[From Bill Powers (930411.2330 MDT)]
Bob Clark (930411.0345 EDT) --
Thanks for the analysis, Bob -- on several levels, you might say.
I take your point about all this problem-solving taking place
without any intrinsic errors. In the terms in which we used to
think of intrinsic error, you're right (aside from any minor
periods of elevated heart rate, blood pressure, and adrenalin).
I no longer think of intrinsic error as limited to purely
physiological variables. For example, the presence of chronic
significant error signals in any control systems of the brain is
itself an indication of something amiss, and would fit the basic
definition of an intrinsic error. It's also possible that the
scope of the reorganizing system has evolved along with the
structure of the brain that permits us to develop higher levels
of control systems. So I don't object too much to your concept of
the DME, which apparently operates in terms of criteria
considerably more advanced than physiological states
("appropriateness" for example). Perhaps your DME is simply a
more evolved version of the primitive "negentropy system" with
which we began almost four decades ago.
I do have one argument with your DME, which is that it seems to
have many capabilities that I would rather see as aspects of the
learned hierarchy. In my development of ideas about levels in the
hierarchy, I tried to isolate types of perceptions that at least
in principle could be controlled by learned control systems.
Anything of that nature clearly doesn't belong in the system
responsible for shaping organization, because what is learned is
not present at first, yet the process of reorganization has to
work from the beginning. I see too much that is systematic and
algorithmic in your descriptions of the DME and what it does. If
those were stripped away and assigned to the learned hierarchy
instead, I think our concepts would come much closer together.
I hope you find these comments interesting and "useful"
(whatever that means!)
Indeed.
ยทยทยท
-----------------------------------------------------------
Avery Andrews (930412.1030) --
In the case of an ECS, however, there is clearly zero potential
to form a representation of D from P alone, tho it may be
possible to represent certain kinds of D from P plus additional
information of various kinds (e.g, the external force acting on
a steering wheel, from perceptions of muscle tension and wheel
motion).
I agree. When we consider the variables and signals as messages,
we look for the systematic relationships among them. It's then a
simple matter to determine when we have enough knowns to solve
for the remaining unknowns in a system of equations. D can't be
determined from P alone because that is too few variables for
solving the system of equations. The forms of the functions must
be known, or else enough must be known about the variables to
determine the forms of the functions. This is just basic systems
analysis.
---------------------------------------------------------------
Best to all,
Bill P.