More consciousness and control; misc.

Subtitle: Confessions from a misspent youth and the method of levels

[From Bruce Gregory (960503.1140 EDT)]

(Bill Powers 960503.0530 MDT)

  This is close to the feeling that occurs at a successful end-point of
  the method of levels, except that the sense of "i" and the world of
  consciousness in general are not inaccessible -- they are just, for the
  moment, irrelevant. The only "sense of self" that remains is a sense of
  calm watchfulness. This is why I and others find the term "Observer"
  especially appropriate.

More years ago than I care to think about I did the est training.
Despite its California-pop psychology reputation, I found it
extremely powerful, and I am grateful for the opportunity it provided
me. The heart of the training was a lengthy process called "The
Anatomy of the Mind." The picture of the mind that was presented a
stimulus response structure. The point, however, was not to present a
model of the mind, but to have the participants encounter the
mechanical way in which they respond to life. The message was that
one is, and always will be a totally automatic machine, devoid of
spontaneity and meaning. ("Your life doesn't mean anything, _and_ it
doesn't _mean_ anything that it doesn't mean anything.") "Getting it"
consisted of realizing the truth of this observation. The power of
the process emerged because of the paradoxical effect of accepting
this conclusion. I noticed that it somehow stunned the perpetual
"voice-over" in my head into silence. (What's the point of paying
attention to your voice-over if it's just a mechanical subsystem
blathering on because that is what it was designed to do? Sort an
internal 3CPO.) I found, in Bill's words, the only "sense of self"
that remained was a sense of calm watchfulness.

The anatomy of the mind was a remarkably liberating process. Since
the outcome sounds so much like that of the method of levels, I can
only conjecture that they bear some, possibly deep, relationship to
each other.

Bruce G.

[From Bill Powers (960503.0530 MDT)]

Gary Cziko 960503.0218 GMT responding to:
John Burch (960502.1430)

     When I got to this point, I immediately thought of the times I have
     (too often) lost consciousness via trauma (I now wear a helmet
     while riding my bike). So I was intrigued to see you continue:

     [John:]
     >Has anybody ever been hit over the head so that when you came to,
     >you didn't know who you were, where you were or anything else for
     >a few moments?

This is close to the feeling that occurs at a successful end-point of
the method of levels, except that the sense of "i" and the world of
consciousness in general are not inaccessible -- they are just, for the
moment, irrelevant. The only "sense of self" that remains is a sense of
calm watchfulness. This is why I and others find the term "Observer"
especially appropriate. Of course it's a different experience to awake
and find that there is nothing to watch! I, too, have experienced this
after unconsciousness, due to a head blow and also while beginning to
recover from anaesthesia. Thanks for making the connection for me.

It's difficult to say whether that Observer was present during the
unconsciousness, because memory is a function of the hierarchy, too! The
return of memory and other functions is confusing, because in the normal
state the Observer is also present, but the activities of consciousness
fill experience, while the calmly watchful Observer does nothing very
"interesting." Even what you see of my activities right now is the
hierarchy working away, not the Observer observing. One of the useful
things your hierarchy can learn is that the Observer, not it, is in
charge. One can see how this sort of thing could be turned into some
pretty elaborate ghost stories -- not by the Observer, but by the
hierarchy.

Burch's proposal concerning the "50-millisecond delayed memory" really
only accounts for part of the phenomenon, if it really does account for
it. We still have to ask what it is that is evaluating the delayed
experience as well as the concurrent one. The proposal is problematic in
that it almost has to entail an _unnoticeable_ time lag -- it would not
do to propose a 1-second lag, because we would notice such a lag. When
you explain any phenomenon in terms of an effect that can't by
definition be observed, you are rather at a loss for a way to verify the
explanation.

Gary:

     I have found this "coming to" to be both a fascinating and
     terrifying experience.

The terror, too, is part of what "comes to," isn't it?

     At first just a feeling of existence, a certain sense of self, but
     not of Gary Cziko. Then gradually memories start coming back,
     first of the immediately preceding events, and then of who I am and
     what my life has been about. I have described this to others as a
     feeling of being reborn, like coming into existence for the first
     time and then going through 40-odd years of cognitive development
     in 15 seconds or so.

This "certain sense of self" -- it isn't a sense of "I", is it? That is,
as you say, it's not a sense of Gary Cziko, or his attributes, desires,
thoughts, intentions, (or intensities, sensations, configurations...
system concepts). I would describe it, in the initial moments, as
similar to watching someone else wake up. There is a moment when
identification takes place; when the Observer joins with the system and
suddenly "this is me."

David Goldstein told me of a psychiatric journal, the _Journal of
Dissociation_. It's not quite what you might think, knowing that
"dissociation" has been viewed in the past as an undesirable condition.
These psychiatrists have evidently stumbled across this phenomenon in
their patients (and apparently in themselves). In trying to figure out
what this state is, they have come to see it as involving a "Helper,"
some rather mysterious entity that seems to play a part in resolving
conflicts or even integrating multiple personalities (as in the case of
one of David's patients, who said that of course the Observer had been
there all along). So, fellows, maybe we're all a bunch of nut cases! But
_nice_ nut cases, you understand.

···

-----------------------------------------------------------------------
Peter Cariani (960502) --

     ... there are numerous examples where "firing rates" (itself a very
     vague term without any mention of which neurons are involved or
     what kinds of time windows are used for the "rate") do not (and in
     some cases, cannot) explain the perceptual distinctions that are
     made -- where other factors such as temporal discharge patterns and
     spike latencies appear to constitute the "signal".

Since rate is simply 1/interval, and changes in firing rate are
certainly "temporal discharge patterns" or changes in spike latencies,
it seems to me that there is no fundamental difference here. If a neural
signal looked upon as a rate of firing changes rapidly as a function of
time, there will of course be changes in the temporal discharge patterns
and spike latencies -- how could it be otherwise? A neural signal
carrying a low-order version of a complex tone of steady pitch, such as
the sound of an oboe, will consist of superimposed harmonics that,
neurally represented, will produce a complex pattern of firing
frequencies or interspike intervals (or neurotransmitter
concentrations), whichever way you want to look at it. But at a higher
level, the identification of the sound as that of an oboe can be
represented as a signal of constant frequency (or interspike interval),
as long as the sound is still that of an oboe. Of course _naming_ the
instrument is a matter for still higher systems.

The representation of pitch is a singularly complex problem because the
range of audio frequencies overlaps the range of neural firing rates
(which could also be described as an overlap of audio periods with the
range of interspike intervals). As far as I know, this problem is unique
to audition, occuring in no other sensory madality. I commend you for
investigating this complex phenomenon, and would be very interested in
seeing anything recent you have discovered about it.
---------------------------------
     Next to these problems, those of operationally-defining how to
     recognize self-producing sets of signals from some set of
     identified and observed signals is trivial. Consider a set of
     chemical reactions between substrates A-Z, where some combinations
     of substrates produce other substrates, e.g. A + B -> C, and M + O
     + Z -> D + W + F. One can make a directed graph of the reaction
     network and using graph-theoretic (or logical entailment)
     procedures find those sets of substrates that form closed loops,
     i.e. the members of the set through their interactions regenerate
     the members of the set.

Sounds useful. Graph theory, however, can represent only the map of the
interconnections -- the dynamics and other quantitative considerations
aren't contained in a map of logical entailment. In biochemical
reactions there are certainly closed loops, but some of them are control
systems and the rest are not. The difference depends on the loop gain.
If the loop gain is less than 1, I would not call the result a control
system even though there is a small amount of negative feedback. A true
biochemical control system probably has to include an enzyme, which can
insert a gain of as much as (or at least) 50,000 into the loop. That is,
a change in the effect of the enzyme on a reaction is reflected (via the
product) back to the enzyme (by, for example, shifting it from the
active to the inactive state) to produce an effect that is 50,000 times
as large as the initial change. But the system is stable. So not all
closed-loop systems have the same properties.

I believe it was Warren McCulloch who first misled cyberneticists
(Proceedings of the 5th Macy Conference) by telling them that a closed
loop (even a negative feedback loop) can't be stable if it has a loop
gain greater than 1. This is something to ponder when considering
biochemical control systems with closed-loop gains in the tens of
thousands.

See Hayashi and Sakamoto (1986), _Dynamic Analysis of Enzyme Systems_,
Springer-Verlag (particularly Chapter 6).
-------------------------------------
     How about them optical interferometry telescopes that were in
     yesterday's science section of the New York Times? Really
     impressive.

Yes, I had seen that in Science, Science News and Sky and Telescope.
Spots on Betelgeuse! Fantastic.

You're going to have to explain, however, why this observation showed up
in this particular post. Is it something about epistemology? I once
asked an astronomer friend why astronomers liked to transform the
almost-plane wavefronts of light that make up the universe into points
and extended objects, by using lenses or curved reflecting surfaces. He
didn't think that was funny.
-----------------------------------------------------------------------
Peter Cariani (960502.1200 EST) --

     However you want to characterize Hard and Soft accounts, I want to
     make it very clear that I have always been very much opposed to the
     notion that simply making something more complex confers onto it
     special properties.

Roger. I hope the same thing applies to the general notions of re-
entrance or recursion, patterns, and so forth.

     "Complexity" and organization are two very different things: one
     can have very simple control loops and very complex ones, but what
     makes a control loop a control loop, i.e. what gives it its
     functional organization is not determined by its degree of
     complexity. Isn't this obvious? Let's not conflate these things.

I agree, let's not. I can go on; it's not the re-entrance or recursion
that creates the function called control, nor is it the idea of
feedback, nor is it even the idea of "organization." All those ideas
apply to system which are not control systems as well as those that are,
and hence do not distinguish control from non-control.

     Secondly, I am not in any way invoking "magical" processes, andI am
     not proposing that "new physical phenomena," in the sense of new
     basic properties of matter, are being created by the evolution of
     functional organizations that are capable of observation. What does
     happen is that the evolution of a particular organization of matter
     permits there to be informational operations like "memory" and
     "modelling relations" (Pattee, Rosen, Kampis), such that
     "observers" are made possible.What does happen is that the
     evolution of a particular organization of matter permits there to
     be informational operations like "memory" and "modelling relations"
     (Pattee, Rosen, Kampis), such that "observers" are made possible. A
     rock is not an observer, a cell is a primitive observer, a human a
     much more elaborated one.

But that is exactly the step that I find magical. Why should memory make
observers possible? Is a phonograph conscious? Why should "modelling
relations" make consciousness possible? Does the relationship between a
computer model of tracking behavior and the actual behaviour make the
computer conscious? You speak as if we have all agreed on the nature of
the phenomenon that we call "consciousness," but what you are really
doing is saying that whatever the term may mean, it is "nothing but"
memory, modeling relations, and so on. That is not explaining
consciousness; it is reducing the problem to fit the kind of explanation
you prefer. That approach was anticipated some time ago by a fellow
named Procrustes.

Memory and modeling relations may give the observer something to
observe, but how can we say that they ARE the process of observation?

Once you have decided that consciousness is nothing but some set of
particular physical processes, your logic them compels you to suggest
that other examples of similar processes must also entail consciousness
-- even a camcorder has to have a tiny bit of consciousness. But the
basis for such claims is not any evidence of consciousness; it is only
the evidence that similar physical processes are occurring. The
relationship of these processes to consciousness remains, actually,
unknown and putative.

     One is not "required" to consider a system as a "control system" in
     order to describe its physical workings, and one is not "required"
     to consider particle-like properties of light in order to describe
     its wave-like properties (interference).

You have it backward, as least as far as PCT is concerned. We did not
start with "control" and from that derive the idea of a control system.
It went the other way. First, there were systems capable of stabilizing
variables outside themselves, maintaining those variables in specific
states despite disturbances tending to alter them. Our first organized
understanding of systems that could behave like this arose out of
engineering and the artifacts built by engineers. What was not generally
understood at first was that the engineers called them "control systems"
because of their (deliberate and intended) resemblance to human beings
carrying out processes commonly called "controlling."

One is not required to consider a system as a control system _unless it
is observed to be controlling something_ -- that is, varying its actions
on the environment in such a way as to stabilize some aspect of that
environment against arbitrary disturbances, and doing so in a way that
depends on sensing the environment. While the term control has been used
in many different ways, in PCT it is used in one and only one way, which
can be technically defined so it can't be confused with any other kind
of organized behavior.

The analogy with the wave-particle complementarity is spurious, because
in fact the behavior of matter can be seen as involving either waves or
particles, depending on how you set up the experiment and analyze it.
This is not true of controlling in the PCT meaning. If you set up an
experiment with control and determine that control is occurring, the
proper analogy is with setting up a pair of slits and observing that
there is interference, which demonstrates the wave nature of matter. But
there is no second way of setting up the behavioral experiment, that I
know of, which will produce a different result, analogous to showing the
particle nature of matter.

In the wave-particle situation, there are two different and mutually
exclusive results obtained: interference patterns, or lack of
interference patterns. This is not a case where there are two competing
explanations of the same phenomenon. There are, in fact, two different
phenomena. This is not just a difference in points of view; one cannot,
simply by changing interpretations, make interference fringes occur when
they are not occurring. It is necessary to change the physical situation
in order to change the phenomenon.

We have a different situation when we are explaining control phenomena.
Here we have a single phenomenon, the stabilization of variables against
disturbances (shorthand for a more complete description). Merely
changing one's perspective (or the word that is used) does not remove
this stabilization, nor (so far) is there any phenomenon that we call
control in which this stabilization does not occur. There are, of
course, other kinds of phenomena which do not entail control, but their
existence does not negate control, and is not an alternative to control.

     Does a system whose organization is described as a "control system"
     call into being new basic properties of matter? No.

No, not at the lowest level of organization. But there are, as you know,
emergent rules that show up at higher levels, and they can't be reduced
to the rules governing the lower levels. Join an inductor and a
capacitor, and you get rules that are not reducible to the rules
governing either inductors or capacitors. It is possible that
consciousness/awareness/Observing is such an emergent phenomenon. I
don't know. We can easily show HOW resonance emerges from combining
inductors and capacitors (conservation of energy is one way), so we can
prove that true emergence is occurring. But until we can show HOW
consciousness emerges from neural processes, we can't establish the
link.

     Might such an organization permit such a material system to become
     an "observer", in some very primitive sense? I think so.

Until we can establish HOW consciousness emerges from these lower-level
processes (if it does), the best we can do is to explore the apparent
properties of consciousness, as in the comments by Gary Cziko and John
Burch, above.
-----------------------------------------------------------------------
MArtin Taylor (960502) --

Boys with air guns sometimes like to shoot at big shiny balloons full of
light gases :slight_smile:
----------------------------------------------------------------------
Bruce Gregory (960502.1345 EDT) --

     The problem, as I see it, is that the receiver (not on the football
     field, but in the nervous system) _seems_ to be skirting the edge
     of being a homunculus looking at a screen similar to the one in the
     stadium and subject to changing perceptions.

In a sense, it is a homunculus. But the hierarchical concept in PCT puts
a different spin on the little man in the head. Since the levels of
perception and control are specialized, as we move up the levels of
neural function, we strip away one layer after another from what is
left. When we get to the logic/program level, for example, there are no
longer any categories, sequences, relationships, events ... intensities.
There are only logical processes involving logical variables, and of
course the processes at higher levels, principles and system concepts.

When we have stripped away ALL the levels in the hierarchy, what we are
left with is either nothing, or the Observer. The Observer does not
itself do ANY of the things that the hierarchy does. So the little man
in the head has been reduced to a point-receiver.

The usual homunculus problem arises because of duplication of function.
We say that there is a perception of a red apple, in the form of neural
signals. But to identify those neural signals as a red apple, we need
another level of perception that knows that these signals are a red
apple, and another to know that the signals in the second level are a
red apple, and so on to infinity. That's the basic little man problem.

In the PCT hierarchy, however, there is one level concerned with
creating neural signals corresponding to objects, the configuration
level. There are no more configuration levels above it. Higher
perceptions are not perceptions of configurations, but of motion, event,
relationship, category, etc.. Each level of perception occurs only once,
and is the ONLY level where variables of its type are represented as
neural signals. So when you consider the whole hierarchy, you have all
the levels of perception that make up experience, each level existing in
only one layer and supporting only one type of perception. The Little
Man in the Head can be nothing less than the whole brain.

If this is a workable picture of the hierarchy, it gives us a hint about
consciousness. In the world of ordinary conscious experience, all these
levels of perception seem to coexist in the same space. We don't
experience any layers; the layers are transparent. At any given time, we
don't experience all the levels or everything in every level. We can
focus down to a very narrow range, or widen the focus to include more
information at more levels. You can look at an electric light bulb and
focus on its intensity, or you can widen your field and appreciate that
it is a symbol for Thomas Edison. In any experience, there are many
levels at which you _could_ experience it consciously, but usually only
a fraction of what is possible to experience is actually noticed
consciously. And there is only a single space of conscious experience.

Also, if the hierarchical concept of perceptions is right, higher-order
perceptions are functions of lower-order perceptual signals. If your
attention is narrowed down, say, to logical or computational aspects of
experience, you could not be having those perceptions unless all the
lower levels of perception were working and providing signals, each
level supplying the required inputs to the level above it (I'm ignoring
imagination). This would be a direct proof that the same perceptual
signals may exist either with or without consciousness of them --
contingent, of course, on the hierarchical idea of perception. And since
focusing attention on any one level makes perceptions of either higher
or lower levels invisible to consciousness, we can show that awareness,
as Rick Marken suggests, receives information only from "below" its
current locus in the hierarchy.

     Most of us, including you and Rick, I suspect, would be happier if
     the receiver could be modeled by a control system. But happiness
     is not always granted us.

Teew Treew. Anyway, a control system needs more than an input function.
We need volition for an output function, but following this paths leads
into the mists -- reference signal? From where? About what?
-----------------------------------------------------------------------
Ellery Lanier 960502 --

     I was just curious as to whether the two [Powers and Calow] even
     had heard of each other.

Powers hadn't heard of Calow. Still haven't read his book, but will try
to get it sent by stagecoach. My work has been anticipated in many ways
by others who wrote after I did.
-----------------------------------------------------------------------
Rick Marken (960502.1100) --

     Peter Cariani (960502)

     Peter asked:

     >Where is the "observer"?

     I said:

     > "Outside" of the neural signals (perceptions) that are observed;

The correct answer is "where is 'where'?" "Whereness" is a learned
concept that exists in the hierarchy of perceptions. "Inside" and
"outside" are learned perceptions. "The hierarchy of perceptions" is a
learned perception . Whatever has been learned has been learned by the
hierarchy; the Observer observes the result.

Or so, to be consistent, it seems.
-----------------------------------------------------------------------
Best to all,

Bill P.

[From Peter Cariani (960505.1500 EST)]

[From Bill Powers (960503.0530 MDT)]

Peter Cariani (960502) --

     ... there are numerous examples where "firing rates" (itself a very
     vague term without any mention of which neurons are involved or
     what kinds of time windows are used for the "rate") do not (and in
     some cases, cannot) explain the perceptual distinctions that are
     made -- where other factors such as temporal discharge patterns and
     spike latencies appear to constitute the "signal".

Since rate is simply 1/interval, and changes in firing rate are
certainly "temporal discharge patterns" or changes in spike latencies,
it seems to me that there is no fundamental difference here. If a neural
signal looked upon as a rate of firing changes rapidly as a function of
time, there will of course be changes in the temporal discharge patterns
and spike latencies -- how could it be otherwise?

Bill, you haven't thought this through properly (relatively few people
have thought about it seriously, despite its centrality to all of
neurophysiology). Generally speaking,"rate-codes" mean that
the average number of spikes produced within some
time window (usually assumed to be tens to hundreds of milliseconds) is the
"coding variable", the informational vehicle in the neural spike train
signal. Let's say we have a neuron producing 150 spikes/second. A large number
of time patterns are possible in the output spike trains of this neuron,
but in a "rate-code" they all mean the same thing (i.e. the rest of the system/
the detector/ the receiver, behaves the same way). In a temporal pattern code,
it is the time pattern between spikes that is the coding variable, such that
each spike train having the same number of spikes but with a different temporal
structure has a different meaning. The simplest temporal pattern code is an
interspike interval code where times between spikes carry the information.
Our evidence suggests that an all-order interspike interval representation
at the level of the auditory nerve covaries with the vast majority of human
pitch judgments. All-order intervals include time intervals between
successive and nonsuccessive spikes, so they represent the autocorrelation
of the spike train. When all-order interspike intervals are added from
all frequency regions of the auditory nerve, the population-interval
distribution closely resembles the autocorrelation function of the
stimulus (hence this is another means of representing the spectrum
of the stimulus, in the time domain, and it complements spectral
representations (e.g. average discharge rate profiles in auditory
frequency maps). Many, many auditory neurons will respond to 2 stimuli
that evoke different pitches with the same average firing rate, but
their interval distributions will be different, reflecting the
periodicities of the respective stimuli. In a latency code, such
as is seen in auditory localization, echolocation, and electroception,
the number of spikes produced is irrelevant, but their absolute times
of arrival, relative to a reference event (like the delay between the
onset of an echolocation cry and its echo), are crucial.

The representation of pitch is a singularly complex problem because the
range of audio frequencies overlaps the range of neural firing rates
(which could also be described as an overlap of audio periods with the
range of interspike intervals).

It's not a 1:1 relationship. Neurons generally don't produced sustained
firing rates over a few hundred Hz, yet we hear periodicities well above
this that are hard to explain in terms of "rate-place" (place=cochlear
place=tonotopic place=frequency in auditory jargon) models. What one sees
in the auditory nerve and cochlear nucleus is that discharges can skip
cycles, but because there are many neurons in each frequency channel
carrying similar signals, timing information can exist well above the
firing rate of any one neuron. (It goes up to 4-5 kHz in humans where it
is limited by "synaptic jitter." In other systems, phase-locking can
go higher, e.g. 8 kHz in the barn owl).

As far as I know, this problem is unique
to audition, occuring in no other sensory madality. I commend you for
investigating this complex phenomenon, and would be very interested in
seeing anything recent you have discovered about it.

It's really everywhere if you look for it in the literature.
I've written a review of the evidence that I was aware of roughly
2 years ago (I think I sent you (Bill) an early version of it several
years ago):

Cariani P. As if time really mattered: temporal strategies for neural coding of
sensory information. Communication and Cognition - Artificial Intelligence
(CC-AI), 1995; 12 (2-3): 157-219. Preprinted in: K Pribram, ed. Origins: Brain
and Self-Organization, Hillsdale, NJ: Lawrence Erlbaum, 1994; 208-252.

Our work on pitch is in the very final stages of review, and barring unforeseen
calamities or delays should be out in the next few months. I think I posted
a list of putative temporally-coded systems a year or two ago, but I'll
list a few:
Audition: pitch, frequency selectivity, timbre (temporal pattern)
                                             localization, echolocation (latency)
Somatoception: flutter-vibration discrimination, texture pattern (time)
                         localization (Bekesy's experiments with delayed inputs)
Electroception: very, very precise time-of-arrival system
Taste: Covey and DiLorenzo's findings of time patterns associated with
                                                        taste; these time patterns can be used to electrically stimulate
       rats which then make the same behavioral signs associated with
       the tastant; if the time patterns are scrambled, the signs go away
Olfaction: Macrides (characteristic time patterns for odorants, rat; also
       Laurent's locust data, which shows different time patterns for
       different odorants; whereas the ideosyncratic and wide-band nature
       of discharge rates in olfaction makes these codes very difficult
       to realize in practice; temporal factors in phermone detection
Vision: Chung, Raymond and Lettvin's work on interval codes in the frog;
      Reichardt et al's work on motion detection in the fly -- cross-correlates
             time patterns of spikes
      Opticon & Richmond work on time patterns encoding form, texture, color
      Pulfrich effect: interocular time delays create depth illusions
      color: Benham's top, achromatic "subjective" colors induced by
         time-patterns of black and white; Kozak's work on interval
         distributions characteristic of different wavelength stimulation
Pain: complex time patterns for pain (Emmers, Pain: a spike interval coded
              message in the brain, Raven Press) that can be disrupted by
       electrical stimulation.

If people are interested I can post references, but the list is long.

---------------------------------
     Next to these problems, those of operationally-defining how to
     recognize self-producing sets of signals from some set of
     identified and observed signals is trivial. Consider a set of
     chemical reactions between substrates A-Z, where some combinations
     of substrates produce other substrates, e.g. A + B -> C, and M + O
     + Z -> D + W + F. One can make a directed graph of the reaction
     network and using graph-theoretic (or logical entailment)
     procedures find those sets of substrates that form closed loops,
     i.e. the members of the set through their interactions regenerate
     the members of the set.

Sounds useful. Graph theory, however, can represent only the map of the
interconnections -- the dynamics and other quantitative considerations
aren't contained in a map of logical entailment. In biochemical
reactions there are certainly closed loops, but some of them are control
systems and the rest are not. The difference depends on the loop gain.
If the loop gain is less than 1, I would not call the result a control
system even though there is a small amount of negative feedback. A true
biochemical control system probably has to include an enzyme, which can
insert a gain of as much as (or at least) 50,000 into the loop. That is,
a change in the effect of the enzyme on a reaction is reflected (via the
product) back to the enzyme (by, for example, shifting it from the
active to the inactive state) to produce an effect that is 50,000 times
as large as the initial change. But the system is stable. So not all
closed-loop systems have the same properties.

Yes, I agree. I haven't thought through these issues of the loop gains,
and I haven't tried (yet) to distinguish those recurrent networks that
would be considered to be control loops from others that might be
stable or contingently-stable. Whether the gain matters or not depends
upon whether the "signal" is the amount of something, as opposed to
its presence at all or above some threshold or the time period of
some reaction cycle. One can conceive of all sorts of systems,
some of them being control systems, based on these other kinds of
signalling processes. I don't know if the brain MUST be a network
of feedback controllers in this sense, whether there could be other
kinds of stable systems that use different kinds of signals that are
not scalars. Although I think that the hypothesis that the brain is
a network (hierarchy or heterarchy) of controllers is a very strong one,
I'm not completely sure that there are no other kinds of organizations
that would do. (I'm open to suggestion here, and would find the arguments
for the necessity of a control network very useful.)

I believe it was Warren McCulloch who first misled cyberneticists
(Proceedings of the 5th Macy Conference) by telling them that a closed
loop (even a negative feedback loop) can't be stable if it has a loop
gain greater than 1. This is something to ponder when considering
biochemical control systems with closed-loop gains in the tens of
thousands.

People make mistakes. All of us.

See Hayashi and Sakamoto (1986), _Dynamic Analysis of Enzyme Systems_,
Springer-Verlag (particularly Chapter 6).

I'll try to find it.

     How about them optical interferometry telescopes that were in
     yesterday's science section of the New York Times? Really
     impressive.

Yes, I had seen that in Science, Science News and Sky and Telescope.
Spots on Betelgeuse! Fantastic.

You're going to have to explain, however, why this observation showed up
in this particular post. Is it something about epistemology? I once
asked an astronomer friend why astronomers liked to transform the
almost-plane wavefronts of light that make up the universe into points
and extended objects, by using lenses or curved reflecting surfaces. He
didn't think that was funny.

I can't very well say, how 'bout them Red Socks? can I?
There really wasn't anything deeper intended, but on thinking about it,
the astronomers are making use of the periodic time-structure of light
to make detectors that are orders of magnitude more sensitive than those
based on numbers of photons (luminance levels). There is a connection
between neural time codes and temporal cross-correlation on one hand, and
optical computing using interference effects on the other. Phase matters.

     However you want to characterize Hard and Soft accounts, I want to
     make it very clear that I have always been very much opposed to the
     notion that simply making something more complex confers onto it
     special properties.

Roger. I hope the same thing applies to the general notions of re-
entrance or recursion, patterns, and so forth.

     "Complexity" and organization are two very different things: one
     can have very simple control loops and very complex ones, but what
     makes a control loop a control loop, i.e. what gives it its
     functional organization is not determined by its degree of
     complexity. Isn't this obvious? Let's not conflate these things.

I agree, let's not. I can go on; it's not the re-entrance or recursion
that creates the function called control, nor is it the idea of
feedback, nor is it even the idea of "organization." All those ideas
apply to system which are not control systems as well as those that are,
and hence do not distinguish control from non-control.

Great, I'm glad we have some convergence of understanding on these issues.
I'm just not prepared yet to limit the discussion to control systems per se.
I'll think about it.........

     Secondly, I am not in any way invoking "magical" processes, andI am
     not proposing that "new physical phenomena," in the sense of new
     basic properties of matter, are being created by the evolution of
     functional organizations that are capable of observation. What does
     happen is that the evolution of a particular organization of matter
     permits there to be informational operations like "memory" and
     "modelling relations" (Pattee, Rosen, Kampis), such that
     "observers" are made possible.What does happen is that the
     evolution of a particular organization of matter permits there to
     be informational operations like "memory" and "modelling relations"
     (Pattee, Rosen, Kampis), such that "observers" are made possible. A
     rock is not an observer, a cell is a primitive observer, a human a
     much more elaborated one.

But that is exactly the step that I find magical. Why should memory make
observers possible? Is a phonograph conscious? Why should "modelling
relations" make consciousness possible? Does the relationship between a
computer model of tracking behavior and the actual behaviour make the
computer conscious? You speak as if we have all agreed on the nature of
the phenomenon that we call "consciousness," but what you are really
doing is saying that whatever the term may mean, it is "nothing but"
memory, modeling relations, and so on. That is not explaining
consciousness; it is reducing the problem to fit the kind of explanation
you prefer. That approach was anticipated some time ago by a fellow
named Procrustes.

And I'm always criticizing AI and alife people for "conceptual
Procrusteanism", i.e. only allowing discussion of those things that
a computer can do. I should have been more clear. I don't assume that
an observer need be conscious. The process of observation entails
that an organism or device can make a measurement, record the result,
and act contingent upon that result (e.g. report the observation,
or run it through a predictive model and report a prediction). If one
is in a roomful of other (human) observers watching a science
experiment, it's possible for each observer to make the measurements,
record them, compute their consequences in a model, report the
results, and ascertain if the results are all the same and/or consistent
with subsequent measurements. No consciousness is needed. The problem
of Other Consciousnesses never comes up. Memory and modelling relations
allow artificial devices to participate as "observers" in this arena,
but, no, this does not necessarily confer upon them "conscious awareness".

Memory and modeling relations may give the observer something to
observe, but how can we say that they ARE the process of observation?

Once you have decided that consciousness is nothing but some set of
particular physical processes, your logic them compels you to suggest
that other examples of similar processes must also entail consciousness
-- even a camcorder has to have a tiny bit of consciousness. But the
basis for such claims is not any evidence of consciousness; it is only
the evidence that similar physical processes are occurring. The
relationship of these processes to consciousness remains, actually,
unknown and putative.

     One is not "required" to consider a system as a "control system" in
     order to describe its physical workings, and one is not "required"
     to consider particle-like properties of light in order to describe
     its wave-like properties (interference).

You have it backward, as least as far as PCT is concerned. We did not
start with "control" and from that derive the idea of a control system.
It went the other way. First, there were systems capable of stabilizing
variables outside themselves, maintaining those variables in specific
states despite disturbances tending to alter them. Our first organized
understanding of systems that could behave like this arose out of
engineering and the artifacts built by engineers. What was not generally
understood at first was that the engineers called them "control systems"
because of their (deliberate and intended) resemblance to human beings
carrying out processes commonly called "controlling."

The ideas also came out of physiology as well as engineering.
My remark was to some other part of your response, where you implied that
I was requiring "consciousness" to be in the description. It doesn't
need to be in the description; it doesn't add to a "physical"
description, since its "observables" are incommensurate with
"physical" ones.

One is not required to consider a system as a control system _unless it
is observed to be controlling something_ -- that is, varying its actions
on the environment in such a way as to stabilize some aspect of that
environment against arbitrary disturbances, and doing so in a way that
depends on sensing the environment. While the term control has been used
in many different ways, in PCT it is used in one and only one way, which
can be technically defined so it can't be confused with any other kind
of organized behavior.

So my point here is that if the term "control system" were never invented,
there is nothing that deters one from describing a material system in
terms of mechanics or kinematics or thermodynamics, and there is nothing
that compels one to consider the material system in terms of a
"control system". (I'm perfectly willing to say, however, that some
aspects of the system will remain badly understood until one has the
concept of a "control system". However, its description as a control
system does not add in any way to these other "physical" descriptions.
Matter does what it does.)

The analogy with the wave-particle complementarity is spurious, because
in fact the behavior of matter can be seen as involving either waves or
particles, depending on how you set up the experiment and analyze it.
This is not true of controlling in the PCT meaning. If you set up an
experiment with control and determine that control is occurring, the
proper analogy is with setting up a pair of slits and observing that
there is interference, which demonstrates the wave nature of matter. But
there is no second way of setting up the behavioral experiment, that I
know of, which will produce a different result, analogous to showing the
particle nature of matter.

In the wave-particle situation, there are two different and mutually
exclusive results obtained: interference patterns, or lack of
interference patterns. This is not a case where there are two competing
explanations of the same phenomenon. There are, in fact, two different
phenomena. This is not just a difference in points of view; one cannot,
simply by changing interpretations, make interference fringes occur when
they are not occurring. It is necessary to change the physical situation
in order to change the phenomenon.

We have a different situation when we are explaining control phenomena.
Here we have a single phenomenon, the stabilization of variables against
disturbances (shorthand for a more complete description). Merely
changing one's perspective (or the word that is used) does not remove
this stabilization, nor (so far) is there any phenomenon that we call
control in which this stabilization does not occur. There are, of
course, other kinds of phenomena which do not entail control, but their
existence does not negate control, and is not an alternative to control.

Different descriptions can entail different sets of observables, and yes,
then the phenomena are different. If there is overlap between sets of
observables then competition in the prediction of those observables can
ensue, but if there is no overlap, one is hard-pressed to cateorically
reject one model in favor of the other.
In the case of a (classical) thermodynamic account of a device,
one would have temperature, pressure, and volume readings, whereas in
its description as a control system, one has the observed values of
the signals, which may or may not directly relate to the thermodynamic
variables. No measurements have been altered; the two models just don't
directly relate to each other; they're framed in incommensurable terms.
This doesn't entail a postmodernist collapse of "observation"; no inter-
ference fringes are created or destroyed. It's more like a
non sequitur. I do think that the wave-particle duality has this
character of incommensurable observables: there are two different
sets of phenomena.

     Does a system whose organization is described as a "control system"
     call into being new basic properties of matter? No.

No, not at the lowest level of organization. But there are, as you know,
emergent rules that show up at higher levels, and they can't be reduced
to the rules governing the lower levels. Join an inductor and a
capacitor, and you get rules that are not reducible to the rules
governing either inductors or capacitors. It is possible that
consciousness/awareness/Observing is such an emergent phenomenon. I
don't know. We can easily show HOW resonance emerges from combining
inductors and capacitors (conservation of energy is one way), so we can
prove that true emergence is occurring. But until we can show HOW
consciousness emerges from neural processes, we can't establish the
link.

     Might such an organization permit such a material system to become
     an "observer", in some very primitive sense? I think so.

So, here, since I'm not assuming that observer=consciousness, this is
tantamount to asking whether a bunch of electrical parts such as
inductors, capacitors, wires, etc might self-assemble over long
evolutionary periods to get something like a robotic device with
sensors, effectors, and some internal capacity for modelling the
world. THis isn't so unreasonalbe, I think.

Bruce Gregory (960502.1345 EDT) --

     The problem, as I see it, is that the receiver (not on the football
     field, but in the nervous system) _seems_ to be skirting the edge
     of being a homunculus looking at a screen similar to the one in the
     stadium and subject to changing perceptions.

In a sense, it is a homunculus. But the hierarchical concept in PCT puts
a different spin on the little man in the head. Since the levels of
perception and control are specialized, as we move up the levels of
neural function, we strip away one layer after another from what is
left. When we get to the logic/program level, for example, there are no
longer any categories, sequences, relationships, events ... intensities.
There are only logical processes involving logical variables, and of
course the processes at higher levels, principles and system concepts.

When we have stripped away ALL the levels in the hierarchy, what we are
left with is either nothing, or the Observer. The Observer does not
itself do ANY of the things that the hierarchy does. So the little man
in the head has been reduced to a point-receiver.

I still don't see how reducing the observer down to a Point Receiver solves
the problem of awareness. (I accept your argument about the homunculus --
your point receiver is not equivalent to the whole organism. Point well
taken.)

Rick Marken (960502.1100) --
     Peter Cariani (960502)
     Peter asked:
     >Where is the "observer"?
     I said:
     > "Outside" of the neural signals (perceptions) that are observed;

The correct answer is "where is 'where'?" "Whereness" is a learned
concept that exists in the hierarchy of perceptions. "Inside" and
"outside" are learned perceptions. "The hierarchy of perceptions" is a
learned perception . Whatever has been learned has been learned by the
hierarchy; the Observer observes the result.

As a neuroscientist, I'd still like some notion of where this Point Receiver
is supposed to be. The explanation just doesn't hang together for me.

Peter Cariani