A revolutionary OS/Programming Idea
Lynn H. Maxson
Lynn H. Maxson" <lmaxson@pacbell.net
Wed Oct 8 23:50:02 2003
Li, Xiao Feng writes:
"Hi, Dear Lynn, what you described are really interesting, but I
believe the adaptiveness for an individual system is not
essential if we have an reproduction and selection mechanism
among a group, where the death of the non-adaptive
individual is even more vital and the adaptiveness is achieved
generation by generation in a species-level instead of
individual-level."
If you want me to agree to a general species survival principle
that there is safety in numbers, you have it. I don't believe I
have said anything to the contrary.
You have to appreciate what Ashby demonstrated with his
homeostat: that a goal-oriented system, basically that
associated with homeostasis in humans, could adapt to
changes in its environment without any form of external
intervention. No "deus ex machina", no finger of God, no
initial instruction set, no starting program.
Moreover the system, the homeostat, responded to external
changes as a whole in an entirely unpredictable,
non-predeterministic manner: state A did not result in a state
B. This flies in the face of control system theory or control
system synthesis. A connection involving feedback
dynamically, i.e. unpredictably, switched between positive and
negative.
The homeostat demonstrated adaptive behavior. If it could
not achieve homeostasis within some indeterminable interval
of time, it failed, i.e. died. In short it constantly kept
attempting to maintain homeostasis until its resources failed it.
Now Ashby was pillored for this work, because he eliminated
the need for God, the "deus ex machina". You would think
scientists would welcome such a demonstration.
Unfortunately it also rendered unnecessary the need to apply
control system synthesis to adaptive behavior in living
organisms or prevalent belief among those in this thread that
it was even possible to do so.
Now we haven't established the basis for adaptive behavior in
living organisms. We have this one demonstration of it
occurring outside the realm of control system synthesis. We
speculate, something far short of demonstrate, that at some
given future time when we can "realistically" emulate the
neurons of the brain we will be able to write the program.
If we cannot, then artificial intelligence will remain as such:
artificial, never real. That doesn't make it useless. Because
we may never "crossover" does not mean we should not
continue the quest for its improvement. That it's not the way
we humans actually adapt does not detract from the value to
us of its advancement.
Yes, in general there is safety in numbers.<g>