[Martin Taylor 960426 13:30]
Bill Powers (960425.0100 MDT) to Hans Blom
I think you have a mental picture in which all the levels of control
deal with the same type of perception (I've had this argument with
Martin Taylor, too).
I was not aware of that, since I have no such mental picture. If you
thought I did, then maybe that's one reason so many of our arguments
have ended rather than being resolved.
Perhaps you are thinking that the different levels of a multilayer
perceptron "perceive" the same kind of thing because the operation
performed in the perceptual function is the same at all levels. If
you think this, that might be one reason you made the quoted statement.
But it is not true. Different levels of an MLP have the same kind of
input function (weighted sum and squash, with possible input shift
register, differentiator or integrator at the different input sources).
But what the output corresponds to in the environment differs greatly
at the different levels.
For example (and not particularly trying to stick with the levels as
proposed in B:CP), consider the simple behavior of beating a drum in a
steady rhythm. What we perceive as one steady rhythm is actually derived
from a periodically changing set of lower-order perceptions, produced by
repetitive actions. We perceive a simple variable called "tempo," which
can become greater (faster) or less (slower): a scalar variable.
There is not much problem in producing this kind of change of "type" of
perception from a series of layers whose input functions are the same,
provided that the input variables have some aspect of time in them--such
as allowing the input to be a shift register, or include differentiators
or integrators, all of which are neurally very easy to do. The fact that
one level "perceives" intensities that an outside observer sees as changing
whereas another level "perceives" a tempo that an outside observer sees
as invariant does not mean that the functions delivering those perceptions
differ in character.
Maybe you are thinking of some other basis for old arguments. But not since
very long before ever hearing of PCT (if ever) have I conceived that "all
the levels of control deal with the same type of perception," unless
"the same type of perception" is taken to be the level of a variable on
a signal path. I do believe all "types of perception" are represented in
this "same" way, regardless of what, if anything, an external observer
would relate the signals to in the outer environment.
Martin