X-Message-Number: 7936
Date: Mon, 24 Mar 1997 09:04:47 -0800 (PST)
From: Joseph Strout <>
Subject: on symbols and computing

I think Thomas makes a strange distinction between conventional computers,
like the ones on our desks, and other computers such as brains, neural
nets, and perhaps other analog computers.  Thomas, it seems that you use
the word "computers" to mean only serial instruction-processors, which get
their data via a keyboard and display results via a monitor.  That is our
conventional type of computer, I agree, but I see it as only a convention.
There are many other types of computers besides this, and they all
compute.

By "compute" I mean that they perform complex operations on and
transformations of information.  For example, we can look at an image
(say, of an apple) and convert this information to a word ("apple").  A
conventional computer can take a number and convert it to its square root.
Each computer can do many other things as well, of course.

Note that when one type of computer tries to imitate a very different type
of computer, it does so slowly and inaccurately; consider your desktop
computer trying to recognize an apple, or you trying to find the square
root of 59371.  I think this is why Thomas says that we are not
symbol-processors, and that our brains are fundamentally different from
computers.  I'd agree that we're very different from *conventional*
computers, but we are still computers.  It's a matter of terminology, and
not very important, except that it can be misleading either way.  If you
say only that brains are computers, you may be led into a false sense of
simplicity, and think that we just need a big supercomputer to reproduce a
brain.  If you say brains are NOT computers, some will be led into a false
sense of difficulty, and think that no brain emulator could ever be built.
Most accurately, we should say that a brain is a very different type of
computer than today's conventional computers.  (It's also very different
from conventional neural networks.  It's perhaps most like a hybrid, both
sequential & parallel.)

As for symbols having no meaning to a computer -- well, that is mainly
because they are disconnected from the world.  Our brains receive nothing
but action potentials; what meaning do THOSE have?  None, of course,
except that they are correlated in interesting ways.  Indeed, since they
arise ultimately from our senses, which ARE connected to the real world,
these action potentials have statistical properties which let our brains
compute a self-consistent model of things being "out there".  A
conventional computer would have the same opportunity if it were in a
robot, connected to a rich set of sensorimotor capabilities.

Finally, as to Searle's Chinese room: it seems odd to imagine that the
system of rule cards, Chinese dictionaries, and a speedy little man who
doesn't speak Chinese could be aware.  But what if you shrunk the whole
system down, and replaced the single speedy man with a team of men -- or
even 10^12 men, none of whom speak Chinese, and who are all just following
rules and passing cards around, and stuck the whole lot into a skull? 
Why, then you'd have a brain, and it's not at all odd to imagine that it's
aware.  Just food for thought...

Best wishes,
-- Joe

,------------------------------------------------------------------.
|    Joseph J. Strout           Department of Neuroscience, UCSD   |
|               http://www-acs.ucsd.edu/~jstrout/  |
`------------------------------------------------------------------'

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=7936