[From: Bruce Nevin (Mon 930412 09:43:53)]
"Trail-blazing ants prefer the straight and narrow", Stephen Day,
_New Scientist_ 27 March 1993: 16.
Black garden ants collectively determine a shorter route to food
located by one of them as a byproduct of each controlling a
perception of making fewest (sharp) turns en route. Sounds like
an excellent phenomenon to model a la "gather" demo. Read all
about it!
···
-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-
Same issue of _NS_, p. 40: Karl Sims set up a computer art
exhibit at the Pompidou Centre (wherever that is, the exhibit
probably will travel to Australia and US). Uses a parallel
Thinking Machines machine and something like GA, images on
multiple screens. The images that attract the most people are
allowed to continue evolving.
-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-
( [Avery Andrews 930410.0814] ) --
Something that tends to go wrong in PCT discussions, it seems to me,
[...] I think it would
be helpful if there was a preliminary phase of questioning, to figure
out whether there is actually any substantive disagreement, before
the argument starts, and more effort to figure out what people are
actually trying to say, once it has.
It would be helpful if, when someone raises a flag about
ambiguity of terms, all the participants had some kind of
"Uh-oh!" perception and thereafter specified which meaning they
intended of the identified ambiguous word.
In this case, it would have been most helpful if the flag that
several people raised--the observation that there are two very
different words that happen to have the same spelling and
pronunciation "information"--had resulted in people clearly
labelling which one they meant each time they used it. A lot of
the diputation took the form of "You said XYZ, and that's absurd
because PQX!" "That's right, I said xYZ, because ABx, and what
the hell does PQ have to do with x?" "You keep bringing in AB,
and so-and-so showed that ABYX ..." and so on and on. As Mary
might say, yuk.
Potentially Inflammatory Opinion: It was hard for participants
to practice elementary semantic hygeine with respect to
"information", I think, because equivocation over the two senses
has become institutionalized and has helped to make information
theory a career builder because those who control the purse
strings don't understand it or Entropy but as everyone knows this
is the Information Age and everybody wants to be Informed because
Information is Power, etc. Yuk again. Go read Shannon's article
on the information theory bandwagon.
( Ken Hacker [930410] ) --
( [Mary Powers 9304.11] ) -
Possibly off the wall department:
Communication in the ordinary sense is made possible by having
some reference perceptions pre-set in sender and receiver. (For
language: the contrasts between utterances, the features of sound
that symbolize or locate the contrasts, the shapes of words in
terms of these elements, the variation of word-shapes depending
on what other words are said, some conventionalized aspects of
word dependencies and word order.) IT-sensible: no ensemble, no
measurable reduction in uncertainty, without discrete, pre-set
elements.
On the face of it, this would seem to require that the sender be
monitoring her behavioral outputs (e.g. pronunciations of words)
for conformity to social norms. This is how I put it some time
ago. However, the private, CS-centric reference perceptions of
one's own speech are not identical to the perceptions of others'
speech. One can speak without normal auditory feedback of one's
own voice (though experiment is needed here). Kinesthetic
perceptions probably play a role in production at least of
consonants, and of course they are not a part of our perception
of the speech of others (except maybe in imagination). So
learning to conform to norms involved learning to control
perceptions of one's speech such that adults and other children
assented (in various ways) that one's behavioral outputs were in
conformity to the norms. In other words, they let one know that
in their perceptions (different from the perceptions one was
controlling) one was perceived as "a person like us".
If the perceptions that we control in order to be "normal" do
come to consciousness, it is not in terms of behavioral outputs
that might or might not conform to a social norm, but to how it
feels to be "me" expressing "myself". (So the language learner
could not learn a proper pronunciation of French because of
discomfort holding his face in that "prissy" way.) Social
correction in terms of behavioral outputs are accordingly
difficult to connect with how it feels to produce them, and
attempts to conform to them feel artificial until appropriate
CS-centric reference perceptions are established, control of
which has the effect of conformity to the observer's reference
perceptions. But such new reference perceptions might conflict
with a perception of "being" or "expressing" "myself".
Especially if one were an adolescent.
I said "communication in the ordinary sense". But the social
processes by which infants learn "normal" (shared) settings for
some essential reference perceptions obviously involve
communication, with some bootstrapping. Perhaps the
uniformitarian hypothesis (in a form something like "you're like
me, and your perceptions are something like mine") is innate?
( [Avery Andrews 930412.1030] ) --
Bateson suggested that the elementary unit of information in the
semantic sense is a difference that is transformed into another
difference ("a difference that makes a difference") around a
feedback loop. This formulation presumes discrete differences,
so we might do better to say "changes that make changes" since we
are talking about a continuous analog representation.
I believe that use of the term "percept" might constitute
sneaking "information" in by the back door. A percept is "an
impression of an object obtained by the use of the senses"
(Webster). The perceptual signal p might constitute a percept
for an outside observer or possibly for a higher-level ECS, or
rather for its input function. The information (the percept) is
created by (or, perhaps better, is located in) the structure of
the input function, which structure as Bill pointed out is a
product of evolution and reorganization. The perceptual
signal--either one of the lower-level signals p coming into the
input function or the new signal p going into the comparator from
the input function at the higher level--is not a percept, it is
just a signal.
This might be why there is no percept at the intensity level.
There's no information there, only a continuously varying value.
You get discrete information at the sensation level, where the
boundaries of a discrete sensation are marked by a threshhold of
intensity. Or am I missing something here? This approach might
confuse perception with attention.
Bruce
bn@bbn.com