X-Message-Number: 9333
From: Thomas Donaldson <>
Subject: Re: CryoNet #9324 - #9331
Date: Sat, 21 Mar 1998 16:50:09 -0800 (PST)

To Mr. Metzger:

As I said in my most recent posting, do not pride yourself upon your knowledge
of computing, since you clearly do not know what recursion is as formally
defined. This issue makes me suspect your own expertise. I would hardly mind
that (I've already said that I never took any formal course in computer
science) were it not that you seem to want to beat me over the head with
your supposed expertise. 

Furthermore, let us suppose that we accept the worth of Turing machines as
statements of the ultimate possible. You state yourself that use of such
ideas, just as does use of Carnot machines, allows us to show that some
kind of computing are impossible. That's fine, but we were discussing not
whether simulation of a human being by computer was impossible but whether
it was POSSIBLE. The two are not the same, as you should know. That does not
even require mathematics, it requires only elementary logic.

Sure, if you wished to use Turing machines to show that simulation of 
a human being was somehow NOT possible, and your arguments were correct, then
you would have a good argument. Clearly that isn't what you aim to do, so
please cease bringing in irrelevant arguments about Turing machines.

If we want to discuss what is possible, we must use some stronger notions.
Since infinite tapes remain clearly impossible, we will have to exclude
them. This alone affects our ideas of just what is and what is not possible.
Among the ways it does so is that any possible computer must operate at
a finite accuracy. Nor can you solve this problem of finite accuracy just
by increasing it one more time: it remains inevitably finite. For that 
matter, we do not know (and cannot prove in general) that a given calculation
will terminate at all. Calculations which will never terminate are rife;
of course in practice we handle them by accepting some finite accuracy. 

No matter where you got your education, I would assume that you did just
a few courses in mathematics. No where in your postings have you come
anywhere near to answering my statements about chaos. Of course for an
ideal Turing machine, no chaos in our calculations will arise; but that
says nothing at all, since such machines cannot exist. Finite accuracy has
consequences. It is not at all sufficient just to make the observation
that neurons are electrochemical machines. You must also show that when
connected as in brains, they will collectively behave in such a way as
to DECREASE the inaccuracy of your simulation, or at worst leave it the
same. When I bring in chaos here I am not talking about small inaccuracies;
I am talking about a situation in which your simulation will show an
EXPONENTIALLY INCREASING level of inaccuracy the longer you run it.

What would that mean for your simulation? Well, at first it would cease to
act as I would have acted or think as I would have thought. At that point
you might still say you were at least simulating a human being. But then
it would get even worse, as it began to act as no human being at all would
act. And finally it would behave so wildly that it simply collapsed: the
computer on which you ran it would fail due to lack of memory or processing
ability. 

At this stage, of course, when we discuss brains I cannot honestly claim
either that they will or will not have the special features required for
calculations of their state not to blow up as I describe above. Nor do you
have the ability to claim that they will NOT blow up. The most I can say
is that many other situations which seem far simpler WILL show chaos and
WILL blow up; indeed, MOST situations have that feature. It is only because
when we do a course in differential equations (say) we look at the simplest
situations that we might feel that blowups are rare. For what it is worth,
my own understanding of how brains work suggests that you will get chaos.
It is far from a sufficient argument against that to observe that neurons
operate electrically with discrete pulses. (Chaos can even occur in 
digital systems). 

Why is it so important to you that we will someday be able to simulate
you in a computer? You seem to fall into vituperation every time someone
questions that idea. You have done so from the beginning of this discussion.
I would think that if you were truly sure of yourself here that you would
be far more patient. I do not claim all knowledge, nor can you. We might
learn from one another. But that is not how this discussion has gone.

Turing machines are good ideas, though they fail to answer many other
questions we might have about computing. When we talk about simulating a
human being, we are talking about one of those questions. Simulation is
very useful too, but when we do it we need also to know its limits, or
we will fall into error. And it's certainly true that we know more about
how brains work now than even (say) 10 years or 5 years ago. The questions
we are discussing will someday be decisively answered; I have given my
reasons for believing that such answers will show, not so much the
impossibility of a simulation, as the vast amount of computer power
required to do so for even a few minutes.

Finally, some side issues. You seem to attribute various beliefs and 
attitudes to me that I do not have. Among others, diamond may be stronger
than bone, but it is also more brittle and requires more energy to 
produce. Perhaps we may devise better materials than bone in the future,
but diamond seems too simplistic. The energy required to produce some
material plays an important role: we can't just look at the end result.
That is a good reason why we now use bone. Wheels, too, work best on
flat surfaces, whether man-made or not. Legs have lots of advantages
when surfaces are NOT flat. And that is why we have legs. This does
not mean that we cannot improve ourselves, if improvement means that
we can change ourselves to live better in our PRESENT (or future) world.
But we should not so easily decide something is superior until we
look at ALL its features rather than just a few. The Daleks would be
hopeless when faced with randomly jagged surfaces, while human beings
would traverse them easily.

			Best wishes and long long life,

				Thomas Donaldson

  
a   

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=9333