[From Rick Marken (920816)]
I want to thank Bill Powers and Avery Andrews for both,in their own
ways, making this a very pleasant weekend for me (despite the fact
that the temperature is in the high 90s -- I've enjoyed, however,
thinking about the fact that it's winter in Australia; most refreshing).
Avery Andrews (920815) --
That was an EXCELLENT post.
You make several excellent points. First, you correctly point out that
the PCT use of the term "control" differs from its use in many other
contexts, where it is used as a synonym for "cause" or "determine". I
get so used to using "control" in the PCT sense that I forget that it
has these other connotations. Your point is very well taken (and put).
Since I'm currently listening to the third movement of a delightful
Mozart sonata, I'll even accept your view that you:
see PCT as a subtype of interactive AI
as long as we can agree that, for living systems at least, it is the
correct (or currently best) subtype.
I conjecture that the way to get the crowd people to navigate
would be to give them more sophisticated perceptual functions
Yes! Right now the crowd people "adjust" by basically reorganizing
(I think there is a bit of random jitter in those higher order
references); that's why they rarely get stuck in traps or keep
pressing against the same spot on the wall (as in your example).
If you want a crowd element to get around obstables more efficiently
(which means, control a higher order variable rather than reorganize)
then you have to figure out what they might control in order to
properly reset the lower order variables.
But remember, when you build intelligence into the model based on
you OWN understanding of the environment, you might be putting in
a capability that would not work in other types of environments
(and would have to be reorganized out -- that's probably why re-
organzation occurs in real systems). What I mean is; if you put
in a control system that let's your crowd person efficiently skirt
a wall perpendicular to the path of approach, the system may not
work for other kinds of wall (or walls with doors). This is
just off the top of my head -- but you get the gist?
But you're right; making a "cleverer" control system probably means
finding a higher order perceptual variable (like some measure of
rate of change of a variable?) that can be used to adjust the reference
for another perception so that the little crowd person can get
around the wall more efficiently. Good exercise. Of course, the
real first step is to see what perceptions efficient wall skirters
(like your dog) actually control when skirting walls. That would
take some real ingenuity -- testing for your "wall skirter's" controlled
variables. Once you know what the dog is controlling, then it's a
piece of cake (sort of) to build the model.
Best regards
Rick
···
**************************************************************
Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)