X-Message-Number: 9330
From: Thomas Donaldson <>
Subject: Re: CryoNet #9302 - #9310
Date: Sat, 21 Mar 1998 00:43:10 -0800 (PST)

For Perry Metzger et al

To Perry Metzger et al:

You seem drunk on Turing machines, despite the plain fact that they have
proven inadequate models even of significant computation. Not only that, but
you also show a remarkable ignorance of the mathematics involved in any
kind of simulation. The Siegelmann counterexample, alone, suggests to me
that some caution should be used when thinking about just what the range of
the Turing machine model might be. (And for Mr. Metzger in particular, it
does not help that we can add more tape when needed. Similar things can be
done with theoretical devices like that of Siegelmann, and it is the height
of illogic to allow such additions in one case and not in the other). 
As for Turing machines, I refer here to those which might be physically built:
if we allow infinite tapes, sure, you may be able to do lots of computations
--- though the time such computations take remains very important to us.
(Again for Mr. Metzger: don't pride yourself too much on your knowledge of
computing. My own books have been packed away, but I looked it up elsewhere:
the formal definition of recursion does not require a stack).

First of all, until we fully understand how brains work, it is not sufficient
to simply say that neurons are electrochemical machines. Depending on what
you mean, I can agree with that, but as a statement it carries far too little
information for any useful simulation. You need a much more precise notion
of just how those electrochemical machines work before you can even start.
Tell me about the electrical reactions IN DETAIL, and the chemical reactions
IN EQUAL DETAIL and also the interactions between them IN DETAIL. This has
not yet been done by anyone. WE know far less about simulating brains than
we know about simulating airplane wings or the weather. Not only that, but
unless brains have very special features (say, for instance, that they are
literally digital, but other features would work as well) then any simulation
of them will inevitably involve chaos, divergence, and failure after only a
short time. Merely being electrochemical machines is far from enough.

Given the complexity of human brains, it seems very unlikely to me that 
we won't find chaos in our simulation. Even laws as simple as Newtonian
gravity cause such divergences when you try to work with more than 2 bodies.
It is a wild assumption that human brains, or even for that matter
the brain of any warm-blooded animal, will not show at least equal chaos if
simulation is attempted.

Second, reality and simulation should not be confused. It is not that
we cannot form theories and do calculations about what will happen:
make a simulated rainstorm and examine its effects, for instance. The
problem with reality is that it will always contain features we have
not included in our simulation. Even for simulation of a human brain,
what that brain perceives and how it reacts will be strongly affected
not just by its internal state but by the world in which it lives. And
that world has a disobedient streak: it is always coming up with something
not in our models, or surprises us in other ways. To seriously
believe otherwise is basically to believe that we have now completely
understood the Universe. Sure, supposing that you can create a simulated
human being in a simulated reality (that is, supposing against the 
likelihood of chaotic behavior in your equations) then that simulated
person will still fall short of reality when the unexpected happens ..
in reality.

Of course, if people really come to believe their computers, then no
doubt we will see another age devoid of progress or exploration. Good
show! So far we have been lucky that such times have eventually fallen
apart: Reality intruded, perhaps brutally, and people stumble about in
dismay until they come to understand what has happened. Enough of that
and no doubt most of the human race (and its constructions) might be
wiped out. 

Even forgetting the probability of chaos, which in practical terms destroys
any possibility of simulating a person, the disobedience of reality can still
happen. It can occur both outside us, or in simulation of a person. If
it happens outside us, then we have to adjust our ideas. If it happens
in simulating a person ie. the actual person and the simulated one 
diverge badly, then we have just learned something important about how we
work. And remember that to be accurate a simulation must not only work 
when the person sits quietly at home reading a book, but when they go out
and talk with others or deal in any way with the world. (Yes, we are 
finite, but far more can affect us that is outside us than is inside us).

A simulation must even deal with events when the person becomes damaged;
our simulation of a brain must be able to deal with such things as 
brain injuries and their consequences, and all the different diseases
to which brains are subject. You might reasonably ask why diseases are
important (don't we want to do away with them?). They are important because
they occur due to features of our brain which when NOT stressed by disease
turn out to be quite useful. Glutamate, the chemical which plays a major
role in learning, can also play a major role in brain damage; mania
and depression both happen because of extreme chemical events which 
normally produce creativity. With great effort you might produce a being
who does not suffer from these faults, but at that point you are no longer
making a simulation. For that matter, such a being may not be at all easy
to produce: we need rain even though we do not like it when we are rained on.
Some things inescapably have two sides. But much more than that, internal
events may have consequences which may surprise us every bit as much as 
external events in reality can surprise us.

Basically, to claim that we will ever simulate a human being seems to me
to be next to the claim that we will someday completely understand the 
universe. No theory, which is a mere construction of words (and equations
which form a particularly developed language of their own), will even
fully predict the universe, any more than I can give you now the complete
decimal expansion of pi. Whenever we try to use language to describe
something (which is just what will happen, in a highly elaborated way,
if we try to simulate a person) that language will inevitably fail because
it is finite, while the states possible to human beings LIVING IN THE WORLD 
are not. (I will put that another way: a lens is finite, but can show 
an infinity of stars. And sure, you can describe something statically, but 
we wanted an active simulation, didn't we?).

			Best wishes and long long life to all,

				Thomas Donaldson   

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=9330