> From: Cowley, CT <ctc196@soton.ac.uk>
>
> After the last lecture, (12,2,97) I was wondering exactly how analog
> and computation models actually help to explain the mind. It seems that
> they are merely a way to describe what the brain can do, and to some
> extent how it works. However, all of these things can be modeled on a
> computer, and I would not believe that a computer, at least one capable
> of symbol and image manipulation, has a mind. It does help to explain
> how the brain works, but not how a self aware creature (i.e. us) actually
> understands, and so I can't see how it helps our understanding of the
> mind itself. Am I missing something very obvious?
It is true that a computer can SIMULATE just about anything, including
furnaces, airplanes and tidal waves, but that doesn't mean that a
computer can heat, fly or flood your basement.
When we use a computer to model something computationally, we are not
claiming that the thing we are modeling is a computer, computing.
The thing we are modeling may be a furnace, heating; or an airplane,
flying. When we simulate something computationally we do it to
understand how that thing works. For example, if, before any plane had
been built, we had had computers, and if we succeeded in simulating
a plane in the air with a virtual plane in the virtual air, we might
have saved a lot of expensive trial and error: We might have been
able to avoid false starts, with plane designs that could never fly.
We would test them out by simulation, and if they failed, we would try
another design, until we found one that worked.
Let's call the theory that the same sort of thing is what we're doing
when we do cognitive modeling "weak computationalism" (WC). WC does not
claim that the brain is computing; it simply uses the computer to test
out theories of how the brain does what it does. WC would be how most
sciences, including astrophysics and mechanical engineering would
use the computer.
Strong computationalism (SC) would be the theory that it is not just
that we can SIMULATE the brain and the world in order to understand
how the mind works: According to SC the brain is literally doing
computations, just as the computer does. According to SC cognition
IS computation. SC is much more controversial. See the discussion
of Searle's Chinese Room Argument in Chapter 1.
Computational, analog and neural net models are all attempts to explain
how the brain can do what all the things it can do. When you have a
model that can do it too, a model you can understand, because you
designed it, then you have at least one possible explanation of how
the brain can do what it does. Without the model, all you have is utter
mystery. With the model, you have the beginning of an explanation.
Now analog models and neural networks can all be SIMULATED on a
computer, just as a plane and a furnace can. But if they work, it
doesn't mean the brain is a computer, any more than the plane is a
computer. It simply means that using analog representations and
neural nets, we can do things that the brain can do. If the model
is correct -- if it "scales up" to being able to do EVERYTHING we can
do, then it is a pretty safe bet that that's how the way the brain does
it. (Not through computation, but through analog processes and
neural nets rather than just their simulations on a computer.)
There may be parts of cognition, though, that are based on the brain's
literally doing computations (symbol manipulation). Although pattern
recognition and categorisation is better explained by analog processes
and neural nets, reasoning, language and problem-solving are probably
better explained as involving real computation (symbol manipulation)
in the brain.
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:49 GMT