[From Bruce Gregory (2010.08.06.1914 EDT)]
[From Bill Powers (2010.08.06.1530
MDT)]
BP: OK, I’ll try. But first, could you summarize your idea of what
the present understanding of this series of events is? I’d like to see
what the PCT explanation will be compared with.
BG: I find that very strange. Why are you interested in my flawed
understanding? It involves Hebbian learning, which you have dismissed, if
I recall.
Did I say it was flawed? I just want to know what is being put up against
PCT. You seem to think current theory is adequate. So what does it say
about this particular
case?
BG: Current theory is clearly inadequate. PCT is the wave of the future. Why are you so obsessed with inadequate theory. I don’t get it.
BG: On the third trial I say “B”. Thereafter I say
“B” when I see “A”. I think I understand the
conventional and outmoded explanation (I remembered that the card with
“A” on one side has a “B” on the other
side.)
BP: What is the current explanation for this? For example, how does
current theory explain how remembering that the card with A on one side
has a B on the other side leads to saying “C”? You have
described what happens, but what is the current explanation for what
happens?
BG: Again, why do you care? Surely there are many wrong models. Why
are you interested in my wrong model? Help me out here.
You seem to think that because I am offering an alternative theory I will
not listen to a proposal concerning a different theory. I don’t offer my
own theories just to be different. I don’t think my theory offers all the
answers. Does Hebb’s? I assume you know how Hebb’s theory would apply to
the examples you’re using. How does it apply? Do I have to justify myself
before
asking?
BG: I am profoundly puzzled by your apparent obsession with theories you know and I know are based on a fundamentally flawed view of human behavior. Was is this all about, anyway?
BG: I have been told by a usually reliable source that there is no
mechanism in PCT that allows a perceptual signal to become a reference
signal.
BP: I’m afraid that your usually reliable source told you wrong, or
you misunderstood what the source told you. See B:CP chapter 15, Memory,
and particularly Fig. 15.3 and the pages leading up to it for the
proposed way in which perceptual signals become reference signals, by way
of memory. I don’t know what the correct mechanism is, because there has
been no research on this proposal.
BG: O.K. I’ll let my source know that he is misinformed.
Are you saying that instead of looking up the reference and seeing for
yourself? How about taking responsibility
yourself?
BG: I am sorry. I thought CSGnet was here to help people understand PCT. I was mistaken. I apologize.
BG: How is the reference signal “B” established? (I assume
that the controlled perception involves saying “B” when I see
“A”. Is this correct?)
BP: I would agree that the task is to produce the correct perceived
relationship between what you say and what is on the first side of the
card.
BG: Good. We have a starting point.
BG: Is the reference level set by a higher level controlled
perception?
BP: No. It is set by the output function of a higher-level control
system.
BG: Sorry, that’s what I meant. What is the reference level for this
higher-level control system?
BP: How am I supposed to know why this person wants to learn French, or
to go through the flashcard exercises you describe? Is there only one
possible reason which PCT ought to be able to describe?
BG: Again, I misunderstood the purpose of CSGnet. Sorry.
BG: Is the reference level the result of e-coli reorganization?
Yes, I would assume so. The reorganization would have applied to the
higher-level output function, or the weightings of the signals it
produces. It is possible that a systematic method of memorizing might
have been used, but that would be unusual except in a trained linguist or
someone who has learned to use mnemomic tricks.
BG: Can we test this assumption in any way? Does e-coli reorganization
imply that my responses are random?
BP: During trial-and-error learning of a skill, probably. In the present
case, which appears to involve only memorization, probably not unless the
person has never tried this before.
BG: I understand. E-coli reorganization does not apply and it is my job to figure out why. O.K.
BG: Do the responses included
German words for example? Or are they confined to French words? How does
the mechanism select the set from which it makes random selections? Can I
assume that one-trial learning is inconsistent with this
model?
BP: I think you’re confusing reorganization with memorizing. Chapter 14
of B:CP on learning starts with this:
“Several types of phenomena are commonly subsumed under the term
learning. I will call them *memory, problem-solving programs,*and reorganization.”
Your example is concerned with memory, not reorganization. Of course
coming to the point of deciding to learn French vocabulary may have been
the product of reorganization – there’s no way to generalize about that.
But the actual task you describe appears primarily to be one of
remembering the French word that goes with the English word. Perhaps
Hebbian learning applies to that kind of task.
BG: O.K. So Hebbian learning might apply to memorization. And possibly to other forms of learning. But not to tracking experiments. Have I got that, at least, right?
Reorganization, in its technical PCT meaning, is the method of last
resort when more systematic methods (themselves learned through
reorganization) fail. It is also the method of first resort before any
other methods have been learned and when memory alone is
insufficient.
BG: O.K. So conventional thinking is not entirely wrong. Or is it? Have I missed something, yet again
BP earlier: I suggest that you make a list of a few important
questions like these, and re-read B:CP to see if you can find any answers
to them. That might save you some time.
BG: So far I haven’t made much
progress. Can you point me to the appropriate pages in B:CP which answer
the questions I’ve raised.
BP: OK, I’ve been doing that. But why not actually read the whole book?
You ask so many questions that are answered in the book that it makes me
wonder if you ever actually read the whole thing, or gave it enough
detailed attention to understand it. I don’t mean to pry, but is there
some reason for not simply studying the book until you understand what’s
in it?
BG: I thought I had. But clearly I was mistaken.
People often read it two or three times – and sometimes jokingly
ask me when I put all that new material in that they missed the first
time. You seem to have missed a great deal that is quite plainly stated.
I hope I’m not mistaken about how plain it
is.
BG: I confess that I have only read it once. I will refrain from posting to CSGnet until I have read it at least twice more.
BP earlier: But please do describe what the conventional explanations
of these observed phenomena are. I’ll try to relate PCT to them and
describe where it’s different.
BG: I’m still puzzled. But O.K. Here is a Hebbian response.
When I see “A” and then “B” a connection is formed.
The connection is weak and therefore my response may not be “B”
but rather another word from the collection, say “C”. Each time
I see “A” followed by “B” the connection is
strengthened and incorrect responses such as “C” are inhibited.
I suspect most neuroscientists would see things something like
this.
The previous neuroscience explanation for this strengthening effect
(long-term potentiation) has proven incorrect.
BG: That’s a relief to know. Perhaps they should recall the Nobel Prize of Eric Kandel and the N.A.S. election of Dale Purves. I have never been impressed by the ability of the Nobel Committee or the National Academy of Sciences to recognize scientific contributions, anyway.
The strengthening effect
occurs only when the delay between the first stimulus and the second one
being associated with it is not more than 50 milliseconds. I have
proposed that this strengthening serves to build new synapses
specifically for aborized fibers from the same originating axon (all of
which reach the target cell at very nearly the same instant). In other
words, it facilitates the kind of mechanism I proposed in B:CP for
changing the weights of incoming signals by splitting the incoming signal
into multiple parallel paths each of which synapses with the dendrite
tree. See chapter 3, Premises (which did not include the neurological
corroboration now available).
BG: Needles to say, I am impressed by your ability to have anticipated all the neuroscience of the past fifty years. Quite an accomplishment I would say. I now appreciate that everything I want to know is contained in B:CP, had I the wit to comprehend it. No wonder Rick has such contempt for me. I am clearly trying to play in a league well beyond my capabilities.
I wonder if this has advanced us toward understanding at all.
BG: It has confirmed the view that I simply lack the mental ability to understand PCT. That is, of course, not your fault.
Bruce
···
On Fri, Aug 6, 2010 at 6:31 PM, Bill Powers powers_w@frontier.net wrote: