NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/bit/listserv/csgl/1541

Text File | 1992-11-18 | 6.0 KB | 114 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!elroy.jpl.nasa.gov!usc!zaphod.mps.ohio-state.edu!uwm.edu!psuvax1!psuvm!auvm!VAXF.COLORADO.EDU!POWERS_W
X-Envelope-To: [e-mail]
X-Vms-To: @CSG
Mime-Version: 1.0
Content-Transfer-Encoding: 7BIT
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Wed, 18 Nov 1992 07:06:29 -0700
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         "William T. Powers" <[e-mail]>
Subject:      Complexes of perceptions
Lines: 99

[From Bill Powers (921117.2000)]

Note to Len Lansky: just set the margins in your word processer wider --
say, to 79.

Martin Taylor, Rick Marken --

The reason I use words like intensity...system concept is to avoid
getting embroiled in loose language, including terms like "standard,"
"value", "belief," and so on.

Martin, you say you would like to use "belief" to mean a structured
perception. I think we agree that what you mean by a structured
perception is a collection of scalar perceptual signals. But a
collection of signals, just by coexisting, does not create any special
kind of perception. I feel my right big toe against the floor and I feel
my left forefinger hit the "f" key. That does not create a structured
perception.

What creates something structured out of a collection of scalar
perceptual signals is a _higher level perceptual function_. If I make
the toe-sensation at the same time as the finger-sensation I perceive
simultaneity, a perception of relationship. The effect of this higher
level perceptual function is to represent the presence of a particular
kind of structuring AS A SIGNAL, in this case a signal standing for the
presence of a relationship called simultaneity. But the presence of
simultaneity means nothing unless it is perceived and explicitly
represented.

The basic tenet of HPCT is that nothing is perceived unless it is
represented as a perceptual signal. This includes what you call
"structuring" of a set of signals. If a structuring is perceived, that
is only because there is a perceptual function organized to represent a
collection of inputs in terms of that kind of structuring.

This is just one aspect of a much deeper point I have been trying to
make since BCP, where I brought it up as Rosenblatt's Principle. I have
spoken about it on the net as the difference between implicit and
explicit functions. In any collection of n perceptions, there is an
infinity of implicit functions of those perceptions. That is, you could
construct an arbitrary function of, for example, the form ax1 + bx2 +
... or any other function, assign arbitrary values to the coefficients,
and continuously compute that function of the perceptions xn. Actually
doing that by using a computing device that outputs a signal
continuously representing the value of that function makes the function
explicit. To give that function of the variables any effect in a model,
the model must contain the machinery for computing it. It is not enough
for a bystander to know that such a function COULD be computed; that has
no effect in the model. The function must actually be computed if it is
to have any meaning in the model, or any effects on anything else in the
model.

In some neural-network models I have seen, perceptual processes are
treated as the existence of organized patterns throughout some volume of
the brain. The problem is that those patterns are being recognized bythe wrong
 brain, the brain of the observer. All active networks carry
signals that are exhibiting some pattern; which pattern depends on what
regularities the observer is prepared to perceive. There might be some
patterns that the observer finds especially easy to see -- synchronism
and so forth. The catch is that these patterns simply represent the
internal working of some part of the brain, but do not necessarily have
any significance to any other part of the brain. For a pattern to have
significance elsewhere in the brain, something must RECOGNIZE the
pattern and indicate its presence or state by a signal sent to other
parts of the brain. In short, there must be a perceptual function that
converts one of the infinity of possible implicit patterns into an
explicit signal. There is no such thing as "the" pattern of signals.

For a brain modeler to speak of any pattern of signals in a network,
that modeler must be using a human perceptual function capable of
indicating presence of that pattern. To the human observer, there can be
just a hint of the pattern, a pretty clear sense of the pattern, or a
perfect prototype pattern. There is a scale of perception ranging from
hardly any to the maximum possible. So in the human observer, the
pattern is represented as a scalar perception.

So why does the human modeler not put that same sort of function into
the model? I claim it is because most human modelers see no connection
between the system they are modeling and themselves. They reify their
own perceptual interpretations. If they see a certain pattern in the
behavior of a model, they assume that the pattern actually exists in the
model, and that by its very existence it can have specific effects. They
do not realize that their own pattern-recognition ( or better, -
construction) abilities are creating a significance in their own brains
in the form of a perceptual signal. Yet they believe that somehow that
collection of signals in the model, without any function to recognize
it, can have significance in the model just by being in the model.

This is a very difficult point to communicate. I hope it's getting
across, at least a little. A "complex of perceptions" appears as
something unitary only to a perceptual function capable of perceiving
that kind of thing and signalling that it is present. Two different
perceptual functions receiving the same set of lower-order perceptions
can easily perceive different "complexes" in it. The complex is in the
perceptual function of the beholder.
---------------------------------------------------------------
Best,

Bill P.