From Rick Marken (961130.2210)
David Goldstein ( 961130.1015)--
I think Rick did a great job with his web page.
Thank you. Thank you.
Rick, how hard was it to learn Java?
I don't know. I haven't learned it yet;-)
Actually, Java is said to be an easy language to learn. But the ones who
are saying it are mainly computer science types, it seems. I'm finding it
rather tough because it is object oriented (OO) and I have had zip
experience with OO; It is also network oriented, which makes some
things unusual; for example, Java won't store data locally -- for
security reasons -- so you have to build your own server to save data;
that's why my programs don't save data; I haven't learned how to do it yet.
So Java can be learned; but it's not a walk in the park (for me, anyway).
Bruce Abbott (961130.1355 EST)
Do you _know_ that all control-learning depends on random
reorganization, that there are no more efficient processes available
to rat or human?
No. But I'm not objecting to your model because you are proposing
a non-random reorganization system. I am objecting to it because the
system you are proposing cannot possibly learn to control. At least,
this is the way it seems to me based on your description of the model.
You are proposing a system that repeats the perception that occurred
prior to a reward. I can't believe that such a system can learn to control.
But maybe I'm missing something. Why not just program the model so we
can see how it works?
Bruce:
What empirical observations DO you see as offering difficulties for HPCT
as now constructed? Surely after 14 or so years you've run into _some_.
Me:
Are you asking "what empirical observations _have already been_ made that
offer difficulties for HPCT as now constructed?" If so, then obviously the
answer is "none".
Bruce:
Why do you suppose that is?
Because if you mean "what _future_ empirical observations DO you
see as offering difficulties for HPCT" then the only possible answer
I can think of is the one I already gave: any observation that is
inconsistent with a prediction of the model.
So a mechanism that uses the perceptual signals available to it to determine
which reference(s) to vary could not possibly learn to control, but a
mechanism that merely chooses them at random (i.e., e-coli reorganization)
can?
If your mechanism uses perceptual signals to determine which references
to _vary_ , then it probably can learn to control and I withdraw my
objections. You kept saying that your model selects a particular perceptual
reference to repeat -- the reference for the perception that occurred before
the reward. If what is learned is how to _vary_ a reference signal, then
what your system is learning is transfer functions -- and everything is
fine. Again, why not program it up so we can see how it works.
Me:
If you don't want your neighbor to kill you, then there are basically
three ways to deal with the problem; 1) control him 2) negotiate with
him or 3) defend against him.
Bruce:
There are only three ways to implement Option 3: (1) Make yourself
unavailable (hide, or lock yourself in a fortress) (2) get him before he
gets you, and (3) control him.
Controlling and defending are two different things. When someone
tries to hit you and you put up your hand to block it, you are simply
resisting a disturbance. You are not controlling the position of the hand
(the disturbance); you are controlling the perception of the relationship
between the hand and your face and the hand is a disturbance to that
relationship.
So if you want to live a normal life it comes down to two options:
control or negotiate...I'm all in favor of the latter, but before it can
happen, my neighbor has to be willing to listen. If he won't, then there
is nothing left to do but control .
There is always nothing to do but control. But you don't have to control
your neighbor to defend against him. If your neighbor wants to hurt you
and that's that, then you will defend against him; you will control your
perception of being hurt.
Controlling your neighbor means getting him to behave in a particular
way. If you want to make him _not want to hurt you_ or if you want to
make him _have the right values_ then you would be trying to control him.
When you defend, you just resist a disturbance to a controlled variable.
When you control a cursor, for example, you _defend_ the perception
of the cursor from the disturbance; but you don't try to control the
disturbance. You _would_ be controlling the disturbance if you tried to
make it be a constant rather than a sine wave, for example.
As I recall, negotiation was the option Nevil Chamberlin chose. When that
failed, he was forced into Option 1.
No. He was forced only to option 3 and even then he was "forced" only
because he was controlling for things like the survival of the nation.
Tracy Harms (1996;11,30.17)
"defend" in Rick's useage does not mean *solve*. It clearly
precludes control; you can't slip control back into it.
Excellent, Tracy. This is exactly what I meant (as noted above).
Best
Rick