From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rutgers!uwm.edu!cs.utexas.edu!uunet!tdatirv!sarima Mon Dec 16 11:01:36 EST 1991
Article 2091 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rutgers!uwm.edu!cs.utexas.edu!uunet!tdatirv!sarima
>From: sarima@tdatirv.UUCP (Stanley Friesen)
Newsgroups: comp.ai.philosophy
Subject: Re: Searle (was.....)
Message-ID: <316@tdatirv.UUCP>
Date: 12 Dec 91 19:42:36 GMT
References: <YdEwb8S00iUz02j910@andrew.cmu.edu>
Reply-To: sarima@tdatirv.UUCP (Stanley Friesen)
Organization: Teradata Corp., Irvine
Lines: 75

In article <YdEwb8S00iUz02j910@andrew.cmu.edu> fb0m+@andrew.cmu.edu (Franklin Boyle) writes:
|
|What I mean is that when you develop new knowledge -- generalizations,
|analogies, etc. -- it must be physically causal in order to affect the
|functioning of the system. In a pattern matching system, in order for
|any data (knowledge) to be causal, it must have a matcher which
|physically detects it in order for it to be able to trigger the
|appropriate function.

Or the pattern matcher must be attached to an input transducer and an output
transducer that convert sense data into the format it needs and its output
into operational responses (which then effect the input sense data, cpmpleting
the feedback lopp and allowing self-correction).

And I maintain that it is exactly at this level that the human brain is
preprogrammed, so that equivalent preprogramming in a computer is not at
all inconsitant with cognition.

The problem is you keep coming back to the phrase 'physically causal',
which as I use it is perfectly capable of refering to a programmed robot
with sensors and effectors.

In what way is a human brain 'physically causal' that a robot is not?

That is what I do not understand.  What is it in your definition of
'physical causality' that eliminates an operational robot from consideration?

|How are these matchers created so that they can effect the appropriate
|responses except through feedback or deliberate programming?

They can't be, the human brain involves both feedback *and* programming.
In the case of humans the programming was done by natural selection.
In the case of computers it will be done by humans.

I see no difference between these except the irrelevent detail of the
source of the programming.

| My problem
|with higher level cognitive functions is that one has to be able to account
|for the physical creation of these matchers, and feedback from the
|environment does not seem plausible in that case.

Quite, and it is not plausable in humans either.  Our genetic program
includes a design for a complex set of matchers that are grown organically,
but operate cybernetically when finished.
|
|An alternative to pattern matching (for which the pattern structures are
|*matched* and, therefore, only causal to the extent that the match *triggers*
|a response) is structure *transmission*.  In pattern matching, no structure is
|transmitted beyond the matcher because the matcher physically "fits" it.

Fine, this seems to me to be correct, and applicable to the human brain.
The pattern matchers in the human CNS are only causal to the extent that
they are connected to effectors (for example muscles).

Transmission is insufficient to explain human cognition, since the cognitive
operations of a human are far removed, structurally and physically, from
the original stimuli.  Even our conscious awareness of sensory inputs is
at a *very* high level of abstraction.  We are aware only of objects and
relationships, not of individual sense signals (try to locate the pixels
in your visual field - you can't do it, they have been abstracted out by
the time the visual input reaches the cognitive level we call consciousness;
but they *are* there at the physical input level, indeed even at the level of
input to the cerebral cortex).

|So, in a pattern matching system, when patterns are input, there better
|be a matcher there to handle it (the matcher could be "more general" than the 
|input pattern).  Otherwise, it will do nothing more than sit in memory.

This seems to be about how humans respond to uncategorized inputs.  Often
we simply do not notice them.  Why is this a problem?  This is just how
such things work.  Our own system is neither complete nor perfect.
-- 
---------------
uunet!tdatirv!sarima				(Stanley Friesen)


