From: Osher Doctorow, Ph.D. [log in to unmask], Sat. Sept. 15, 2000,
11:06PM
Coming from logic-based probability (LBP), which is a combination of logic
(or its set/event analogues) and probability/statistics which I introduced
in 1980, I find what looks like considerable confusion in computers used to
simulate life/evolutionary/complexity processes. The claim is frequently
made that equilibria are transcended by adjustment to novelty, continual
change, and similar things/processes. By trial and error, it is expected
that via cellular and other types of automata (some of them continuous), the
"glorious adaptability of human intelligence" will be attained or
approximated.
I find logical and other errors in both equilibria and trial-and-error
adaptation to novelty. The curious thing is that attacking equilibrium
theory is like beating a dead horse - equilibrium theory is not a very deep
theory anywhere except perhaps in some exceptional branch of science where
it will probably eventually be superceded by a better concept. We probably
owe the obsession with equilibrium theory to economists and some statistical
mechanics people. Economists and assassinated peacemakers comprise the only
two Nobel Prize categories which appear to be awarded more for the sake of
good will than for lasting knowledge. I will deal with statistical
mechanics elsewhere, since it is a super-complicated matter.
Trial-and-error adaptation to novelty really needs to stand by itself, and
this is where its logical and philosophical basis/ontology runs into
trouble. To illustrate what I am talking about, I have recently proposed
in my writings both on and off the internet that history needs to concern
itself with repeated errors (which are technically fixed or invariant
points) rather than with telling long stories of battles, leaders, eras,
etc. Imagine if history were actually to teach us how to prevent serious
errors from repeating! Yet history departments, in most of the USA at
least, do not do that as their main emphasis (or even as an important
emphasis). Somehow, the trial and error human being, plus the best that
historians can be in analytic and synthetic skills and problem solving, have
left us with the impression that history is a bunch of stories - and thus
politicians with their backgrounds in history and related political science
end up with not only endless legal case stories to learn but with historical
and political stories, and they keep repeating the same errors as past
politicians when they attain power.
Why do I bring this up only now, when mathematicians and physicists and
other scientists have had the untaken opportunity to argue my points up to
now? Basically because probability/statistics is only now beginning to
seriously attack problems of dependence of events, influence of one event on
another, "causation", etc. Evolutionary computing needs to keep up with
the latest results in our field, just as we need to keep up with
evolutionary computing. I think that both fields can help each other, but
it is dangerous for them to ignore each other's results. Just for your
information, there are two schools that are doing the
dependence/influence/causation work: my LBP school and the Bayesian
conditional probability (BCP) school which is much more popular but often
fails versus LBP in studies of rare events, llower dimensional/boundary
events, events which are contained in (subsets of) other events, and events
which highly influence other events. You can read abstracts of 46 of my
over 100 papers on LBP at the Institute for Logic of the University of
Vienna, http://www.logic.univie.ac.at, selecting ABSTRACTS and then BY
AUTHOR and then my name.
Osher Doctorow
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|