internal feedback; emitting behavior; Design review

[From Bill Powers (960302.0630 MST)]

Shannon Williams (960302.11:00) --

     Besides, you do use internal feedback paths because some of the
     elements in an HPCT hierarchy depend upon loops that never see the
     external world.

Only in the imagination mode. In the normal mode of operation all loops
at all levels are closed via the environment.

     Stop reading me as if I were saying 'it is possible simply to
     "emit" an organized behavior'.

This is hard to do when you follow that statement by this:

     When you talk about two people solving their conflicts by "going up
     a level", you are not talking about changing one's focus from
     emitting behavior to determining what behavior to emit.

I thought you didn't want to be read as saying that it is possible to
"emit" an organized behavior. "Emitting behavior" is _never_ a
consideration in HPCT, whether you're talking about levels or not. All
that a higher system can do is specify (in the form of a reference
signal) the perception that the lower system is to (try to) produce.

     Every bit of your evidence which supports HPCT, supports a control
     loop with neural nets. A control loop with neural nets has the
     added advantage of explaining learning.

I agree that neural nets will be useful as a way of modeling both
perception and output systems. So far, however, nobody knows how to work
them into a PCT model.

     I guess I will return to this list when I have a working product.

That's beginning to sound like a good strategy. I hope you can write it
in a language that I can compile and run (Pascal or C).

···

----------------------------------------------------------------------
Best,

Bill P.