[UNKNOWN] Re: [UNKNOWN] Re: PCT and Values

CSGnet Archive CSG2004
1

[From David Goldstein (2004.11.11.1944 EDT)]
[Bill Powers (2004.11.10.0701 MST)]

Bill,

I don't think I completely understand your answer.
A perceptual signal is not a perception but it means a certain perception
within a given control system. I understand this.
I also understand that a reference signal is not a perception.
But why wouldn't it mean a particular perception within the same control
system?

How are reference signals from a higher level control system assigned to the
comparators of the next lower level systems?

David

ยทยทยท

----- Original Message -----
From: "Bill Powers" <powers_w@FRONTIER.NET>
To: <CSGNET@listserv.uiuc.edu>
Sent: Wednesday, November 10, 2004 9:56 AM
Subject: [UNKNOWN] Re: [UNKNOWN] Re: PCT and Values

[From Bill Powers (2004.11.10.0701 MST)]

David Goldstein (2004.11.10.1755)--

[ Rick Marken (2004.11.09.0855)]

A reference state is a particular state of a perception which can vary.

So a reference state is a perception.

As a perception, it is built from lower level perceptions, I believe.

I can't tell who said this, but it needs to be corrected. A reference
signal is NOT a perception. For a comparison to take place, you need two
signals: the perceptual signal, which is the current value of a
perception,
and a reference signal, which is a DIFFERENT signal set to the value that
is desired.

In PCT as of 2004, all signals are one-dimensional, meaning that they can
vary only in magnitude (frequency). So a perceptual signal has a magnitude
that can vary, and a reference signal has a magnitude that can vary but is
set to one particular magnitude at a given time. The action of a control
system causes the perceptual signal to vary until its magnitude is the
same
as the magnitude of the reference signal. A reference signal does not come
from a pattern-recognizing function; it is simply a signal set to have a
specific value.

A perceptual signal is not patterned in PCT. It indicates by its magnitude
the degree to which a pattern is present in the set of inputs (lower-order
perceptual signals or stimuli), but it does not itself carry that pattern.
The pattern is recognized by computations in the neural network, the
perceptual input function, that generates the perceptual signal. The
pattern has no existence outside the neural network, or the computations
taking place in it.

I am simply describing here the basic principle of analog computing,
perhaps in a form that is strange. In an analog computer, what makes one
voltage mean one thing and another voltage mean something entirely
different is the nature of the physical process that receives one voltage
and generates the other. The physical process IS the "computation."
Integration of a voltage, for example, is done by

1. applying the voltage to a resistor to produce a current proportional to
the voltage;

2. Passing the current into a capacitor which charges up at a rate
proportional to the current, and

3. Detecting the charge on the capacitor as a voltage that is the output
of
the computation.

If the input voltage represents a force, the output voltage represents a
velocity. The input voltage, however, does not apply a force to anything,
nor does the output voltage have a velocity. They are both just voltages
with specific values at any given instant. If you did not know the
computation that derives the second voltage from the first one, you could
not tell what the second voltage means. If the first voltage could not be
identified as the coming from a force transducer, or a knob labeled
"force", you could not tell what it meant, either. Analog computation is
done without symbols, and the only rules of computation are those built
into the physical structure of the computing elements. Although patterns
can be recognized and acted upon, there are no patterns in any of the
signals.

This is the whole basis of the PCT model. It is the basis of all the
computer demos developed in the last 30 years in PCT, and also in most
other fields that involve simulations.

Does it explain experience satisfactorily? No, I don't think so, not in
the
sense of explaining why the world appears to us as it does. It explains
controlling very, very well. It predicts behavior very well. It even
explains well how it feels to control things. But you'll notice how the
"very"s drop off.It does not explain at all why red looks like red or why
a
square looks square. Conscious experience is, so far, unaccounted for.

So a reference signal is not a perception. But then a perceptual signal is
not a perception, either. They are both just signals that can vary in
magnitude. What makes one a perceptual signal is that it comes out of a
perceptual input function. What makes the other a reference signal is that
it DOES NOT come out of a perceptual input function, and that it goes to a
comparator where it specifies a magnitude to be matched. The action of the
control system makes the reference signal into a reference signal, because
only the control system as a whole can make the perceptual signal's
magnitude match the magnitude specified by the reference signal, thus
demonstrating the primary effect of the reference signal.

Best,

Bill P.

2

[From Bill Powers (2004.11.11.0645 MST)]

David Goldstein (2004.11.11.1944 EDT) --

I don't think I completely understand your answer.A perceptual signal is
not a perception but it means a certain perception within a given control
system.

I think it IS a perception -- that is, I think that what we experience as
perceptions are neural signals rising level by level in the brain, going
from peripheral locations toward central locations. Those are perceptual
signals.

Or perhaps what we experience are the chemical concentrations inside
neurons that are affected by neural signals travelling from one neuron to
another in the afferent pathways. What am I sure about? That what we
experience is not the external or internal physical world itself that
perceptions are supposedly about.

I also understand that a reference signal is not a perception.
But why wouldn't it mean a particular perception within the same control
system?

Think of the thermostat control on the wall of a room. There is one
indicator that shows the current room temperature -- say it's 68 degrees.
That indicator is like a perceptual signal. Then there is a second
indicator, in the form of markings on a dial or something similar. You can
predetermine the room temperature by changing the setting of this
indicator, say to 72 degrees. That is like a reference signal. The second
indicator does not show the actual room temperature; it shows what
temperature you want. The temperature you want is 72 degrees, but that is
not a perception of the room temperature, which is 68 degrees.

The only similarity between perceptual signals and reference signals in a
given control system is that they are associated with the same variable --
air temperature in the case of the thermostat. That "meaning" is
established by the fact that it is a change in air temperature that causes
the reading on the thermostat, the perceptual signal, to change. That in
turn gives meaning to the reference signal. Because the perceptual signal
means air temperature, we say that the reference signal is a temperature
reference, even though the very same signal would have a different meaning
if it were connected to the comparator of a different control system.

The nature of the sensor, the input function, is what gives meaning to the
perceptual signal, and also to the reference signal. Otherwise they are
just signals; you couldn't tell by measuring them what variable they
represented, or even (without experimenting) which one was perception and
which one was reference.

How are reference signals from a higher level control system assigned to the
comparators of the next lower level systems?

I think that is in large part a matter of learning and reorganization. A
higher system becomes organized when something (the "reorganizing
system"?)finds the connections to the next lower level of comparators that
will allow controlling the higher system's perceptual signal. We know that
synapses are always forming and disconnecting -- slowly -- even in the
adult brain. Very probably, at the higher levels there is also active
switching by the output functions, so the same higher system can change
which lower systems it is using for control. Plenty of room there for
conjecture and experimenting with models.

Best,

Bill P.