RE: CSI and algorithms:

From: Chris Cogan (ccogan@telepath.com)
Date: Tue Sep 19 2000 - 02:24:53 EDT

  • Next message: Ralph Krumdieck: "New species"

    At 12:50 PM 09/18/2000, you wrote:

    ><< Chris:
    >But, further, as I showed in the rest of the post, random processes will,
    >given time enough and a cumulative process such as I describe, *do*
    >generate every *possible* string of information of a given length. This is
    >not open to serious question, because it is too trivial to refute.
    >
    >
    >
    >Nelson:
    >The key thing to ask is can natural processes produce CSI. Can natural
    >selection/random mutation produce CSI from scratch. That is something
    >evolution cannot do. Evolution works with things that are already lying
    >around.
    > >>

    Hmmmm. Since the accumulation of random variations that I describe (without
    any selective factor at all) will in time produce *any* string of
    information of any finite length, then it has to produce any strings that
    are CSI. It will do this from scratch (unless you mean something like,
    "from nothing whatever"), because all that's needed is replication and
    sufficiently exhaustive variation.

    Evolution does work with things that are lying around, such as water,
    carbon, hydrogen, oxygen, etc. It works with things that replicate and
    vary, and do it in such a way that the effects of variation can be
    cumulative over time. It works with DNA, which consists entirely of
    substances "already lying around").

    The only possibility for failure of the described mechanism to produce CSI
    is that CSI turns out not to be applicable to any *possible* string of
    information. But this would be Dembski's fault, not a failure of the
    evolutionary varying process.

    If we are considering a situation in which there *is* selection (i.e., some
    variations don't reproduce), then the selective mechanisms determine
    whether CSI is produced or not (assuming that it is logically possible for
    DNA to contain CSI at all). Even if selection is *also* random, if the
    amount of selection is set at sufficiently high levels as to allow long
    information strings to be produced, CSI strings will still be produced,
    though detecting them in the morass of non-CSI strings might be difficult,
    and it will usually take longer.

    For any set of well-defined instructions for building something, the
    instructions themselves do not *necessarily* have to be CSI. They may
    instead represent a simple algorithm that, applied over and over, produces
    the complex result (the Mandelbrot set is an example of very rich
    complexity that arises via a simple algorithmic process).

    Further, evolution itself is a kind of algorithmic process (and,
    essentially, a very *simple* one, at that) that can demonstrably produce
    *any* degree or "kind" of complexity. Whether natural genetic evolution
    *actually* does or has produced such complexity is a different issue, but
    it is obvious that such a process *can* produce such complexity, by the
    very randomness and exhaustiveness of the process (especially without any
    selection at all).

    Selection *reduces* the production of variations, and indirectly *channels*
    them. This is as far as selection, natural or otherwise, can go toward
    "creating" anything, because it cannot directly control the variation
    process. All it can do is keep some variations from serving as a "place" to
    put *further* variations (by preventing reproduction of those variations).
    That is, for example, if every step in evolution in one direction is
    prevented from replicating, such variations will only produce a kind of
    statistical "fuzz" in that direction, fuzz that is consistently "shaved"
    away by selection. This leaves resources and space for *other* variations,
    but it in no way can directly *produce* those variations. Selection limits
    and shapes the overall variation process by controlling which *already
    produced* variations get to replicate.

    Thus, while selection may *appear* to generate a new species from an old,
    it is really merely *allowing* a new species to evolve from an old one, and
    trimming away anything that deviates too far from doing so. Selection thus
    allows successful variations to become "amplified" by replication.

    It is the variations that are the real "driving" force in evolution. Since
    it is clear that variation and reproduction alone *will* yield *all*
    degrees and types of complexity, ID must rest on the idea that selection
    *prevents* this from happening. And yet, it is the effectiveness of
    selection to do much that is most vehemently denied by ID theorists.

    But still, it is *possible* for selection to select *only* for simplicity
    or for non-CSI. Clearly, in the real world, this does not seem to be
    consistently happening, or we ourselves would long ago have been selected
    out of existence, as would *all* forms of life (apparently, because all
    forms of known life would represent CSI, if I understand CSI correctly).
    Therefore, we conclude that selection is NOT preventing CSI.

    What, then, *can* prevent CSI from evolving? One other thing: The dreaded
    "macroevolutionary barrier," much talked about (in other terminology or
    only implicitly). *IF* there is some mysterious, undetected "barrier"
    between "kinds" (at some level of "kinds"), such that it prevents every
    population of kind A from ever evolving into kind B, then we'd have something.

    But, *what* would we have? That's right: some force intervening to
    *prevent* crossing that boundary, some force that *prevents* suitable
    variations from occurring.

    And, what would that be, mathematically? It would be highly *non-random*
    variation, variation which might still be random in some respects, but
    which would be highly *constrained* so as to keep it from producing a
    borderline string of kind B from a borderline string of kind A information.

    Since, mathematically, *all* possible variations will be produced if there
    is no selection, and if it is not *selection* that is preventing the
    "right" variations from occurring to result in CSI, then it must be some
    other force, either inherent in the situation (i.e., DNA and its
    environment at replication time), or from *outside* the situation that are
    *preventing* full randomness, and thus preventing exhaustiveness.

    Now, I grant you that, even if both variations and selection are random,
    *many* CSI cases would either never occur or would be killed off when they
    did occur (not *all* CSI is biologically "fit"). However, as long as *some*
    strings make it through the gauntlet for enough generations, CSI will be
    created. This is another *mathematical* fact, not speculation. The question
    is whether variation is sufficiently exhaustive in its randomness and
    whether selection is in fact too weak to prevent it from occurring.

    My answers are:

    1. Variation is not truly random (especially in the cases of species with
    large genomes of "library" information, etc.), but that it is
    *sufficiently* random(like) to be able to create all possible DNA strings
    of any length, given enough time. It is sufficiently random as to be
    exhaustive over enough time.

    2. Natural selection very *often* prevents prospective examples of CSI from
    being created, but it also allows many to be created. That is, I think that
    selection, while strong, does not always filter out incipient CSI, and
    that, therefore, some CSI *is* allowed to evolve. That is, there is no
    natural "macroevolutionary barrier" between a population of a species and a
    later species that has the same genome with some variations (between
    pre-hominid apes and modern man, for example).

    That is, it would seem that the evolutionary "algorithm," as *apparently*
    present in nature, *must* eventually produce CSI, or make laws of
    probability look very silly.

    That's what I think. If ID-theorists can show *either* that the variations
    are not sufficiently random to produce CSI, *or* that selection is
    sufficiently sever and sufficiently *biased* against CSI that it can't
    occur, I'll be *very* interested to see such a demonstration.

    SO:

    1. Is variation not random *enough* to produce CSI?

    2. Is natural selection so severely biased *against* CSI that it *prevents*
    it?



    This archive was generated by hypermail 2b29 : Tue Sep 19 2000 - 02:29:03 EDT