future and saw the Internet, as well
as computers in everyone's pockets.
In his book, The Wired Society, he proclaimed that by the year 2000, computers
and other devices linked by fibre optics, telephone lines and radio waves
would allow millions of people to exchange electronic mail, shop for
merchandise, trade stocks, work at home, take classes, pay taxes, plan
vacations and entertain themselves. He added that entrepreneurs who jumped in
early would reap vast fortunes.
Not bad.
So when Martin makes a forecast, people listen. In a new book, After The
Internet: Alien Intelligence, he insists that we are on the cusp of a
discontinuous leap in what computers can do and that the changes coming,
properly guided, will lead us all to a land of milk and honey.
``It's not just a question of computers becoming more powerful but rather of
their developing a different kind of intelligence,'' Martin says.
``We will have machines that are a billion times more intelligent than we are,
but only in narrow, specific ways. In the 1960s and 1970s, the
artificial-intelligence people kept telling us that in 20 years computers
would be as intelligent as people. We cannot now get close to programming what
a mosquito does, much less a human being.''
We must abandon the false promise of artificial intelligence - the general
term for technologies that aim to emulate human cognition - and understand,
embrace and exploit the alien nature of computer thinking, he says. There will
be close, synergistic partnerships with machines. Humans will undertake
creative tasks, leaving the drudgery of realizing them to evolved computers
that are simultaneously mysterious, powerful and oddly naive.
Martin likens them to human idiot savants, and as such, he believes we can
keep them under control - ``a person with general intelligence will always
find ways to control a person without such intelligence.''
Such predictions are backed up by credentials as imposing as Martin's
6-foot-5' frame. Now 67, he was the father of computer-aided systems
engineering (CASE), a fundamental breakthrough in the 1980s that automated
software development.
Multinational CEOs and other corporate Brahmins still pay him up to $35,000
(U.S.) a day to tell them what the future will bring and apparently find it
money well spent.
Martin has started a half-dozen wildly successful software and consulting
companies, including James Martin Associates, which brought Microsoft founder
Bill Gates to Bermuda in 1990 to try to buy it. (Martin declined, saying:
``The company was growing. I saw no point in selling.'')
Bold, dead-on prognostication has made the reserved, soft-spoken Martin a
millionaire many times over, but he contends that accumulating wealth was
never his sole aim. Rather, he says, what he enjoys is imparting his latest
vision. He wants his clients, and the rest of us, to realize that ``human
history has been a story of desire outstripping the means to achieve it, but
that may well be reversing.''
In an era when it is fashionable to fear the fast-changing, high-tech future,
Martin contends we should climb aboard and grasp the reins firmly.
``We need to think hard about the new power we human beings have and make sure
we use it to our benefit,'' he says. ``We need to keep talking about these
changes. We need to become masters of complexity, and we need to do it rather
quickly.''
What could a new wave of technology do? Imagine, Martin says, hardware and
software that vastly expand our ability to meet human desires. Imagine a
technological world that plumps pillows, smoothes paths, and cleans, greens
and beautifies in the precise ways that we want.
For example, take the present hassle of proving you are who you say you are:
cards to access automatic teller machines, credit cards, driver's licences and
so on.
Martin says your wallet will grow thin as programs like TrueFace recognize
your unique appearance. Doors will open, cars will start. Pay phones, vending
machines and parking metres will automatically bill you.
Medicine will prevent illness rather than provide remedies after the fact.
Intelligent software will comb health and genetic data, foreseeing not only
the likelihood of, say, a heart attack but also prescribing therapies to keep
it from happening.
As for the environment, Martin believes the whole planet could follow the path
of Agar's Island, his home in Bermuda. An abandoned British military munitions
storage site that he purchased in 1995, the island is becoming a tropical
fantasy.
``It is a great example of how the planet's bad places can become very nice
places to be if technology is put to use intelligently,'' he says.
The engine of this transformation will be what he terms alien intelligence -
computer smarts evolved so far beyond human thinking that they can do things
no conventionally programmed computer can, in ways no human being can
understand.
Martin explains that computers are beginning to use Darwinian-like
evolutionary programs to ``breed'' solutions to problems - and unlike
biological evolution, the computers can spawn thousands of new generations per
second.
By using genetic algorithms, programmers the world over are already
replicating evolution to tackle virtually every area of human enterprise, from
improving steel quality in European mills to maximizing production schedules
to more precisely aiming radiation therapy beams in British hospitals.
Even hardware can evolve. The field-programmable gate array chip, invented in
the 1980s but just now reaching its potential, gives a glimpse of how future
computer circuits will design themselves.
Unlike a typical silicon chip with etched immutable switches, or ``gates'' in
chip parlance, the field-programmable chip can be reconfigured instantly and
endlessly. Better yet, it can be bred rather than designed to do a task.
Closely linked to the rise in evolutionary computation, Martin predicts, will
be an explosion in artificial neural networks: computer hardware and software
that link powerful parallel processors so that they roughly but narrowly
emulate the learning capacity of the human brain.
Martin predicts that in as little as 10 years, what he calls breeding
factories - vast buildings as big as a General Motors assembly plant filled
with parallel-processing supercomputers - will dot the Earth.
Such factories will breed software and field-programmable chips capable of
astounding feats, then spread their products worldwide via the Internet.
One concern, voiced by Sun Microsystems' co-founder Bill Joy, is that if
computer intelligence continues to increase at a compounding rate, computers
will quickly render human beings useless, or even turn on them.
Martin disagrees. Computer intelligence, he emphasizes, has always been so
different from human thought processes that it's wrong-headed to see the two
as locked in a race toward similar goals. Still, he concedes, ``we are going
to make mistakes.''
A couple of 800-passenger jets colliding in mid-air due to faulty
alien-intelligence software is the sort of mischance that could make the
public ``go hysterical,'' Martin says.
``While we will make mistakes, what most of the pessimists forget is that
human beings are not passive. Once things begin to go wrong, humans have great
skill at thinking up ingenious ways to correct the problem.''
Martin smiles. ``I can think of no reason why we should stop doing that.''
©¿©¬
Stay hungry,
--J. R.
Useless hypotheses, etc.:
consciousness, phlogiston, philosophy, vitalism, mind, free will, qualia,
analog computing, cultural relativism, GAC, Cyc, Eliza, cryonics, individual
uniqueness, ego
This archive was generated by hypermail 2b30 : Fri Oct 12 2001 - 14:40:19 MDT