[From Bill Powers (970610.1312 MDT)]
Peter Burke (970610) --
If we look at some of the levels however, we see the possibility of
levels within levels. Consider programs for example. Some programs, >for
example, the program a student follows to go to the store to buy >a
textbook, is at a lower level than a program a student follows to >take a
course. Thus, we have programs within progarms, and some >program are at
"higher levels" in the hierarchy than other programs.
My question is whether these are two different hierarchies >(conceptually)
or whether the highest program is still lower than >the level of
"principle" in the HPCT system.
You raise an interesting question that has also occurred to me. Looking at
a computer program, you can see that there is a main section, which calls
subroutines, which in turn can call more subroutines, with no limit to the
number of "levels" that could exist. But in another sense, in that all
these units are programs, they are all at the same level. They all require
the same set of underlying computing capabilities which would be useless,
for example, for judging the distance between two visual objects.
As I originally conceived the idea of levels, I thought of layers that
could be added onto an existing system which already had a certain number
of functional and self-contained layers. Each new layer (thinking
phylogenetically) would use the lower layers as its means of action, and
also receive (copies of) perceptions of the kinds already handled in the
lower layers, combining them to form new kinds of perceptions. Each new
layer would add some specific new type of computing and control capacity to
the system, although without the new layer the system would be able to
function perfectly well, doing whatever is possible to do up to that level.
This leads to the idea of "comparative PCT", in which we look for organisms
that lack all levels after a certain point. It's sort of like the idea of
"the man riding the horse riding the alligator," but with the levels being
defined very differently, by function rather than by analogy to specific
organisms or brain structures.
I first ran into the idea of levels within levels while considering what I
call the configuration level. A chair is a configuration, yet if you look
more closely it is made of legs, seat, perhaps arms, and a back, which are
also configurations. And each of these configurations can be seen as made
of smaller configurations, right down to the cells in the wood and the
structures within the cells. I finally decided, just to get on with more
interesting matters and not because I had proven anything, that these were
not really levels of configuration perceptions. What they show is the
result of applying the _same_ capacity to perceive in terms of
configurations, but on different size scales. The configuration level is
the first layer in which perception of configuration is abstracted from the
field of lower-level sensations. Organisms that lack this layer could not
perceive objects, although they could perceive and control sensations. What
this layer does is perceive and control configurations in all sizes,
orientations, or positions.
The configuration level is much stranger than it seems at first. There's
more than just "object recognition" going on. For example, if the whole
visual field is filled with little triangles, one sees "triangleness"
wherever one looks, but it is the _same_ sense of triangleness in every
instance -- even if the small triangles are arranged to make one big
triangle. But if one of the triangles is changed to a square or a circle, a
_different_ perception arises, that of squareness or circleness, and it
stands out from the rest. It's like a second perceptual signal that has
joined the previous one, in a different place in the world of experience --
not a spatial place, but a place in a space where different positions
distinguish different configurations regardless of where they are on the
retinas. This is what makes people think of perceptrons.
The same problem arises at most other levels. Are there events composed of
smaller events? Are there relationships among relationships? Categories of
categories? Sequences of sequences? My inclination is to say no, what we
are experiencing in each case is simply the same capacity to perceive
eventness, relationshipness, categoriness, or sequenceness, applied on
larger or smaller scales. There is only one layer, in each case,
specialized to create perceptions of each type out of inputs from lower
levels.
The main thing that is different between my concept of a hierarchy and
others is that in mine, there is no single organizing principle that
carries you from the bottom to the top level. J. G. Miller's idea of living
systems uses basically the principle of size: molecule to organelle to cell
to organ to organism to social system. This amounts to applying the same
level of perception, configurations, to the world on different size scales.
In my hierarchy, each new level introduces a new principle. Sensations have
nothing in common with configurations; there is nothing in configurations
that suggests the idea of transitions; transitions do not suggest a
space-time grouping into events, and so forth. Each new level, as it were,
changes the subject.
Also, my concept of hierarchy differs from certain others in that the
levels in mine are physically distinct from each other. In one concept of
hierarchy, you begin with a set of elements, and create a hierarchy just by
conceptually grouping them: little groups are circled to set them off, then
the next level consists of bigger circles containing sets of littler
circles, and so on to the circle that encloses all the others. No matter
what "level" you look at, it is made of the same physical elements. In my
hierarchy, different neurones make up the different levels, with
connections between levels consisting of neural pathways that carry
information from one place to another. When you look at the configuration
level, you find no signals that represent intensities or sensations, and in
a creature having this as the highest level, there are simply no signals
representing events, relationships, and so on, at all. There are no higher
layers that can compute these things.
Each level in the PCT hierarchy, therefore, has to be thought of as a
physically distinct layer, specialized to perceive and control the world in
one specific way, different from the way in which any other level perceives
and controls. No one level knows that there are any other levels, higher or
lower. Each level is a kind of _idiot savant_, being very good at
extracting particular aspects of the world, but knowing _only_ that world.
It's only the whole brain that knows all the levels -- the whole brain and
this strange phenomenon called awareness, which seems able to receive
information from anywhere in the hierarchy, although not the whole thing at
once.
To get back to your original comment:
Think of a program running in a computer. In order for it to run, the
computer has to have a specific kind of organization; the organization of a
radio or a power plant would not suffice. We know that in at leat some
computers there must be data storage, means for calculating logical
functions, means for retrieving and executing instructions. Whatever the
actual components of the computer, they are of the kind that is necessary
to run a program -- ANY program. The nature of the computer is such that it
can run simple programs, complex ones, programs with internal levels of
organization, or anything you can conceive of as a program. But there are
no levels of organization in the computer itself; the computer is simply
the medium in which programs can run, and its organization is quite
independent of WHAT program is running.
This is how I think of the program level of the brain. Its organization has
nothing to do with WHAT programs are running there. It is simply that level
that contains the components necessary, in the right organization, to allow
any kind of program at all to run. The program might concern verbal
language, sign language, mathematics, superstition, strategies of war or
business, boolean logic, symbolic control systgems, or anything
program-like, in single or multiple functional (but not physical) layers.
That's only a matter of what specific programs have been learned.
Studies of higher mental functions, it seems to me, focus primarily on the
particular programs that happen to be running in a brain, or a community of
brains. From the standpoint of PCT, the particular programs are of only
passing interest; all that is interesting about them is what they might
reveal about the basic functions at the program level that are common to
ALL cognitive or program-like activities. Studies of particular logical
ways of reasoning and creating language are only studying the activities in
a brain, not the basic functions of the brain; a parallel would be studying
the printed inputs and outputs of a computer program, thinking that you are
learning how the computer itself works.
So the "levels" we can see in the brain's program level are really just
levels in the programs that are running; they all exist at the same level,
the 9th level in the HPCT hierarchy, where the brain is organized to run
programs.
Best,
Bill P.