X-Message-Number: 11480
From: 
Date: Sun, 28 Mar 1999 22:33:14 EST
Subject: I married a computer

In  a  recent issue of the NEW YORK REVIEW OF BOOKS there is a review by
philosopher John Searle of computer innovator Ray Kurzweil's book, THE AGE OF
SPIRITUAL MACHINES, subtitled WHEN COMPUTERS EXCEED HUMAN INTELLIGENCE
(Viking, 1999). 

Searle's review is captioned "I Married a Computer." As it turns out, some
people literally have married computers--because at one time the word
"computer" meant a person who does computations. I might add that many people
have married typewriters--because at one time the word "typewriter" meant what
we now call "typist."

The main thrust of Kurzweil's book is that computer speed and storage will
increase enormously (no argument there), that we will learn exactly how brains
think and feel (few will argue this either), and therefore computers will be
able to do everything human brains do, and do it faster and better. One
asserted consequence is that we will have the option of "uploading" ouselves
"into" computers.

Searle's review is mostly a reprise of his rejection of the thesis that
present-style computers are or can be "intelligent," let alone conscious. 

So who is right? Neither. Both discussions are seriously flawed. 

Kurzweil does in a couple of spots acknowledge that we may not know for sure
whether a computer is conscious, but he thinks we will believe it when it says
it is. He also says that knowledge of a computer's internal workings will not
tell us whether it is conscious. 

This is wrong, at least one side of it. When we learn the anatomy and
physiology of human feeling, then the presence of something similar in another
entity will be a good indication of consciousness in that entity. After all,
that is why we assume consciousness in other people--they are similar to us,
and we know we are conscious. Lack of similarity, on the other hand, might or
might not disprove consciousness, depending on what we learn about the
requirements of consciousness. Maybe one day some kind of radio-telepathy will
come into play, so individuals can share subjectivity.

Searle does not deny that machines can think and feel. We are machines (I
prefer to say "mechanisms") and we think and feel. But he denies that present-
style computers, regardless of speed or storage capacity, can "think" at all
or "know" anything, because they do nothing but manipulate symbols which have
no internal semantic content. Isomorphism is not everything; the map is not
the territory; a computer simulation of the digestive process will not
actually digest food.   

 Of course, the strong AI people and uploaders will retort that a simulated
digestive process will indeed perform digestion--of simulated food for
simulated people. But they are simply reiterating their mantra that
isomorphism indeed IS everything, that nothing matters except development of
patterns of information, the "information paradigm"--an unproven article of
faith, moderately plausible but no more than that.

Searle's blind spot seems to be his failure to notice that many of the signals
in our brains are "merely" symbols of certain events or conditions, either
external or in other parts of the brain. Yet we use the feedbacks, along with
trial-and-error, to build a picture of the external world; we build a mental
world with semantic content, using mere symbols (in conjunction with our self-
circuits or subjective circuits, which are not yet identified). Of course, we
don't start from scratch as babies; we have a head start through the gifts of
previous evolution, with many propensities hard-wired in right off the bat.
With enough understanding, the same could be done for computers.

As for Kurzweil's sins, they appear to be mainly those of sloppiness and
hurry. Doubtless his book was just one of many projects on his agenda, and
perhaps he didn't do much review or revision. Searle catches him in various
errors and failures as an expository writer. One gross error which I did not
notice Searle mention was Kurzweil's repeated statement--conveyed as fact, not
minority opinion--that quantum decoherence requires a CONSCIOUS observer, not
just an interaction. (For example, Schroedinger's cat remains in a
superposition of dead/alive states until some outside conscious observer takes
a look; and in fact the whole universe was in limbo until conscious observers
caused its whole history retroactively to gel.)

As far as I recall, neither Kurzweil in this book nor Searle in various
writings have dealt with David Deutsch's put-down of Turing computers
(including all present digital computers)--that they are classical, while the
real world is quantum. Can a classical system simulate events in a quantum
world? To some extent, obviously yes, since scientists do use computers, and
even hand calculations, to derive quantum results, and the rules of quantum
mechanics now in use are written in classical language. But one suspects that
any attempt actually to emulate a person, or a significant part of the real
world, with a Turing machine would sooner or later run into insuperable
problems, if indeed there were not insuperable conceptual problems at the
outset, as I have discussed elsewhere.

And yet again, does all this, or any of it, bear on the real problems of
ordinary people? The answer, however unwelcome, is a flat yes. Ordinary people
are not compelled to try to understand such esoteric matters--but the
alternative is to accept the dicta of authority, which leaves you with the
almost equally difficult problem of choosing the most authoritative authority,
and then deciding which dicta are the important ones. There are no cheap and
easy answers. If you want to understand the choices in life and death, you
have to pay attention and do your best to make intelligent bets. Nature has
neither malice nor mercy. Freeze, please.

Robert Ettinger
Cryonics Institute
Immortalist Society
http://www.cryonics.org 

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11480