Emergence of behavior through software
Lynn H. Maxson
lmaxson@pacbell.net
Mon, 25 Sep 2000 21:51:22 -0700 (PDT)
Billy wrote:
"The part where someone started believing that "universal machine"
has ANY connection whatsoever to reality."
Massimo Dentico wrote:
"This is the fun part of your message: you *seem* covertly despise
the philosophy and then you propose the same theme of a
philosopher like Penrose."
Then finally Kyle Lahnadoski wrote:
"But I suspect that QM is just a statistical approach to an
unknown deterministic process."
First off I have to apologize for not being familiar with Penrose.
As I said early on in this thread my reference relative to the
brain is Antonio R. Demasio's "Descartes' Error: Emotion, Reason,
and the Human Brain". I am not quite the blunt disbelief of
Billy, don't wish to argue an unprovable belief in either
determinism or non-determinism, or dispute that at the quantum
level the observation (which involves quanta) interferes with
(becomes part of) the process: the Heisenberg principle of
uncertainty.
At one (earliest) point in my career as a technician my
responsibility included diagnosing and correcting computer logic
circuits (in the days of tubes, diodes, resistors, and
capacitors). Much has changed in that time of what was used to
construct computers, but basically nothing has changed relative to
how. It is still logic circuits aggregated into instruction sets.
IMHO that's the key here, the instruction set, particularly the
fixed instruction set. No matter how elaborate, sophisticated, or
levels of meta-programming, no matter how high the level of the
HLL, it all occurs within the instruction set. No executable
software exists no so translated into the instruction set of the
machine for which it is intended.
Basically it does not make any difference how high a level of
abstraction occurs in our HLL of choice. When it comes to
execution, to the components of that execution, it never lies
outside the boundaries of the instruction set.
That occurs in reflection, in reflection on reflection, in
reflection on reflection on reflection, and so on for as much as
we choose (as it is our choice and not that of the software).
That does not mean all mimicry is alike, only that all mimicry is
mimicry. It is mimicry because we instituted and recognize it.
No software written to date has any means on its own to change the
levels or have any cognizance of what it is doing. Any such tests
are those which we have instituted.
Beyond the instruction set we have the linear memory. While we
may construct elaborate non-linear aggregates (data and programs)
for them to execute we must translate them into a linear form.
For software to exhibit higher level abstractions on its own
(other than the patterns we have encoded into it) it must overcome
the limits of linear memory and gestalt patterns and families of
patterns as well as define them conceptually and give them names
(means of reference).
When you look at the human brain and nervous system with what
little we have learned of it and then look at a computing system
of hardware and software, both of which we know to the most
intimate detail they are different constructs entirely. There are
no fixed logic circuits in the brain (and, or, and not), no linear
memory, no linear addressing, no instruction set. As what is
there has been sufficient over time to allow us to construct
computers and software, i.e. that their components have a
realization, the question arises can the reverse also occur?
Therein lies the crux of our differences. Can the computer do for
the brain what the brain has done for it? Even with extensive
assistance from us? If it is von Neumann architecture, Turing
computational rules, fixed instruction set, fixed internal logic
circuits, and linear addressable memory, I say no. There's no
"magic" in that box.
Fare believes otherwise, that you can go up levels of abstraction
and that at some point in that upward path you achieve a
capability not present in any of the lower. Something additional
happens entirely free of all that has gone before. If I
understand Kyle Lahnakoski correctly with his purely deterministic
universe, this doesn't happen even in the brain: everything that
occurs can be accounted for by everything below it. What cofuses
me is that he offers this in support of Fare.<g>
A natural question lies in asking the conditions under which this
"spontaneous generation" occurs. If it is levels of abstraction,
then how many levels is it? What is the magic number? Where has
it occurred. Certainly not in any of the examples he has
furnished. He says in commenting on one example that we cannot
fathom the result, i.e. we cannot in an interval which we can
commit follow the logic which produced the results. However we
can write software with deterministic logic that can produce
results which we cannot replicate on our own. It still doesn't
mean that anything "extra" occurred only that we used a tool as an
extension of our capabilities. It does in the large what we can
only do in the small. It extends our limits. Good tool.
Fare is entitled to his opinions and the means he has chosen for
his path to discovery. If at some point his opinion becomes
provable fact in a scientific sense, then no such argument pursued
here will continue. I wish him well. Personally I don't feel any
of it is necessary to achieve the goals or meet the requirements
of the Tunes project or the Tunes HLL. If we achieve them without
the need for something extra, then I question even bringing it up.
Let's just say it is an example of Occam's Razor.