Re: CSI, GAs, etc.

From: SZYGMUNT@EXODUS.VALPO.EDU
Date: Tue Oct 03 2000 - 13:26:37 EDT

  • Next message: DNAunion@aol.com: "Re: muliplte persona alert!"

    Richard,

    You wrote:

    ===========================================================
    Chris, I'm probably just repeating something you've already written, but I
    I'd like to restate it in my own words...

    If a computer simulation faithfully models a natural process, then it is not
    displaying any more intelligence than the natural process itself (and
    probably *less*, since real computer simulations are actually simplified
    models of natural processes). Thus, a computer simulation of evolution is no
    more intelligent than the natural process of evolution itself.

    The intelligence of the computer designer or programmer is irrelevant. Is a
    computer simulation more intelligent if it's run on the latest Cray
    supercomputer than if the identical simulation is run on an old TRS-80? Is a
    computer program more intelligent if it's written by Einstein than if the
    same program was written by a trainee programmer? Of course not. The
    intelligence of the simulation is purely a function of the algorithm that is
    being executed. And, if the simulation is running the same algorithm as
    natural evolution (or a simplified version of it), then the simulation is no
    more intelligent than the natural process of evolution.

    Phew... it's *difficult* explaining the obvious!

    When IDers claim that there's a fundamental distinction between the
    algorithm as it's executed in nature and the same algorithm as it's executed
    by a computer, because the latter is somehow infused with the intelligence
    of the programmer, they're effectively denying the value of computer
    simulations in general. Oh dear...there goes *another* field of endeavour
    sacrificed on the altar of Intelligent Design, before which knowledge
    and reason must bow their heads.

    Richard Wein (Tich)
    ==============================================================

    I don't think you really addressed Paul Nelson's point here. I believe
    you are correct that the intelligence (or intelligent infusion of
    information) of the programmer is not an issue **if the computer program
    faithfully models a natural process**. For example, I trust that
    Monte Carlo simulations of an experiment in particle physics will
    give results that agree with experiment, if the relevant physics equations
    are faithfully written into the program and if the numerical techniques
    are handled carefully. If such simulations do *not* agree with experiment,
    then there is a possibility that there is a need for *new physics* that
    is not modeled by the simulation.

    I believe Paul was pointing out that so-called genetic algorithms (GAs)
    are not really faithful models of "random mutation and natural selection".
    My knowledge of GAs is limited, but I believe that programmers spend a
    considerable amount of time designing such algorithms in order to efficiently
    solve a given problem. Since their goal is *not* to faithfully reproduce
    conditions for RM & NS but rather *to solve a specific problem*, they choose
    special cases of "mutation" conditions that allow for quick problem solving.
    They have enough foresight and can perform enough trials to know what
    parameters will be best suited to a particular task. I think this is what
    Paul was pointing out (although he can certainly speak for himself).

    I'm not claiming that no such *realistic* evolution simulations exist,
    only that I don't really think that GAs faithfully model a natural
    process. THAT is the real issue. You may think that they do, in which case
    a detailed discussion of the nuts and bolts of GAs is in order. But if
    they don't, then GAs become completely irrelevant to questions of origins and
    evolution. Right?

    Thanks,

    Stan Zygmunt
    Dept. of Physics and Astronomy
    Valparaiso University



    This archive was generated by hypermail 2b29 : Tue Oct 03 2000 - 13:24:52 EDT