June Gloom

CSGnet Archive CSG2005
21

[From Bill Powers (3005.06.22.1252 MDT)]

Richard Kennaway (2005.06.19.2148 BST) –

I’ve just put the current
version on the web at


http://www.cmp.uea.ac.uk/~jrk/Robotics/ArchyDistrib2005Jun.zip

There’s a readme file that describes how to use it.

Still trying to figure it out – so far all that happens is that things
topple over and fall to the ground. But very realistically!

I don’t recall if I ever made
available the paper I presented at Control 2004, but it’s now linked to
from

http://www.cmp.uea.ac.uk/~jrk/Robotics/

Yes, I saw it and I still feel that the opening arguments should have
some Bach playing in the background. Beautiful. I especially like your
succinct comments about Rodney Brooks’ subsumption architecture: "
In subsumption, higher-level agents
operate instead of lower level agents; in HPCT, higher-level
controllers operate by means of lower level
controllers."

I don’t think the difference could be stated more accurately or in
fewer words.

Best,

Bill P.

22

[From Bill Powers (2005.06.22.0108 MDT)]

Jason Gosnell (2005.06.20.2005CST) –

This seems to be the most
critical thing to me. You can’t really fool your

nervous system–not for long anyway. It’s funny that the mind or ego
or

whatever can set this up. It seems that any successful movement must
include

things as they are now–before I can move–an acceptance of “what
is” I mean

as noting what my actual experience is before idealizing a solution.
Maybe

it’s OK as a hypothesis to have an idea, but one can follow this and
ignore

error signals for quite some time…

It’s not so much “fooling” your nervous system, I think, as
trying to contradict what it has become set up to do. You can try to
exert “will power,” but that arouses opposition. After all, why
would action feel like “trying” if there were no internal
objections to it? I think people in Zen have been saying things like that
for some time.

Does anyone else have an
organized or simple approach to the MOL?

Working on it. Tim Carey and several of his colleagues (in Scotland) now
use nothing else in their counselling practice. Right now the procedure
is a little formal and awkward, but the goal is to figure out the right
approach that will just focus on what works about it without any special
language. What we try to do is to get people to become aware of
the viewpoint they had previously been aware of other things from.
This requires some hard listening, but as one becomes more adept at
picking up signs of a higher level, it gets easier and more relaxed.

I have just a totally random
approach of stopping and asking myself what is going on within me
now?

That’s probably as good as any when you’re trying to do it by yourself.
It’s not easy, though – thoughts flit by before you even realize they
happened, and you miss things that another person might catch if you were
talking out loud. And it’s easy to forget to do it again, and again…
you get too interested in what you find and get stuck at another
level.

What’s this all about? There’s a
lot of thinking BS I sometimes get caught

in and confused by–my habitual way of thinking–sometimes I
magically

plunge through that right into an awareness of the problem…and then
from

there is just an ongoing process of observation and responding
differently.

Even more difficult than getting to the problem sometimes is coping with
or

experiencing the feelings. So, this is another kind of skill–how to hold
a

feeling in awareness and not get pulled into identification or disowning
it.

I know one way is to assist with self-talk: "I am aware that the
feeling of

anger is in me now", or saying Hello to anger even and even
imagining

breathing with the feeling. Sometimes that helps massage it a bit.
Somehow

the relationship to the feeling is modified through the self-talk. Is
an

explanation of this assistance from internal speech included in PCT? I
can

see that it allows the observer to become more active–there is some

distance from the experience and yet some connection.

Good thoughts, and very similar to mine. Speaking aloud to myself (even
when I’m alone, as I am a lot) seems a very different experience from
just thinking the same things. But again, a lot of the problems in solo
MOL are simply forgetting to keep going. So you start being aware that
you are angry, that’s a good step. But you have to go on to notice what
it is about, and what you are wanting to DO when you’re angry, and how
you feel about that, and why you don’t do it, and so on. If the problem
doesn’t resolve as soon as you become aware of it, the real problem is
probably not at that level. It really helps to be talking with someone
else who knows what is going on, and who can say “Whoops, what did
you just say?”

I have included an interesting
set of questions for breaking identification

with thought by Byron Katie…

Is it true?

    includes

the variation on this question…what’s the Reality of it?

Can I absolutely know it is true?

includes…where’s my proof?

How do I react when I think that thought?

Who would I be without that thought? or, what would my experience be
without

that thought?

At some point I think you have to pop out of the thought level completely
and get out of the “thoughts about thoughts” framework. That
gets you to the experience without thought position, though I find that
very hard to maintain, not to mention boring (there’s a background
thought for you!).
This is all very personal stuff. We aren’t all organized the same way –
even if we share types of levels, I doubt that we share the
details. “Seek out your own salvation with diligence,” were, as
I recall, the Buddha’s last words. He probably had this mostly figured
out, or as figured out as it gets, and realized that it’s not going to be
the same experience for everyone. And you just have to start from where
you are and take it one step at a time. Reorganization doesn’t follow a
predictable path. If it did, it wouldn’t work so well.

Best,

Bill P.

23

[From Bill Powers (2005.06.22.0140 MDT)]

Bjorn Simonsen (2005.06.21,13:20 EST) –

The output signals from all the
levels mentioned above become reference

signals for levels below. It is difficult to say what happens, but
my

control at the different levels are hindered. Maybe I become anxious, but
I

am sure I don’t feel very well.

No, it isn’t easy to change the value of a reference signal _in a
wanted

way_ by just saying No … when I control a perception at a lower
level.

Right. Good story, by the way.

Maybe we some times have to move
all the way to the System Concept level if

we are going to change our behavior. But it isn’t predictable how the
wires

become connected if we change for other perceptions at “the top
level”.

A lot of the problem is that we think of changing our behavior, but of
course when we think that, we are opposing some other control systems
inside us that need that behavior for what they are controlling.
So all we’re doing is consciously taking one side of a conflict, the
other side of which is not in consciousness. That doesn’t work. We have
to change what we want, not our behavior, and we have to change it at a
high enough level. The change in behavior will follow.

When I quit smoking I didn’t experience any great surge of will power. I
just somehow knew that I wasn’t going to smoke any more – that I was
going to let it go. Mike Acree sent me a book on stopping smoking, but I
make the decision to let it go before I even started reading the book. I
thought something like, “OK, if this book can make me stop smoking,
I give it permission to do so.” Then I read it, and never had
another cigarette. Don’t ask how I did that. I don’t know. Maybe the book
played a part - do recall a few helpful ideas. But they wouldn’t have
helped if I had still been resisting.

Best,

Bill P.

24

[From Bill Powers (2005.06.22.0151 MDT)]

Bjorn Simonsen (2005.06.21,16:00 EST) --

Our nervous system doesn't know anything, it just functions.

E.G. The brain functions because particular transmitters conduces to the
"transport" of neural signals. Acetylcholine is the prototype of many
diverse chemical substances that can be released from diverse nerves and
neurons in the brain as the all-important link in the signaling process.

That is like saying that a radio functions by electrons and holes moving through transistor, resistors, and capacitors, so the radio doesn't produce any music. Reductionism explains nothing, because there are different laws of nature that come into play at different levels of organization. You could organize acetycholine and all the other neurotransmitters differently and end up with a nonfunctioning brain, just as you could wire up transistors, resistors, and so on at random and end up with a nonsense device that did nothing useful. What makes the brain work as it does is the organization of its parts, not the parts themselves.

I say that our nervous system knows everything there is to know, because that is one of the things it is organized to do.

Best,

Bill P.

25

[From Bjorn Simonsen (2005.06.22,10:30 EST)]

From Bill Powers (2005.06.22.0151 MDT)

I really hope you get enough sleep.

What makes the brain work as it does is the
organization of its parts, not the parts themselves.

Yes and No. The brain is at one moment organized as it is. Then you start
smoking for the first time. After 10 seconds the nicotine reaches the brain.
Nicotine (a supplement to acetylcholine) works as a transmitter
(acetylcholine) on _one_ type of receptors. The stimulation of the receptor
becomes far greater than would normally be the case for acetylcholine
itself. Now, more channels for sodium (or alternatively one of the other
ions) opens. The entry or the exit of ions are reflected in transient change
in potential difference in the target cell. This change will be greater or
lesser because of nicotine.

The action potential in the receiver cell becomes greater (or lesser)
because of nicotine.

The brain is organized as it is. After nicotine the reference value becomes
greater (or lesser).

The reference value in a loop is changed. Now I thought the perception would
change.
Nicotine is a part of the organization. The effects of the brain is also
dependent on the parts.

Am I still wrong?

bjorn

26

[From Richard Kennaway (2005.06.19.2148 BST)]

[From Bill Powers (3005.06.22.1252 MDT)]
Still trying to figure it out -- so far all that happens is that things topple over and fall to the ground. But very realistically!

The robots themselves won't look like they're doing anything, but the control systems are indeed hard at work! (As is often the case with well-functioning control systems.) 'f' turns on a random force applied to the broomstick balanced on the left-hand robot. The amplitude is such that after a few hundred simulated seconds, it eventually overwhelms the robot's ability to counteract it.

You can reach into the world and apply forces to objects with right-click-drag, so you can do things like pick up a block and drop it on the robot's back. Or gently pull on the top of the broomstick and see it remain stationary, while the robot underneath it moves to oppose the disturbance.

I also discovered that pageup/pagedown zooms in and out -- helpful if your mouse doesn't have a middle button.

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

27

[From Bjorn Simonsen (2005.06.22,11:58 EST)]

From Bill Powers (2005.06.22.0108 MDT)

Working on it. Tim Carey and several of his colleagues (in Scotland)
now use nothing else in their counselling practice. Right now the
procedure is a little formal and awkward, but the goal is to figure out
the right approach that will just focus on what works about it without
any special language. What we try to do is to get people to become
aware of the viewpoint they had previously been aware of other things
from. This requires some hard listening, but as one becomes more
adept at picking up signs of a higher level, it gets easier and more

relaxed.

I should not have any thoughts about MOL, I am a too little specialist in
MOL. But I am interested. I could have kept my thoughts for myself, but if
Djerzinski is correct saying "a thought that evolves without any partners to
discuss often risks to drown in isolation and insanity, ...". Therefore:

MOL is a method where the clients are helped to discover their own conflicts
(or a conflict between two people) and helped to control a perception at a
higher level where the original conflicts are fringed out of consciousness
(the way I understand MOL). The method is more educational than medical. It
is more like training than curing.

May I be frank to paraphrase a technique from Stewart B. Shipiro's "The
selves inside you". He was a coach for the Norwegian defense (personnel
development). He knew something about Berne's transactional analysis (and
much more) but nothing about PCT.

Here are my thoughts:

The client sits in chair A, one of 4 chairs and tells the Adviser who sits
in chair D about his failure. The other chairs are placed as below.

C D

A B

When the client in chair A has described his failure, the Adviser in chair D
asks him to take place in chair B. And he adds: "OK, we have now heard your
message. Let us hear the message in full. Please sit here and talk to
yourself about what you heard from chair A".

Ask the client to change between chair A and chair B and talk to himself in
the other chair.

When the Adviser realizes that the client is able to understand his own
conflict, he asks the client to take place in his chair D, and he adds: "Now
you sit in the observers chair. Can you describe a common pattern in what
you described in chair A and B". (The last statement is important and
substantial for the method and must be further developed).

After the client has described a common pattern in the conflicting messages
from chair A and chair B, the Adviser asks the client to take place in chair
C and he adds: "Can you imagine a way to control your perceptions of "the
world" within the common pattern you mentioned in chair D". (Also this
statement must be further developed).

"Describe how you control these perceptions".

When the Adviser realizes that the client is able describe how to control
his perceptions at a higher level, he asks the client to take place in his
chair D again. And he adds: "Please tell me what you have taken part in in
this session".

Let me make this more specific.

Oliver (a NN) is my client. He tells me he is a failure. He is bone-tired
and at the same time he is so behind with the report he is working with and
responsible for. "I am never going to amount to anything" he said.

Telling me this, he sits in a chair A, and I sit in chair D, the observer's
chair, about 2 meters apart.

I ask Oliver to sit in chair B and really tell Oliver in chair A who he is.
Oliver in chair B points to chair A and criticize: "Come on. Get off your
butt and do something! You are not interested in anything but yourself. I
have nothing but contempt for you. I am disgusted with you."

I ask Oliver to go back to chair A to receive and respond to the message
from chair B. He said: "I don't feel I have any energy. I'm doing the best I
can. I am a parent and I must earn money and I must do my job for the
children."

Oliver changes between chair A and chair B and responds to the messages from
the other chair.

I stop when I feel Oliver knows the "two Olivers". Then I ask him to take
place in my chair, the observer's chair. And I say: Sit here. Now be the
observer of these two. You heard everything. What are your comments?".
Later: "Can you imagine a way to control your perceptions of "the world"
within the common pattern you mentioned in chair D"

When I feel Oliver describes the common pattern very well I ask him to take
place in chair C. And I add: "There are many ways and you may make many
programs to accomplish your wish to be a responsible family father. Please
describe a program to be a responsible family father where you avoid the
conflict you described in chair A and B".

After Oliver has described the new program I ask him to go back to my chair,
the Advisors chair and tell me what he had taken part in.
He said:
"When I was seated in chair A and B I told you about my failure and my
judgment about myself. In chair D, I discovered my conflict.
You asked me about a common pattern in my descriptions from chair A and
chair B. "The common pattern is my wish to be a responsible family father
who gives the family safety. Enough money and enough care".
Then you asked me to take place in chair C. There I described a new program
how to be a responsible family father earning a lot less money in a new less
responsible job".

........
I think it is important that the Adviser don't tell to much. I neither think
it is important that the Adviser asks if the client is satisfied. It is more
important that the client get insight and knowledge about himself.

bjorn

28

[From Bill Powers (2005.06.22.0904 MDT)]

Bjorn Simonsen (2005.06.22,10:30 EST) –

Yes and No. The brain is at one
moment organized as it is. Then you start

smoking for the first time. After 10 seconds the nicotine reaches the
brain.

Nicotine (a supplement to acetylcholine) works as a transmitter

(acetylcholine) on one type of receptors. The stimulation of the
receptor

becomes far greater than would normally be the case for
acetylcholine

itself. Now, more channels for sodium (or alternatively one of the
other

ions) opens. The entry or the exit of ions are reflected in transient
change

in potential difference in the target cell. This change will be greater
or

lesser because of nicotine.

The action potential in the receiver cell becomes greater (or
lesser)

because of nicotine.

The brain is organized as it is. After nicotine the reference value
becomes

greater (or lesser).

What is there in the explanation up to this point that makes the
reference value change? Why doesn’t a perceptual signal simply change? I
assume that all the facts you report are correct. But what effect do the
changes you report have on all the rest of the brain? Why is an
increase in stimulation of a receptor treated as a good effect or a bad
effect by the rest of the brain? Why isn’t it just treated as an
increase? The result of an increase of a signal will depend on what the
reference level for that signal is, and how the associated control system
is organized. Why isn’t the sudden change caused by the nicotine simply
treated as a disturbance, and corrected?
You could make a similar presentation for the effects of any stimulus on
any part of the nervous system. But there is nothing in such a
description to predict how the whole system will act when such a change
happens. To make a prediction, you have to know what all the control
systems are, what their parameters and reference settings are, and how
they relate to the other control systems in the same brain. The chemistry
of neurotransmitters is never going to explain that.
To return to my analogy with the radio: you could tell me everything
about transistors, capacitors, inductors, resistors, power supplies, and
so forth, right down to the details of electron and hole flow, and that
would still not tell us anything about what the circuits in the radio do.
Before you can say what the circuits do, you have to know how all these
parts are connected together, and what happens as a result of connecting
them together one way rather than another. You have to know, for example,
that connecting a capacitor in parallel with an inductor creates a
resonant circuit that will have some natural frequency of oscillation.
Neither the capacitor nor the inductor has a natural frequency of
oscillation (discounting the parasitic capacitance between the windings
of the inductor). A natural frequency of oscillation is an emergent
property of the organization made up of a capacitor and an
inductor connected in parallel.

This is what I mean by saying that there are natural laws at every level
of organization which are not simply summaries of the laws that hold at
lower levels.

The reference value in a loop is
changed.

Why do you say that? How do you know the nicotine isn’t affecting a
perceptual signal, or the way a comparator works, or an output signal?
How do you know it isn’t changing the gain of the system without any
direct effect on signal amplitude? You see the problem. The chemical
details tell us nothing about what the brain is DOING, or how its parts
are connected to each other.

Now I thought the
perception would change. Nicotine is a part of the organization. The
effects of the brain is also dependent on the
parts.

Am I still
wrong?

You’re not wrong at the level of describing the detailed chemical effects
of nicotine. I assume you have looked that up and got it right. But I
think any implication that this explains the behavior of nicotine addicts
is wrong.

Best,

Bill P.

29

[From Bill Powers (2005.06.22.0931 MDT)]

Richard Kennaway (2005.06.19.2148 BST) --

The robots themselves won't look like they're doing anything, but the control systems are indeed hard at work! (As is often the case with well-functioning control systems.) 'f' turns on a random force applied to the broomstick balanced on the left-hand robot. The amplitude is such that after a few hundred simulated seconds, it eventually overwhelms the robot's ability to counteract it.

My impression is that these systems aren't very stable. Why should the effects of a randomly-varying disturbance accumulate? They don't do so in models of tracking behavior -- those models will just go on varying their outputs and counteracting the disturbances forever, as long as you want to let the simulation run. The effects don't accumulate in the inverted pendulum demos, either. I'm puzzled by the damping factors that seem to be build into objects, the linear and angular damping factors. Why are they there? Why not just masses that behave normally? It just looks to me as if there are some inherent instabilities in the current design.

You should be able to bang on these systems any way you like (within reason) and for as long as you like without making them lose control.If they're stable at first, they should go on being stable.

I wish all this were being done in Delphi so I could take part. Darn my stupid brain. As it is I can only look at outward appearances without any idea of what's going on under the hood (i.e., bonnet).

Best,

Bill P.

30

[From Bill Powers (2005.06.22.0942 MDT)]

Bjorn Simonsen (2005.06.22,11:58 EST) –

May I be frank to paraphrase a
technique from Stewart B. Shipiro’s "The

selves inside you". He was a coach for the Norwegian defense
(personnel

development). He knew something about Berne’s transactional analysis
(and

much more) but nothing about PCT.

Here are my thoughts:

The client sits in chair A, one of 4 chairs and tells the Adviser who
sits

in chair D about his failure. The other chairs are placed as
below.

C
D

A B

Etc.
One of my ideas about MOL is that it is part of every psychotherapy that
works. In other words, all the successful therapists are successful
because they somehow manage to get people to back off from one point of
view and find out what higher-level systems inside them are doing,
solving problems along the way. One implication is that the
differences between all the different approaches are unnecessary
decorations; they don’t do anything to help and they may actually
interfere with the process. That’s my opinion about the chairs in your
example. I don’t think it’s necessary to use this indirect way of
examining your own thoughts from another point of view to achieve results
that are just as good or maybe better. If you asked for a technical
explanation of the effects that moving from one chair to another will
have, according to whatever the underlying theory is, I don’t think you’d
get an answer. It’s just something someone thought up. The basic idea of
having a person describe a problem, and then describe attitudes toward
and thoughts about that problem, and then move to still another
viewpoint, is perfectly in line with MOL. But what is the rest of this
technique about? I think the chairs are just a distraction, a little
pretending in a situation where what we really want is close attention to
what is actually going on in one’s head.

And why does the therapist choose what the client is to talk about, and
how the client is going to solve the problem? The therapist can’t
possibly know the best route for untangling the client’s difficulties –
what to deal with first, so the next problem becomes accessible, and so
on. And the therapist certainly can’t know what solution will work for
the client, temporarily or permanently. I don’t think the therapist can
do anything but get in the way by interfering in the natural processes of
problem-solving that are going on in the client. What the therapist can
do is keep the process working in the right direction – up a level, that
is, not toward any preconceived end.

If you looked into any psychotherapeutic techniques that have more than
average success, I think you’d find that they all involve doing something
that calls a person’s attention to the thought or attitudes or feelings
behind what is in the foreground of attention. And they would do this
repeatedly, not once or twice, exploring whatever problems come up as
this process goes on but always aiming toward a higher-level view.
Whatever else a successful psythotherapy entails is just a matter of
style or atmosphere, and either has no effect at all or slows the
progress of therapy.

My opinion is that a therapist should do just the things that have a
beneficial effect, and forget the rest.

Best,

Bill P/

31

[From Richard Kennaway (2005.06.22.1916 BST)]

[From Bill Powers (2005.06.22.0931 MDT)]
My impression is that these systems aren't very stable. Why should the effects of a randomly-varying disturbance accumulate? They don't do so in models of tracking behavior -- those models will just go on varying their outputs and counteracting the disturbances forever, as long as you want to let the simulation run. The effects don't accumulate in the inverted pendulum demos, either.

They don't accumulate, it's just that the disturbances I generate are normally distributed, so eventually a random peak comes along that is bigger than the control system can deal with.

I've just put a new version up at
http://www.cmp.uea.ac.uk/~jrk/Robotics/ArchyDistrib2005Jun22.zip
It ensures that the disturbance is never greater in magnitude than a given amount. I believe that the robot will now balance the pendulum indefinitely. I'll leave it running overnight and see what happens.

I've added a visual indication of the maximum allowed force.

Actually, there's one small effect that may cause the robot's feet to creep on the ground, and the pendulum to creep on the back of the robot: both the feet of the legs and the foot of the pendulum are spherical. So the slight rolling around they do can result in them shifting. This could eventually result in loss of control, e.g. because the pendulum slips off the edge. I could fix the pendulum to the robot with a ball and socket joint; the real solution to the feet creeping will be to make the robot take a step whenever it finds it has to reach too far from its normal posture to keep the pendulum upright.

I'm puzzled by the damping factors that seem to be build into objects, the linear and angular damping factors. Why are they there? Why not just masses that behave normally? It just looks to me as if there are some inherent instabilities in the current design.

The linearDamping and angularDamping values I just copied from other Novodex example programs as recommended default values. They fill the world with slightly viscous air. You can set them to zero and I find the simulation works just as well.

I believe their main purpose is to damp out any fictitious energy that might be introduced to a simulation by numerical inaccuracy in the solution of the differential equations.

You should be able to bang on these systems any way you like (within reason) and for as long as you like without making them lose control.If they're stable at first, they should go on being stable.

I wish all this were being done in Delphi so I could take part. Darn my stupid brain. As it is I can only look at outward appearances without any idea of what's going on under the hood (i.e., bonnet).

Someone on one of the Novodex support forums asked about a Delphi-compatible version as well, but I don't know if it figures in their plans.

-- Richard Kennaway

32

[From Bjorn Simonsen (2005.06.22, 20:45 EST)]

From Bill Powers (2005.06.22.0942 MDT)

...... One implication is that the differences between all the different
approaches are unnecessary decorations; they don't do anything
to help and they may actually interfere with the process. That's
my opinion about the chairs in your example. I don't think it's
necessary to use this indirect way of examining your own thoughts
from another point of view to achieve results that are just as good
or maybe better.

That is good enough for me. Thank you. You did develop the MOL and you have
an experience doing the MOL. I put the chairs in a drawer.

And why does the therapist choose what the client is to talk about,
and how the client is going to solve the problem?

I must have explained myself very bad. It has never been in my understanding
of Mol that the therapist should choose what the client should talk about.
I think I wrote:

When the client in chair A has described his failure, ..

and this description should be his comment after the therapist had wished
welcome and asked what if he could be to any help.
I neither thought that the therapist should choose what the client should
talk about. I must really have expressed myself very bad. I have some
experience with Carl Rogers. And I see some resemblance in roles and rules
in Client centered therapy and MOL.

I don't think the therapist can do anything but get in the way by
interfering in the natural processes of problem-solving that are going
on in the client. What the therapist can do is keep the process
working in the right direction -- up a level, that is, not toward any
preconceived end.

I absolutely agree. I thought you could read that role out from my comments
to change chair.

I am thankful for your comments.

Bjorn

33

[From Bill Powers (2005.06.22.1252 MDT)]

Richard Kennaway (2005.06.22.1916 BST)--

My impression is that these systems aren't very stable. Why should the effects of a randomly-varying disturbance accumulate?

They don't accumulate, it's just that the disturbances I generate are normally distributed, so eventually a random peak comes along that is bigger than the control system can deal with.

I think it's more than that.If you put the viewpoint close to the robot with the pendulum and put a brief impulse into the pendulum, you can see a lot of damped oscillations that go on for some time afterward. This says that by applying timed disturbances of small amplitude you can make the oscillations build up to a considerable degree. A random disturbance will have some component at the critical frequency, and if the phase doesn't shift too fast the pendulum will amplify the oscillations at that frequency. I think that's where the loss of control comes from.

The response you want after an impulse disturbance is a smooth return to the starting conditions, with no overshoot, as in the inverted pendulum demo. Try adjusting the rate feedback to get rid of the multiple oscillations after a brief disturbance. You'll probably have to adjust it in all the degrees of freedom. To get perfect stability at both levels of control, the lower-level response to a disturbance should look as much like a simple exponential return to initial conditions as possible, so the higher system thinks it's controlling an environment with a single integral lag in it. If you have mass in the control loop you need rate feedback (derivative of position) at the lower level (like the "phasic" component of the stretch reflex). Otherwise you're always working close to a 180-degree phase lag, which produces the instability.

The linearDamping and angularDamping values I just copied from other Novodex example programs as recommended default values. They fill the world with slightly viscous air. You can set them to zero and I find the simulation works just as well.

I believe their main purpose is to damp out any fictitious energy that might be introduced to a simulation by numerical inaccuracy in the solution of the differential equations.

That's likely; I've had to put damping into some of my models for the same reason. I like "fictitious energy." Good term. In closed-loop systems you don't have to be concerned with fictitious energy unless the model is really bad, but when you want to show the open-loop behavior of the environment, sometimes it has to be dealt with.

Someone on one of the Novodex support forums asked about a Delphi-compatible version as well, but I don't know if it figures in their plans.

I guess I could resurrect my Visual C++, but I'd feel as if I were diving into the deep end of the pool without my water wings. I hang onto Delphi because it takes care of a lot of Windows stuff I don't understand.

Best,

Bill P.

34

[From Bill Powers (2005.06.22.1313 MDT)]

Bjorn Simonsen (2005.06.22, 20:45 EST)--

I must have explained myself very bad. It has never been in my understanding
of Mol that the therapist should choose what the client should talk about.

My apologies -- I didn't read your words carefully enough.

Best,

Bill P.

35

Bill Powers (2005.06.22.0108 MDT)

[Jason Gosnell (2005.06.22.18:55CST)]

Bill Powers (2005.06.22.0108 MDT)

Reorganization doesn’t follow a predictable path. If it did, it wouldn’t work so well.<<

As always, your feedback is appreciated: I now think this is so…there is something about the seeming disorder of it that makes it go as needed…it like wobbles. Or maybe that is the only way it can go as that is how things are generally going…in seeming or apparent disorder…even though there may be some organization to this.

At some point I think you have to pop out of the thought level completely and get out of the “thoughts about thoughts” framework. That gets you to the experience without thought position, though I find that very hard to maintain, not to mention boring (there’s a background thought for you!).<<

That’s right I think. The questions below (all self-questions; observer questions; koans; even Socratic styles) only work if you do as Zen master Dogen says…“take the backward step–withdraw energy from deluded thought–and shine the light of awareness inward.” This is actually the way he describes the basic move of “meditation.” So, if you use questions they must penetrate through the wall of repetitive story lines. In Zen and her method this is done by placing more attention in the body or sometimes they say the inner energy field of the body. So, the heart and body or your “felt-sense” needs to be included somehow. What does PCT say of this? Is this like including all of the different levels of control–getting in touch with the experiences that are not caught in the story perhaps? Or, is this merely one way to contact the observer–to have more “presence” as it were rather than having most all of the energy in the head? Or, rather than us having the energy pumped into thinking, the energy of facing the self is available when turning the light of awareness inward–that is, energy is not supplied to thought-desires-gaining-gaining things…but is supplied more and more to a pure form of awareness itself?

This whole issue of shining the light inward doesn’t preclude the whole activity of imagination though…at least I think not. Otherwise these fresh and empowered ideas one gets wouldn’t materialize in the field of awareness. This is an interesting issue as it seems to me that there are two types of imagination…1. Is basically your programming as it stands now–your stories of self/other etc., 2. Is an imagination that comes out of the observer’s field. This is hard to describe, but there is a kind of imagining that seems empowered by the observer–it may use previous learning–language/ideas, etc, but it is not exactly limited or bound-up by the previous learning. There is something about having more of the observer present that changes all of this whether it is thinking, imagining or feeling…or even action. I think so anyway at this time. I don’t know–perhaps it is another phenomenon.

Dogen then goes onto say after taking the backward step…“…then your original face will be manifest.” This I think is referring simply to the observer which includes all of experience and is free from it at once. This observer presence I think is the original face and it includes the whole of experience. My understanding now is that “it” (original face–religious feeling-experience–etc.) is not apart from everyday experience as people might imagine. But, the observer is not completely limited by his background of learning…so it is the original face–not contained or stained by experience. If so, then this move allows for some freedom in the midst of difficult or even pleasurable experiences.

This is all very personal stuff. We aren’t all organized the same way – even if we share types of levels, I doubt that we share the details. “Seek out your own salvation with diligence,” were, as I recall, the Buddha’s last words. He probably had this mostly figured out, or as figured out as it gets, and realized that it’s not going to be the same experience for everyone.<<

The story is that he developed different approaches to help different students, so there are a lot of methods in classical Buddhism. But, supposedly he always maintained that quote…sometimes translated as “Be a light unto yourselves.”

Regards…Jason Gosnell

All electronic mail communications originating from or transmitted to Bridgeway Center, Inc. are subject to monitoring. This message and the information contained in it, which may consist of electronic data attachments, are the confidential and proprietary communications of Bridgeway Center, Inc. and are intended to be received only by the individual or individuals to whom the message has been addressed If the reader of this message is not the intended recipient, please take notice that any use, copying, printing, forwarding or distribution of this message, in any form, is strictly prohibited. If you have received this message in error, please immediately notify the Bridgeway Center, Inc. Privacy Officer at (850) 833-7540 and/or forward the message to hipaa@bridgeway.org and delete or destroy all copies of this message.

36

[From Bjorn Simonsen (2005.06.23,11:00 EST)]

From Bill Powers (2005.06.22.0904 MDT)

What is there in the explanation up to this point that makes the
reference value change? Why doesn't a perceptual signal simply
change? I assume that all the facts you report are correct. But
what effect do the changes you report have on all the rest of the
brain? Why is an increase in stimulation of a receptor treated
as a good effect or a bad effect by the rest of the brain? Why
isn't it just treated as an increase? The result of an increase of
a signal will depend on what the reference level for that signal
is, and how the associated control system is organized. Why
isn't the sudden change caused by the nicotine simply treated
as a disturbance, and corrected?

Your questions put me on a test. I am sorry I will use so many words, but I
cannot do it in another way.

What is there in the explanation up to this point that makes the reference

value change?

Dendrites act as receiving stations for incoming signals from other neurons.
If the signals are strong and/or sustained, they will be conducted along
the branch of the dendrite down to the cell body.
If the new net difference in voltage is large enough within the target cell,
it causes the sodium channels to open. Then a potential signal is generated.
An action potential is always the same size, typically some 90 millivolts.
( [From Bruce Abbott (2004.12.21.1800 EST)] and other places.
Because an action potential cannot be made bigger, the neuron will instead
generate more and more action potentials. The frequency will increase.

Conclusion: The frequency in the addresser neuron releases an amount
acetylcholine in the synapse. On the outside of the target neuron there
are special proteins called receptors. When the transmitter (acetylcholine)
meet the receptor it is formed a new chemical molecule of the two. This
molecule opens the channels for sodium or alternatively one of the other
ions. The entry or exit of any of these ions will be reflected in a
transient change in potential difference in the target cell. This signal
goes to the cell body.
A certain frequency in the addresser neuron results in a certain frequency
in the axon in the receiver neuron.
And Now. .... Then you start smoking for the first time. After 10 seconds
the nicotine reaches the brain. Nicotine (a supplement to acetylcholine)
works as a transmitter (acetylcholine) on _one_ type of receptors. The
stimulation of the receptor becomes far greater than would normally be the
case for acetylcholine itself.

Conclusion Conclusion: With nicotine in the brain some receiver neurons will
excite a higher frequency. The reference value somewhere changes.

Why doesn't a perceptual signal simply change?

Of course the perceptual signal changes. When the reference value is
increased, we wish something more. If the reference value is reduces, we
wish something less. Our actions also changes.

I assume that all the facts you report are correct. But what effect do the

changes you report have on

all the rest of the brain?

My "facts" come from one of us: [From Bruce Abbott (2004.12.21.1800 EST)].
and http://bio.winona.edu/berg/ANIMTNS/actpot.htm and m.m.

Nicotine actually works on one type of receptors, the type that is reserved
for acetylcholine. It doesn't work on the whole Brain. But the actual
receiver neurons) are organized with other neurons. And they will or may
respond to the change.

Why is an increase in stimulation of a receptor treated as a
good effect or a bad effect by the rest of the brain?

I don't know. It depends how the receiver neuron (above) is organized. But I
can imagine (this is just a thought and I will appreciate your comment) that
a reduced stimulation will result in too low (or a near zero reference)
references here and there. And this will result in reorganization.

Why isn't it just treated as an increase?

It is. And you presented your thoughts about the consequence of increased
perceptual signal or reference signal in [From Bill Powers (2005.06.17.0651
MDT)].

The result of an increase of a signal will depend on what the
reference level for that signal is, and how the associated
control system is organized.

Yes of course. It is not easy to forecast where the hare jumps.

Why isn't the sudden change caused by the nicotine simply treated
as a disturbance, and corrected?

It is. All perceptual signals coming from one level is a disturbance in the
input function at a higher level. And it the perceptual signals are
corrected when you take a new cigarette. I liked this last answer. :slight_smile:

You could make a similar presentation for the effects of any
stimulus on any part of the nervous system. But there is nothing
in such a description to predict how the whole system will act
when such a change happens.

Yes I agree. But maybe I mis-understand you. Of course nobody could tell how
the system would act the first time a person (the first person who smoked
tobacco) lighted a cigarette.

To make a prediction, you have to know what all the control
systems are, what their parameters and reference settings
are, and how they relate to the other control systems in the
same brain. The chemistry of neurotransmitters
is never going to explain that.

I absolutely agree. I also agree with your last sentence. But I have not
used the word prediction. I will not use the word prediction when we talk
about nicotine, medicine or other chemicals. I think chemicals have an
effect, but the effect is dependent on parameters, reference settings and
organization.

I think many doctors also think so about chemicals/medicines, but they don't
know anything about PCT.

The reference value in a loop is changed.

Why do you say that? How do you know the nicotine isn't
affecting a perceptual signal, or the way a comparator works,
or an output signal?

There are many loops and many reference signals. Of course I could say that
nicotine affects a perceptual signal. But a copy of that perceptual signal
goes to a higher level, and to a higher level. There the effect is a changed
error and a changed output signal. And an output signal at one level may
become a reference signal at a lower level. Am I wrong when I think a change
in a perceptual signal also may become a change in a reference signal?
Therefore I just wrote "The reference value in a loop is changed".

How do you know it isn't changing the gain of the system without
any direct effect on signal amplitude?

I don't know. But I thought the gain is a result of different
physical/biological circumstances. (I hope to mail another mail about gain
in my economic model).
I am neither sure what to say when you talk about signal amplitudes. Of
course I have seen the scope of a heart machine watching a dying person. But
I can't explain the amplitude if it is correct that: "An action potential is
always the same size, typically some 90 millivolts. ( [From Bruce Abbott
(2004.12.21.1800 EST)]. Because an action potential cannot be made bigger,
the neuron will instead generate more and more action potentials. The
frequency will increase."

I have thoughts about the amplitude on the scope of a heart machine. But the
amplitude there isn't comparable with what happen in a loop. I think the
machine plug is in contact with many neurons and the more neurons it is in
contact with, the higher amplitude. When a person is dying all the neurons
don't stop functioning at the same millisecond.
Because an action potential cannot be made bigger, the neuron will instead
generate more and more action potentials. The frequency will increase."

You see the problem. The chemical details tell us nothing about what
the brain is DOING, or how its parts are connected to each other.

Yes I see problems, and I agree that chemical details tell us nothing about
_what_ the brain is doing. But I think chemical details tell us something
about _how_ the brain is doing in the organization it has.

Of course chemical details doesn't tell us how parts are connected, but I
think chemical details tell us how the connected parts work.

I liked this answer. I feel more sure myself. What can you do with that?

bjorn

37

[From Richard Kennaway (2005.06.23.1642 BST)]

[From Bill Powers (2005.06.22.1252 MDT)]
I think it's more than that.If you put the viewpoint close to the robot with the pendulum and put a brief impulse into the pendulum, you can see a lot of damped oscillations that go on for some time afterward. This says that by applying timed disturbances of small amplitude you can make the oscillations build up to a considerable degree. A random disturbance will have some component at the critical frequency, and if the phase doesn't shift too fast the pendulum will amplify the oscillations at that frequency. I think that's where the loss of control comes from.

This is true. I need to do a proper linear analysis of the system, but playing around with some of the parameters suggests that better results can be obtained if you edit Archy.config, at the point where it says:

  <upper
      propConst=...
and
  <lower
      propConst=...

to set those two numbers to 0.3 and 0.5 respectively. You have to restart the program to make it reread the file. These are the parameters for the bob position and bob velocity controllers.

-- Richard

38

[From Bill Powers (2005.06.23.1517 MDT)]

Jason Gosnell
(2005.06.22.18:55CST)–

Reorganization
doesn’t follow a predictable path. If it did, it wouldn’t work so
well.<<

… there is something about
the seeming disorder of it that makes it go as needed…it like wobbles.
Or maybe that is the only way it can go as that is how things are
generally going…in seeming or apparent disorder…even though there may
be some organization to this.

Look up my E. coli model. E. coli can only change its direction of
swimming at random. It can’t steer. But by tumbling or not tumbling at
the right times, it can make its way up a gradient of nutrient very
efficiently.
I think of reorganization as what you do when nothing else works. It is
powerful because it is random and not rational. A pigeon does not think
rationally that whether it turns left and right in an alternating
sequence couldn’t possibly have any effect on its ability to get food.
Therefore it learns to walk in a figure-eight, because that is what B. F.
Skinner insists on seeing before he will give the pigeon any food. If the
pigeon were systematic and rational it would starve.
In Zen and her method this is
done by placing more attention in the body or sometimes they say the
inner energy field of the body. So, the heart and body or your
“felt-sense” needs to be included somehow. What does PCT say of
this?
Well, being basically an engineer I’m perfectly willing to accept
the idea of kinesthetic senses, but not the idea of inner energy fields.
I don’t think there is any such thing. Why do people who don’t know
anything about physics insist on borrowing words like energy and field
and giving them nonsense meanings? I’ve never liked that aspect of the
Eastern philosophy that we read about over here.
Is this like including all of
the different levels of control–getting in touch with the experiences
that are not caught in the story perhaps? Or, is this merely one way to
contact the observer–to have more “presence” as it were rather
than having most all of the energy in the head? Or, rather than us
having the energy pumped into thinking, the energy of facing the self is
available when turning the light of awareness inward–that is, energy is
not supplied to thought-desires-gaining-gaining things…but is supplied
more and more to a pure form of awareness itself?

Whatever it is, it’s not energy. That term is taken, it already
has a meaning, and its meaning has nothing to do with the subjects
you’re talking about. Define what you mean, and maybe we can find a
different word for it. Though I’m not convinced that you’re speaking of
anything but sensations from the body.

This whole issue of shining the
light inward doesn’t preclude the whole activity of imagination
though…at least I think not. Otherwise these fresh and empowered ideas
one gets wouldn’t materialize in the field of
awareness.

I don’t think they just “materialize.” I think they come from
random changes in synapses, most of which result in new ideas that are
worthless. You have to decide whether the result of a reorganization made
things better or worse; if it made them worse, you reorganize again right
away, discarding the previous result. That’s what I call the E. coli
principle.

This is an interesting
issue as it seems to me that there are two types of imagination…1. Is
basically your programming as it stands now–your stories of self/other
etc., 2. Is an imagination that comes out of the observer’s field. This
is hard to describe, but there is a kind of imagining that seems
empowered by the observer–it may use previous learning–language/ideas,
etc, but it is not exactly limited or bound-up by the previous learning.
There is something about having more of the observer present that changes
all of this whether it is thinking, imagining or feeling…or even
action. I think so anyway at this time. I don’t know–perhaps it is
another phenomenon.

I’m very cautious about going beyond the experiences and observations
into theories about them. The Eastern philosophers who noticed these
phenomena long ago didn’t have any scientific tradition. Their theories
were, it has always seemed to me, undisciplined flights of fantasy, with
no hint of asking “Now, how can I show whether this theory is right
or wrong?” If it sounded good, they went with it. The result isn’t
very impressive. I’m happy to acknowledge their ability to observe and
report experiences, but not with their explanations.

Dogen then goes onto say after
taking the backward step…“…then your original face will be
manifest.” This I think is referring simply to the observer which
includes all of experience and is free from it at once. This observer
presence I think is the original face and it includes the whole of
experience.

Isn’t it a lot clearer to talk about an “observer” than about
an “original face?” Is he talking about an original cheek, eye,
nose, mouth, moustache? I wouldn’t think so – but then why doesn’t he
explain exactly what he does mean, instead of using these metaphors? Why
leave it up to others to guess if you really have an idea? Can you find
any place in Dogen’s writings where he says, “Here is a phenomenon
that I have noticed, but I don’t have any explanation for it”? I
don’t think metaphors give us explanations or understanding. They just
produce what I call “understandingness” – the impression of
having understood something without actually having done so.

My understanding now is that
“it” (original face–religious feeling-experience–etc.) is not
apart from everyday experience as people might imagine. But, the observer
is not completely limited by his background of learning…so it is the
original face–not contained or stained by experience. If so, then this
move allows for some freedom in the midst of difficult or even
pleasurable experiences.

I don’t think the observer learns. The brain learns and knows things. I
do agree that the observer is part of everyday experience.

I’m not really interested in the “religious” trappings of
Eastern philosopy at all. I don’t think that’s how we get to
understanding. As far as I’m concerned, the observer is a phenomenon to
be found in everyone’s experience, and it doesn’t get “stained”
by anything. It simply works the way it works. Some people’s brain become
better organized than others. It seems possible that people can organize
their brains to take the observer into account, and that everything works
better when they do.

I’m sure I sound very stogey, thick-headed, and unimaginative. I wouldn’t
be too unhappy with that, considering the alternatives I’ve
seen.

Best,

Bill P.

39

[From Bill Powers (2005.06.23.1559 MDT)]

Bjorn Simonsen (2005.06.23,11:00 EST) --

>What is there in the explanation up to this point that makes the reference
>value change?

Dendrites act as receiving stations for incoming signals from other neurons.
If the signals are strong and/or sustained, they will be conducted along
the branch of the dendrite down to the cell body.

That doesn't answer my question. Why are you assuming that the effect of the nicotine is on comparator neurons rather than on neurons that are part of perceptual functions or output functions?

Conclusion: The frequency in the addresser neuron releases an amount
acetylcholine in the synapse. ...And Now. .... Then you start smoking for the first time. After 10 seconds the nicotine reaches the brain. Nicotine (a supplement to acetylcholine) works as a transmitter (acetylcholine) on _one_ type of receptors. The stimulation of the receptor becomes far greater than would normally be the case for acetylcholine itself.

Conclusion: With nicotine in the brain some receiver neurons will
excite a higher frequency. The reference value somewhere changes.

I don't see that simply exciting a higher frequency in some neuron means that a reference value has changed. All I can see is that the proportionality constant between the input to some neuron and the output from it has changed. That is not a change in a reference signal. It's just a change in a proportionality constant.

>Why doesn't a perceptual signal simply change?

Of course the perceptual signal changes. When the reference value is
increased, we wish something more. If the reference value is reduces, we
wish something less. Our actions also changes.

Let me be clearer. Why isn't the effect of the nicotine that you speak of simply an effect on the neurons in a perceptual input function, so that for the same input signals they produce larger output signals? Are you proposing that nicotine selectively affects only those neurons that are found in comparator functions?

Nicotine actually works on one type of receptors, the type that is reserved
for acetylcholine. It doesn't work on the whole Brain. But the actual
receiver neurons) are organized with other neurons. And they will or may
respond to the change.

So are you saying that comparators are specialized to use acetylcholine as their neurotransmitters?

>Why isn't it just treated as an increase?

It is. And you presented your thoughts about the consequence of increased
perceptual signal or reference signal in [From Bill Powers (2005.06.17.0651
MDT)].

Sorry, I'm not following your thinking at all. Maybe if you could answer some of the questions I have here I might understand better.

Best,

Bill P.

40

[From Bill Powers (2005.06.23.1612 MDT)]

Richard Kennaway (2005.06.23.1642 BST) --

... set those two numbers to 0.3 and 0.5 respectively. You have to restart the program to make it reread the file. These are the parameters for the bob position and bob velocity controllers.

Yes, much better. Now the recovery is quick and there are no continuing oscillations.

I see that you're having problems of a sort I knew we would run into but have managed to avoid dealing with. It's the problem of what happens when errors get very large. In your simulation the robots not only go crazy, flailing around and flipping over, but they start violating physics, floating up into the air and so on.

It looks as if part of the problem arises when feet lose or make contact with the ground. Such discontinuities produce large instantaneous changes in forces, which create stability problems. Since the program evidently has collision detection in it, can the sensing of foot pressure be introduced?

I don't know if you recall, but some time ago I proposed adding a touch sensor to the lowest loop, so that a negative feedback signal would be produced at the spinal comparator proportional to foot pressure. To the position control system, contact would seem like having moved the foot to a new position, so in fact the lower loop would automatically switch from position control to force control when contact was made. This switching would not require any literal switching (as you have been hinting about using). I suggest this because in the real control systems of the body -- both human and cockroach -- sensors in the skin (or hairs) do in fact have very short synaptic connections to the same spinal motor neurons that are used for position and force control, and the feedback is negative. It's possible that if you simply applied a sine and cosine wave to the appropriate position control reference inputs, the legs would perform a semicircle movement for the upper half of the sine wave, and a force during the lower half when contact with the ground occurs. Then losing contact with the ground would not cause huge error signals. Maybe.

Best,

Bill P.