Prediction (was Learning)

CSGnet Archive CSG2004
41

[From Bill Powers (2004.12.29.1152 MST)]

Bruce Gregory (2004.1229.1323)--

If you think about it, I believe that you will come to the conclusion that
each of steps I described is a necessary one. The more you practice, the
less conscious the steps are likely to be. But unless you recognize a
problem as one you have encountered before (even if it involves a
perfectly routine action) how would you even begin? After all, a door knob
doesn't announce itself by saying, "Twist me and push to open."

What I realize is that you are substituting logic for observation. To
declare that the steps become unconscious is to insure against ever having
to demonstrate that they take place: what is unconscious is by definition
unobservable.

Another way to say this (without invoking unconscious reasoning) is that
once you have established the correct steps of a procedure, you can simply
emit them in sequence without having to work them out again. Of course you
would want a kind of organization that can compare the sequence of steps as
actually executed against the sequence you worked out and are now
remembering, and that can vary your actions if disturbances tend to cause
deviations so the proper sequence will occur anyway. That is what I call a
"sequence control system."

Apropos of that, I get the impression that you think I offered the levels
of control as inventions to flesh out HPCT. If so, that would be precisely
backward. I found from observation, slowly and over quite a few years, that
I and other people seem to perceive the world in certain categories, and
that these categories seem to be universal (Prof. Zhang Hua Xia, who
translated B:CP into Chinese, assured me when I asked that the levels I
proposed fit very naturally into Chinese ways of thinking and perceiving).

The incorporation of these levels into a control-system model of behavior
then required proposing control systems with the properties necessary to
account for the observations, and that required working out how all these
different kinds of control systems could be coordinated and work together.
That led, before very many classes of perception had been observed, to the
idea that higher systems adjusted the reference levels for lower ones,
accounting for the common observation that we seek some goals as means of
achieving others, in a multiordinary way. You can see the basic building
blocks and their organization taking shape in the first two papers on
"feedback theory" by Powers, Clark, and MacFarland in 1960, which was seven
years after the start of this work.

Best.

Bill P.

42

[From Rick Marken (2004.12.29.1130)]

Dan Mayer (2004.12.29.1153) --

Rick Marken (2004.12.29.0845)]

... Perhaps you could go over how they derived this conclusion
[about prediction] from evolution?

Llinas ... believes control was and is an evolutionary answer for
our need to navigate in an unfamiliar environment. But control
by itself is not sufficient in his view.

I thought he was going to derive this from the fact or theory of evolution.
Now he's deriving it from his "view".

What 'consciousness' does, is to provide our control processes, with the
necessary front and back ends to help the control process to be as efficient a
process as it can be.

Again, I'm not seeing how this comes from evolutionary fact or theory. And
in what sense does consciousness make the control process more efficient?
Does Llinas provide any calculations of control efficiency with and without
consciousness?

Our 'predictive' capacity is, at least according to Llinas, evolutions way of
'restricting', or 'limiting' the variability of input,

But that's precisely what control does: it restricts inputs to those that
match the reference for input and reduces (enormously) the variability of
that input that would be created by disturbances (See my "Nature of
Behavior" paper in _Mind Readings_). Control does all this without any
prediction at all. So why does evolution need prediction to limit the
variability of input when people are already doing this themselves with
control? And who is this evolution fellow anyway?

and our ability to
'think' and make 'choices', enhances our ability to correct for error by
providing high degree's of freedom in behavioral selection.

My experience with building control systems suggests that they have all the
df they need to correct error. Does Llinas provide some models demonstrating
that there is an improvement in control that results from the ability to
think and make choices?

Added to this Llinas believes that feelings/emotions are pre-motor initiators
of control.

This guy has a lot of beliefs. Does he ever implement these as models and
test them?

I happen to like Llinas story.

I prefer Jane Austin's, myself.

But we all must keep in mind, that it _is_ just a story. Just like
the work of Bill Powers.

No. Llinas' story may be a story but Bill Powers' "story" is a testable
working model. If you don't know the difference between stories and models
then this list will be of little value to you.

I guess I was thinking of "fundamental" in the sense of the basic
organizing principle. I think there are basically two fundamental
organizing principles that have been proposed as the basis of
life:

Hold on a second, when did the discussion turn to the evolutionary basis of
_life_?

I am _specifically_ talking about the evolution of humans and our ability to
navigate in a changing environment. What are you talking about?

PCT is a model of living systems. So the discussion on this list is about
the purposeful behavior of all living systems and about the evolution of
these systems.

I would just like to know why [Llinas] believes this. Unless someone
can show me what he means in terms of testable, working models, I'm
really not interested.

Suit yourself, it's your story.

Since Llinas seems to have no testable model on which he bases his ideas
about the importance of prediction in human activities I think it's more
appropriate to move the discussion of Llinas to a modern fiction list.

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

43

From Dan Mayer (2004.12.29.1416)

In a message dated 12/29/2004 2:15:03 P.M. Eastern Standard Time, powers_w@FRONTIER.NET writes:

once you have established the correct steps of a procedure, you can simply
emit them in sequence without having to work them out again. Of course you
would want a kind of organization that can compare the sequence of steps as
actually executed against the sequence you worked out and are now
remembering, and that can vary your actions if disturbances tend to cause
deviations so the proper sequence will occur anyway. That is what I call a
“sequence control system.”
Is a ‘sequence control system’ conscious one? I don’t think so. Do you think about how you walk? How do you know that a ‘sequence control system’ exists? Have you identified this through experimentation or observation?

Why do you think your ideas are superior to someone else’s? Have you read Llinas? Have you read any neuroscience work that you can cite for your opinion besides some work done in the 1940’s -1960’s?

I found from observation, slowly and over quite a few years, that
I and other people seem to perceive the world in certain categories, and
that these categories seem to be universal (Prof. Zhang Hua Xia, who
translated B:CP into Chinese, assured me when I asked that the levels I
proposed fit very naturally into Chinese ways of thinking and perceiving).

Excuse me, but at the beginning of this post you stated:

[From Bill Powers (2004.12.29.1152 MST)]

What I realize is that you are substituting logic for observation. To
declare that the steps become unconscious is to insure against ever having
to demonstrate that they take place: what is unconscious is by definition
unobservable. Another way to say this (without invoking unconscious reasoning) is that…

How did you ever ‘observe’ the unobservable yourself?

I guess a billion Chinese can’t be mistaken, but I don’t quite understand your admonishment of Bruce for talking about ‘unconscious’ processes, yet your hierarchy’s first 6 or so levels are all ‘unconscious’ ones. How did you come to them? How do you know they exist? Because Prof. Xia has spoken for 1 billion Chinese, and a few others concur?

I guess you and about 100 people in the world perceive it this way.

I also believe this is less then the number of people who believe the world is really flat

The incorporation of these levels into a control-system model of behavior
then required proposing control systems with the properties necessary to
account for the observations, and that required working out how all these
different kinds of control systems could be coordinated and work together.
That led, before very many classes of perception had been observed, to the
idea that higher systems adjusted the reference levels for lower ones,
accounting for the common observation that we seek some goals as means of
achieving others, in a multiordinary way. You can see the basic building
blocks and their organization taking shape in the first two papers on
“feedback theory” by Powers, Clark, and MacFarland in 1960, which was seven
years after the start of this work.

I think you are mistaken in these beliefs, but I do wish you well and hope these story’s serve you well. I’m afraid there of little use to me.

DM

44

[From Bruce Gregory (2004.1229.1455)]

Bill Powers (2004.12.29.1152 MST)

Bruce Gregory (2004.1229.1323)--

If you think about it, I believe that you will come to the conclusion
that
each of steps I described is a necessary one. The more you practice,
the
less conscious the steps are likely to be. But unless you recognize a
problem as one you have encountered before (even if it involves a
perfectly routine action) how would you even begin? After all, a door
knob
doesn't announce itself by saying, "Twist me and push to open."

What I realize is that you are substituting logic for observation. To
declare that the steps become unconscious is to insure against ever
having
to demonstrate that they take place: what is unconscious is by
definition
unobservable.

Bill, I am saying nothing any different than what you said to Fred.
Problems are solved at the lowest level of a hierarchy. Once you have
practiced something, you do not need to process it at higher levels.
You are not conscious of the motions you make while driving, but they
are not unobservable. You can pay attention to them if you wish. What's
going on here?

Another way to say this (without invoking unconscious reasoning) is
that
once you have established the correct steps of a procedure, you can
simply
emit them in sequence without having to work them out again.

Emit? I act. Is that what you meant to say?

Of course you
would want a kind of organization that can compare the sequence of
steps as
actually executed against the sequence you worked out and are now
remembering, and that can vary your actions if disturbances tend to
cause
deviations so the proper sequence will occur anyway. That is what I
call a
"sequence control system."

I am less interested in the successful execution of a program than I am
in the success of the program executed. You rarely want to carry out a
whole sequence if the first step presents you unexpected problems. A
sequence is a proposed solution to a problem, in my view, not an end in
itself.

Apropos of that, I get the impression that you think I offered the
levels
of control as inventions to flesh out HPCT. If so, that would be
precisely
backward. I found from observation, slowly and over quite a few years,
that
I and other people seem to perceive the world in certain categories,
and
that these categories seem to be universal (Prof. Zhang Hua Xia, who
translated B:CP into Chinese, assured me when I asked that the levels I
proposed fit very naturally into Chinese ways of thinking and
perceiving).

The levels of control map very well onto cognitive stages of
abstraction. I never claimed otherwise. However, one could have
developed these levels without knowing anything about control. (The
notion that one controls principles is a bit of stretch in my view. I
certainly know of no evidence that any such thing happens.)

The incorporation of these levels into a control-system model of
behavior
then required proposing control systems with the properties necessary
to
account for the observations, and that required working out how all
these
different kinds of control systems could be coordinated and work
together.
That led, before very many classes of perception had been observed, to
the
idea that higher systems adjusted the reference levels for lower ones,
accounting for the common observation that we seek some goals as means
of
achieving others, in a multiordinary way.

What needs to be fleshed out, in my view, is how the higher level knows
what reference to assign to the lower level. Reorganization is one
answer, but most of the time we don't resort to random uncoordinated
responses to new problems. That is why I think we apply solutions that
have worked in the past to solve new problems. Only when these
solutions fail do we involve higher cognitive levels. You really ought
to read _On Intelligence_ . Hawkins proposes a hierarchical structure
that you would find familiar.

The enemy of truth is not error. The enemy of truth is certainty.

···

On Dec 29, 2004, at 2:12 PM, Bill Powers wrote:

45

[From Bruce Gregory (2004.1229.1511)]

Bill Powers (2004.12.29.1142 MST)

It seems to me that you're interpreting what I say in some strange
ways, or very selectively. Are you telling me that every time you sit
down to eat, you have to figure out what a knife and fork are for, and
plan how to reach out and pick them up, and how to move them in such a
way as to cut up food and convey it to -- let's see now, ear, pocket,
mouth -- ah yes, the mouth? Is there nothing you have learned to do so
well that you no longer have to make plans, choices, or decisions
about it? I don't mean to invade your privacy, but it seems to me that
to live as cognitively as you appear to do (or think you do) would be
a great burden and handicap.

When I reach out to pick up a fork I expect that it will not weigh five
pounds. I have expectations about how the work will appear and behave.
As long as those expectations are fulfilled I am not surprised. When
they are not fulfilled, I am surprised. This surprise can only be
identified with error if you want to argue, as Rick does, that we have
reference levels for all perceptions, not just controlled perceptions.
Is that what you intend?

For me, expectation is a form of prediction. The outfielder expects the
fly ball to behave in a certain way; he predicts what it will do (it
will not dart about erratically, for example). We can model how he has
learned to catch the ball without incorporating prediction _in the
performance model_, but that does not mean that he has no expectations.
In fact, he has expectations even when the ball is hit to another
outfielder. In which case, he is simply an observer.

I guess I just pay more attention to what I am doing while I eat than
you do. The cognitive burden is not all that large and I suspect one
tends to enjoy the meal more.

The enemy of truth is not error. The enemy of truth is certainty.

46

From Dan Mayer (2004.12.29.1444)

This has been fun, but time beckons and I really must be going, so this will be my last post on this thread and on this list.

In a message dated 12/29/2004 2:41:09 P.M. Eastern Standard Time, marken@MINDREADINGS.COM writes:

[From Rick Marken (2004.12.29.1130)]

I thought he was going to derive this from the fact or theory of evolution.
Now he’s deriving it from his “view”.
Why don’t you tell me what the ‘theory’ of evolution is precisely. When your done, send it to me privately and I’ll tell you whether your accurate or not.

What you don’t seem to grasp and understand, is that each of us has a ‘view’, or a story to tell. Nothing else. Just like it’s all perceptions. All perceptions are only story’s. That includes, you me, Llinas, and Powers.

Again, I’m not seeing how this comes from evolutionary fact or theory. And
in what sense does consciousness make the control process more efficient?
Does Llinas provide any calculations of control efficiency with and without
consciousness?
I think you need to understand what evolution actually is and does before we can continue with this conversation.

But that’s precisely what control does: it restricts inputs to those that
match the reference for input
If this were true there would be no error’s. It doesn’t ‘restrict’ error’s it corrects them. Are you sure you understand the control process?

and reduces (enormously) the variability of
that input that would be created by disturbances (See my “Nature of
Behavior” paper in Mind Readings).
How does your model ‘restrict’ input? Your reference level is not generated by the sensory receptors. Your reference levels come from ‘error’ signals from a higher level in the hierarchy and are then compared to your sensory input. Where and how is your sensory input ‘restricted’?

Please don’t bother answering me, these are just some questions you might want to consider.

Control does all this without any
prediction at all. So why does evolution need prediction to limit the
variability of input when people are already doing this themselves with
control? And who is this evolution fellow anyway?
Darwin.

My experience with building control systems suggests that they have all the
df they need to correct error.
Yes, of course they do, because you supply them with enough df’s. They are not an intrinsic property of any one control process.

Does Llinas provide some models demonstrating
that there is an improvement in control that results from the ability to
think and make choices?
If I told you yes, would it make a difference? I don’t think so judging from your responses to my posts.

I prefer Jane Austin’s, myself.
So do I, on occasion. Different story’s for different purposes.

No. Llinas’ story may be a story but Bill Powers’ “story” is a testable
working model. If you don’t know the difference between stories and models
then this list will be of little value to you.
A model of what? And a story of what? That is, what are you modeling, and what story is your model telling? I’m afraid your ‘models’ and 'story’s don’t match up very well.

PCT is a model of living systems. So the discussion on this list is about
the purposeful behavior of all living systems and about the evolution of
these systems.
OK, I’ll bite. What did PCT evolve from and where might it go?

Since Llinas seems to have no testable model on which he bases his ideas
about the importance of prediction in human activities I think it’s more
appropriate to move the discussion of Llinas to a modern fiction list.
Maybe, but he has nothing over on Powers’ tale of an untestable AND unobservable hierarchy (except by a billion Chinese) and someday a model may very well exist of something closer to Llinas’ story then Powers. But this of course is my story and view.

I do wish you well however in never,never land.

Cheers,

DM

47

[From Bruce Gregory (2004.1229.1512)]

Rick Marken (2004.12.29.1130)

Since Llinas seems to have no testable model on which he bases his
ideas
about the importance of prediction in human activities I think it's
more
appropriate to move the discussion of Llinas to a modern fiction list.

You remind me a lot of George W. Bush. No need to hear opinions that
might conflict with your own. I admit that it makes life a lot simpler.
But the consequences are not always exactly what you intended. Maybe
you should read just a teensy bit more. It would make your book reviews
more persuasive.

The enemy of truth is not error. The enemy of truth is certainty.

48

From Dan Mayer (2004.12.29.1352)

In a message dated 12/29/2004 1:50:14 P.M. Eastern Standard Time, powers_w@FRONTIER.NET writes:

[From Bill Powers (2004.12.29.1115 MST)]

Correct, that is what I mean. I will also add that we tend to solve problems at the lowest possible level of organization, meaning the least cognitive.
There is no such thing as ‘least cognitive’. Either it is, or is not in consciousness

An example has been discussed here several times: computing where a target will be some time in the future and computing a path that will get you there at the right time (as some people conceive of catching a baseball). Once you have experimented with this for a while, you realize that all you have to do is keep the target at a constant azimuth and elevation, if you can, and you will intercept it without any computations at all. So what you initially conceived of, and possibly attempted, as a cognitive task has been reduced to a far simpler tracking task. And that’s how you do it from then on.
Where is this ‘procedure’ stored in the PCT model and how is it ‘called’ into action? Please forgive any possible ignorance here.

But this principle also applies completely within the cognitive levels, whatever you think they are. Once you have found the solution to a problem, you don’t have to work out the method again: you already know the method. You become organized in the same general way a computer becomes organized when you program it. It make take some time to work out the program, but once it’s installed no further “working out” is required. You just run it.
Llinas calls these ‘Fixed Action Patterns’. But he specifies how and where they originate in the nervous system and brain. His ideas about how our nervous systems work are quite a bit different then yours though.

Yes. What we call decision-making or problem solving or planning and all that sort of stuff is really, I claim, the reorganization phase. You know what I mean by “really.” After reading one of my accounts of sequence control, Mary was delighted to showing me how this worked as she copied notes from our little refrigerator whiteboard onto the slip of paper to take shopping with her. She simply imagined herself in the store traveling along the aisles in her standard pattern, worked out long ago, and wrote down the items from the list inb the order that the pattern took her to different sections of the store. So she never had to double back or walk farther than necessary – and she got to the frozen food last so it would have the least thawing time. Once you work out the method, what’s left to decide or plan? No deviations from the sequence required, unless the store is out of something and you have to think up a substitute.
Yes, a decidedly different approach to behavior that simply acknowledges the existence of a mind and moves right on.

You have done a wonderful job of replacing the ‘environment’ and ‘Cognition’ individually as the causes of behavior and replaced it with control processes, that is, in effect, dependent on both the environment and cognition to operate smoothly

Now all you need to do is figure out why and how cognition takes place and you actually may be able to engineer a useful system.

DM

49

Re: Prediction (was Learning)
[Martin Taylor 2004.12.29.15.27]

[From Bill Powers (2004.12.29.1115
MST)]

Correct, that is what I mean. I will also add that we
tend to solve problems at the lowest possible level of organization,
meaning the least cognitive. An example has been discussed here
several times: computing where a target will be some time in the
future and computing a path that will get you there at the right time
(as some people conceive of catching a baseball). Once you have
experimented with this for a while, you realize that all you have to
do is keep the target at a constant azimuth and elevation, if you can,
and you will intercept it without any computations at all. So what you
initially conceived of, and possibly attempted, as a cognitive task
has been reduced to a far simpler tracking task. And that’s how you do
it from then on.

I think I’m agreeing with Bill
in expressing an apparent disagreement with him :slight_smile:

I played cricket for 40 years,
much of the time at a pretty high level, including internationally. I
was considered very good at catching a ball in both the infield and
the outfield. Until Rick presented his model, I was cognitively quite
unaware “that all you have to do is keep the target at a constant
azimuth and elevation.”

I think that in learning to
catch well, I had simply reorganized the visuo-motor controls at quite
low levels, not involving anything cognitive. A ball would come off
the bat, I would “see” where it was going to land and when,
and would be there awaiting its arrival if there was time, or diving
to make the catch if time was a bit short. I don’t think I actually
did “keep the target at constant azimuth an elevation” very
much, at least if there was time to sapre in getting where the ball
was going to be.

Mind you, it’s quite possible
that a movie of me catching a long fly ball might have shown a
different story – perhaps I did conform to the principle, but
perceived myself “in the future” as being in the right place
at the right time. I don’t think so, because now that I am just a
passive observer, I think I see outfielders moving quite quickly for
their first few steps and then making slow corrections as the ball
nears them. That could, however, be the way the algorithm works in
practice. Rick probably knows whether it does, or whether the
simulated outfielder’s pace is more nearly constant during the flight
of the ball.

Anyway, my point is that I
agree with Bill that you need no cognition, no so-called
“thought” when executing skillfully an outfield catch (or
any other really skilled behaviour, probably even skilled mental
behaviour, paradoxical as that may sound).

Martin

50

[From Bruce Gregory (2004.1229.1552)]

[Martin Taylor 2004.12.29.15.27]

[From Bill Powers (2004.12.29.1115 MST)]

Correct, that is what I mean. I will also add that we tend to solve problems at the lowest possible level of organization, meaning the least cognitive. An example has been discussed here several times: computing where a target will be some time in the future and computing a path that will get you there at the right time (as some people conceive of catching a baseball). Once you have experimented with this for a while, you realize that all you have to do is keep the target at a constant azimuth and elevation, if you can, and you will intercept it without any computations at all. So what you initially conceived of, and possibly attempted, as a cognitive task has been reduced to a far simpler tracking task. And that's how you do it from then on.

Anyway, my point is that I agree with Bill that you need no cognition, no so-called "thought" when executing skillfully an outfield catch (or any other really skilled behaviour, probably even skilled mental behaviour, paradoxical as that may sound).

While your conclusion is unremarkable, I would be willing to bet a substantial amount that _no_ major league baseball (as opposed to cricket) player has _ever_ realized that all you have to do is keep the target on a constant azimuth and elevation. That is a modeler's realization, not a player's realization. Only modeler's make computation's, players never do. However, players, unlike modelers, have expectations and act on them.

The enemy of truth is not error. The enemy of truth is certainty.

51

[Martin Taylor 2004.1`2.29.15.41]

From Dan Mayer (2004.12.29.1416) to Bill Powers

I don't quite understand your admonishment of Bruce for talking
about 'unconscious' processes, yet your hierarchy's first 6 or so
levels are all 'unconscious' ones.

Some serious misunderstandings are happening here, either on my part
or on the part of other correspondents.

As I understand HPCT, the perceptions at _EVERY_ level are
independent of consciousness, though the perceptions at _ANY_ level
can be made conscious or will automatically become conscious under
the right circumstances.

The HPCT model doesn't require consciousness to exist. Many different
suggestions have been put forward within the HPCT model, to account
for the fact that we obviously are conscious of some perceptions some
of the time, but not for most of the ones we control, most of the
time. Some of these suggestions might be testable if the precise HPCT
sub-structure used in a test could be determined.

But it's one thing to assert, as Bill quite reasonably does
(following at least a century and a half of Western tradition) that
perceptions come in different categories that have certain
entailments among the categories (you can't have a sequence inless
you have some entity type that can be ordered in space or time, for
example). It's another thing to describe a precise substructure of
perceptual relationships within a specific person at a specific
moment, and to describe the dynamics of
controlling/not-controlling/gain-changing/sensor-shifting the
perceptions involved in those relationships. And that's what you must
do if you are to make precise tests of constructs such as "attention"
and "consciousness."

To recap: HPCT makes no assertion obout some levels being conscious
and others unconscious. It makes assertions about entailment
relations among perceptions, some of which may be demonstrable, some
arguable, and others hard to test. It is an observationsl fact
attested by many people that perceptions claimed to be low in the
HPCT hierarchy can be made conscious only with some difficulty, and
possibly only with deliberate training. But that's true of some
controlled perceptions at very high levels, too (as an observational
fact mapped onto the canonical 11-level hierarchy). One is seldom
conscious of one's system-level or principles-level controlled
perceptions, though they probably are involved in setting multitudes
of lower-level reference vales (some for conscious, some for
unconscious perceptions).

Martin

52

Re: Prediction (was Learning)
[Martin Taylor 2004.12.29.16.44]

[From Bruce Gregory
(2004.1229.1552)]

[Martin Taylor
2004.12.29.15.27]

[From Bill Powers (2004.12.29.1115
MST)]

Correct, that is what I mean. I will also add that we
tend to solve problems at the lowest possible level of organization,
meaning the least cognitive. An example has been discussed here
several times: computing where a target will be some time in the
future and computing a path that will get you there at the right time
(as some people conceive of catching a baseball). Once you have
experimented with this for a while, you realize that all you have to
do is keep the target at a constant azimuth and elevation, if you can,
and you will intercept it without any computations at all. So what you
initially conceived of, and possibly attempted, as a cognitive task
has been reduced to a far simpler tracking task. And that’s how you do
it from then on.

Anyway, my point is that I agree with Bill that you
need no cognition, no so-called “thought” when executing
skillfully an outfield catch (or any other really skilled behaviour,
probably even skilled mental behaviour, paradoxical as that may
sound).

While your conclusion is unremarkable, I
would be willing to bet a substantial amount that no major league
baseball (as opposed to cricket) player has ever realized that all
you have to do is keep the target on a constant azimuth and elevation.
That is a modeler’s realization, not a player’s realization. Only
modeler’s make computation’s, players never do.

I’m not clear whether “as opposed to cricketers”
implies that I think that cricketers have this realization. I said
they (or at least I) didn’t, or intended to. My phrasing “you
need no cognition, no so-called “thought”” was intended
to convey that one needs no cognition or so-called thought to catch a
ball, even in cricket.

However, players, unlike modelers, have
expectations and act on them.

Now, if by this you mean “conscious expectations” I
would say that this is true some of the time, but not always. At least
in the close field, I have caught many balls without having been aware
of anything in the interval between the ball leaving the bat and
finding myself at full stretch on the ground with the ball in my hand.
Outfield catches take more time, and one may well on some occasions be
conscious of an expectation of where the ball is going to land – but
I think even then, it’s not always conscious, any more than is the
sound of your computer fan (of which I bet you just did become
conscious if your computer has a fan).

Martin

53

[From Bruce Gregory (2004.1229.1705)]

[Martin Taylor 2004.12.29.16.44]

Now, if by this you mean "conscious expectations" I would say that
this is true some of the time, but not always. At least in the close
field, I have caught many balls without having been aware of anything
in the interval between the ball leaving the bat and finding myself at
full stretch on the ground with the ball in my hand. Outfield catches
take more time, and one may well on some occasions be conscious of an
expectation of where the ball is going to land -- but I think even
then, it's not always conscious, any more than is the sound of your
computer fan (of which I bet you just did become conscious if your
computer has a fan).

Most of our expectations are not conscious. Normally we only become
aware of them when they are _not_ fulfilled. That is when we are
surprised. As far as I know, all surprise is conscious. When the ball
darts about instead of landing, I dare say you become aware of it.
Sometimes the absence of a sound is unexpected and you become aware of
that. This says to me that at some level outside of consciousness, you
were expecting the sound to continue.

The enemy of truth is not error. The enemy of truth is certainty.

54

[From Rick Marken (2004.12.29.1410)]

Bruce Gregory (2004.1229.1315)--

Rick Marken (2004.12.29.0945)

Great. You could make a real contribution (finally) to CSGNet if you
would describe these tests, the model on which they are based and
why the tests would show that prediction, planning and/or anticipation
are involved in behavior.

Have you ever read any book other than B:CP? Are you dyslexic by any
chance? Anyone who is interested in Hawkins' model and the tests he
suggests is free to read _On Intelligence_. If the book is too much to
ask of you, the tests are described in "Appendix: testable predictions"
pp. 237-245. Perhaps you can get someone to read it to you.

I think we could reduce the noise and acrimony level on CSGNet considerably
if we could focus on models and data. I'm happy to read the Hawkins book
myself but this won't benefit anyone else who is listening in on the
conversation. So while I will order Hawkins' book from the library and read
it when I get a chance, I think it would help this conversation enormously
if you would provide what I asked for above, which is just a description of
Hawkins tests, the model on which they are based and a brief explanation of
why the tests would show that prediction, planning and/or anticipation are
involved in behavior.

I don't think this should be difficult for you to do since you seem to
understand Hawkins' models and tests well enough to have concluded that they
show that prediction is a particularly important basis of human activity.
There must be a model and/or test that clearly shows what PCT misses and
Hawkins doesn't. Presenting it would, I believe, help everyone involved in
this discussion (of the role of prediction in human behavior) get a clearer
idea of what the problem with the PCT model is. Dealing with this topic only
via verbal jousting is not going to get us anywhere.

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

55

Hi Bruce
To me your description below sounds a lot like Case Based Reasoning

http://www.aiai.ed.ac.uk/links/cbr.html

···

I would say that first you need to recognize that the situation you now
face
resembles one you faced in the past. Then you have to recall the
solution
you used on that occasion. They you have to try that solution and see if
it
works. You anticipate (predict) that it will work, otherwise you would
not
proceed with it. Only if it fails do you try something else. Most likely
what you now try will be based on your recognition of a different
similarity
between another problem solved in the past and the one you are now
encountering.

The enemy of truth is not error. The enemy of truth is certainty.

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

56

From Dan Mayer (2004.12.29.1645)

In a message dated 12/29/2004 3:59:22 P.M. Eastern Standard Time, mmt-csg@ROGERS.COM writes:

[Martin Taylor 2004.1`2.29.15.41]

Some serious misunderstandings are happening here, either on my part
or on the part of other correspondents.

As I understand HPCT, the perceptions at EVERY level are
independent of consciousness, though the perceptions at ANY level
can be made conscious or will automatically become conscious under
the right circumstances.
OK, under what conditions does consciousness kick in and how?

The HPCT model doesn’t require consciousness to exist.
Yes, but humans living in a changing and variable environment do, even if it’s only to maintain (that is, clean and feed) an organism. So the model Llinas speaks of is a functioning human being.

What is HPCT addressing?

Many different
suggestions have been put forward within the HPCT model, to account
for the fact that we obviously are conscious of some perceptions some
of the time, but not for most of the ones we control, most of the
time. Some of these suggestions might be testable if the precise HPCT
sub-structure used in a test could be determined.
OK, this seems reasonable. Llinas is in fact offering speculations as to why this stuff is happening, and how this happens in our brain and nervous system. Llinas makes no claim of accuracy. His story is woven from 25 years of neuronal research. He admits at the very beginning that he might be over reaching. It’s a fun story. As B:CP is as well.

But it’s one thing to assert, as Bill quite reasonably does
(following at least a century and a half of Western tradition) that
perceptions come in different categories that have certain
entailments among the categories (you can’t have a sequence inless
you have some entity type that can be ordered in space or time, for
example). It’s another thing to describe a precise substructure of
perceptual relationships within a specific person at a specific
moment, and to describe the dynamics of
controlling/not-controlling/gain-changing/sensor-shifting the
perceptions involved in those relationships. And that’s what you must
do if you are to make precise tests of constructs such as “attention”
and “consciousness.”
I’m not sure about ‘Western tradition’. Llinas is one of the leading scholars in neuro science in the last half century. he is currently the Chair of Neuroscience at The Einstein medical School at N.Y.U and has been a prolific author and experimenter for over 40 years.

Powers might be a wonderful engineer, but he’s no neuroscientist. Especially one who has specialized in the functioning of the brain for over a quarter of a century. If I had, or have a question about ‘perceptions’, I think I would take them to Llinas, thank you.

To recap: HPCT makes no assertion obout some levels being conscious
and others unconscious. It makes assertions about entailment
relations among perceptions,
How can you possibly do this when you have no idea how perceptions are constructed or where they come from?

The entire basis for HPCT is the Hierarchy, which is an untested, unobservable piece of speculation that can never be tested because no one knows exactly what to look for, or where to look. No one has been able to validate the existence of a single level AND label it correctly.

The first level is a misnomer because the environment does not generate the signal. The signal is generated by the sensory receptor, and it changes as it moves up to the sensory cortex. Perceptions are not formed until these signals move to the associated area’s of the brain, where these signals are given meaning. This is neuroscience 101.

some of which may be demonstrable, some
arguable, and others hard to test.

I’m interested in which levels you categorize as demonstrable, which as arguable, and others hard to test and why?

It is an observationsl fact
attested by many people that perceptions claimed to be low in the
HPCT hierarchy can be made conscious only with some difficulty, and
possibly only with deliberate training.

You use the word ‘fact’ very loosely. There is NO fact that a hierarchy does indeed exist, so therefore any levels in that supposed hierarchy are equally as dubious. Second, I am unaware of any testing in the 35 years that this material has been available for testing. is there a reason why no tests have been attempted up to this point, besides the fact that no one knows where to look or what to look for?

So even if I say there is a hierarchy, and if I even give you the first level of intensity. Please show me the second level and how I might test for it, and if you can in fact tell me how, why has it not been done to date?

I applaud your faith in the model. I just don’t share it with the same intensity you seem to

But that’s true of some
controlled perceptions at very high levels, too (as an observational
fact mapped onto the canonical 11-level hierarchy). One is seldom
conscious of one’s system-level or principles-level controlled
perceptions, though they probably are involved in setting multitudes
of lower-level reference vales (some for conscious, some for
unconscious perceptions).

Pure conjecture. Fun, but conjecture none the less. Thanks for a fascinating story.

DM

57

[From Bruce Gregory (2004.1229.1738)]

Rick Marken (2004.12.29.1410)

I don't think this should be difficult for you to do since you seem to
understand Hawkins' models and tests well enough to have concluded
that they
show that prediction is a particularly important basis of human
activity.
There must be a model and/or test that clearly shows what PCT misses
and
Hawkins doesn't. Presenting it would, I believe, help everyone
involved in
this discussion (of the role of prediction in human behavior) get a
clearer
idea of what the problem with the PCT model is. Dealing with this
topic only
via verbal jousting is not going to get us anywhere.

Fair enough. Here follows a brief statement of Hawkins' predictions. I
leave it to you to describe the corresponding PCT-based predictions.

1. We should find cells in all areas of the cortex, including primary
sensory cortex, that show enhanced activity in anticipation of a
sensory event, as opposed to in reaction to a sensory event.

2. The more spatially specific a prediction can be, the closer to
primary sensory cortex we should find the cells that become active in
anticipation of an event.

3. Cells that exhibit enhanced activity in anticipation of sensory
input should be preferentially located in cortical layers 2,3, and 6
and the prediction should stop moving downward in the hierarchy in
levels 2 and 3.

4. One class of cells in layers 2 and 3 should preferentially receive
input from layer 6 cells in higher cortical regions.

5. A set of "name" cells described in prediction 4 should remain active
during learned sequences.

6. Another class of cells in layers 2 or 3 (different from the name
cells referred to in predictions 4 and 5) should be active in response
to an unanticipated input, but should be inactive in response to an
anticipated input.

7. Related to prediction 6, unanticipated events should propagate up
the hierarchy. The more novel the event the higher the unanticipated
input should flow. Completely novel events should reach the
hippocampus.

8. Sudden understanding should result in a precise cascading of
predictive activity that flows down the cortical hierarchy.

9. The memory-prediction framework requires that pyramidal neurons can
detect precise coincidences of synaptic input on thin dendrites.

10. Representations move down the hierarchy with training.

11. Invariant representations should be found in all cortical areas.

The model is too complex for me to describe in useful detail. Those who
are interested can find details in Chapter 6 "How the Cortex Works,"
(pp. 106-176) of _On Intelligence_.

The enemy of truth is not error. The enemy of truth is certainty.

58

[From Rick Marken (2004.12.29.1440)]

Dan Mayer (2004.12.29.1444)

This has been fun, but time beckons and I really must be going

Are you suffering from multiple personality disorder now, Marc?

Why don't you tell me what the 'theory' of evolution is precisely.

We have discussed two on the net. One is the familiar "natural selection"
model and the other is the "directed evolution" model described by Cairns et
al. Both are discussed in the "Random walk chemotaxis" paper in _Mind
Readings_, a paper that was oiginally published in _Behavioral
Neuroscience_. The third theory is "inheritance of acquired
characteristics", which Darwin's "Origin" put pretty much to rest.

I think you need to understand what evolution actually is and does before we
can continue with this conversation.

I'll work on that after I've figured out Economics.

[control] doesn't 'restrict' error's it _corrects_ them.
Are you sure you understand the control process?

I'll study that once I finish with Economics and Evolution.

How does your model 'restrict' input?

Call the variance in input, q, with disturbance present and without control
var (q). Call the variance in input with the same disturbance present but
with control var'(q). It turns out that var'(q) << var(q). It's in that
sense that control restricts input to a very narrow range.

Does Llinas provide some models demonstrating that there is an
improvement in control that results from the ability to think
and make choices?

If I told you yes, would it make a difference?

Of course. I've asked to see these models over and over again. I keep
getting referred to inaccessible publications. The one article by Llinas
that you sent was about neural circuits, as I recall. I don't recall seeing
any clear demonstration of an improvement in control with prediction.

No. Llinas' story may be a story but Bill Powers' "story" is a testable
working model.

A model of what?

Control.

PCT is a model of living systems. So the discussion on this list is about
the purposeful behavior of all living systems and about the evolution of
these systems.

OK, I'll bite. What did PCT evolve from and where might it go?

Er, it's the systems that evolved. PCT was created, though it took a few
more than 7 days to finish the whole megilla. In fact, it's a work in
progress.

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

59

[From Rick Marken (2004.12.29.1500)]

Bruce Gregory (2004.1229.1512)--

Rick Marken (2004.12.29.1130)

Since Llinas seems to have no testable model on which he bases his
ideas about the importance of prediction in human activities I
think it's more appropriate to move the discussion of Llinas to
a modern fiction list.

You remind me a lot of George W. Bush. No need to hear opinions that
might conflict with your own.

It's unexamined opinions that I don't care to hear, whether or not they
conflict with my own. Opinions, as you know, are like assholes; everybody's
got one. So my way of limiting the one's I listen to is to just listen to
those that are informed by scientific evidence. I don't think George is
interested in hearing scientific evidence regarding the merits of any
opinions, most especially his own.

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

60

[From Rick Marken (2004.12.29.1310)]

Bruce Gregory (2004.1229.1738)

Rick Marken (2004.12.29.1410)

I don't think this should be difficult for you to do since you seem to
understand Hawkins' models and tests well enough to have concluded
that they show that prediction is a particularly important basis of
human activity...

Fair enough. Here follows a brief statement of Hawkins' predictions. I
leave it to you to describe the corresponding PCT-based predictions.

1. We should find cells in all areas of the cortex, including primary
sensory cortex, that show enhanced activity in anticipation of a
sensory event, as opposed to in reaction to a sensory event.

...

11. Invariant representations should be found in all cortical areas.

All these are predictions about the brain activity one might observe in
certain (rather vaguely) defined circumstances (for example, does "in
anticipation of a sensory event" mean "temporally before the event occurs"
or "during the time when the organism is 'anticipating' the event"). They
are certainly interesting predictions, but I don't see their relevance to
the issue of the importance of prediction and anticipation in behavior. For
example, if we found that prediction 1 is true, what does this tell us about
the role of prediction or anticipation in behavior? That it occurs? We
already know that. That it improves control? We don't know what is being
controlled in this circumstance or whether the enhanced activity prior to a
sensory event is related to control.

I appreciate your posting these predictions. But I think what we also need
some description of the model that led to these predictions and/or an
explanation of why these predictions lead you to conclude that the PCT model
is missing something important regarding the role of prediction in behavior.

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.