The "Evolutionary Algorithm" as Intelligence by Proxy

From: Chris Cogan (ccogan@telepath.com)
Date: Sun Oct 01 2000 - 21:47:13 EDT

  • Next message: DNAunion@aol.com: "Intelligence without a brain?"

    The "Evolution Algorithm" as Intelligence by Proxy
    (I'm sure others have made some or all of the same points that I make
    below, but some may find my formulation either clarifies things or brings
    to mind understandings that didn't arise from reading others' comments on
    the Intelligence-by-Proxy issue, so here it is.)

    Some ID theorists have been claiming that the "evolutionary
    algorithm" that is used to simulate natural evolution in computers is a
    case of "intelligence by proxy" because the algorithm is written by
    intelligent humans.

    Can this be true? Can it be that this fact alone makes the algorithm a
    proxy for an intelligent designer? As a hint to my main conclusion,
    ask yourself: Is a pile of rocks intelligence by proxy if it happens to
    have been built by an intelligent person?

    But, before you decide, consider some problems with this claim:

    1. The algorithm is a simple, highly mechanical process, and the
    production of variations can be completely randomized for each
    "generation" of the algorithm, and the process can still work to
    produce *any* degree or kind of complexity as long as that degree or
    kind of complexity is possible to the medium (computer-stored bit
    strings, genes, etc.).

    2. The process works by *blindly* generating variations. This is
    *not* intelligence at work. Intelligence design, by ID theorists' own
    claims, is *pre-selection* of results, as when a person intelligently
    writes what he wants to say rather than merely selecting bits and
    pieces as they are produced by some random word-generator in a
    computer.

    3. Even selection may be purely random and the process will *still*
    work, in a large class of cases, to produce any finite degree or type
    of complexity.

    4. Nature, fed with energy and beginning with material that can be
    formed into autocatalytic sets of molecules, can perform this
    "algorithm" without any intelligence at all.

    5. The algorithm is used by human researchers as a means of
    modeling processes in Nature, and is therefore no more "intelligent"
    than what it models. Generally, computer models are *less*
    "intelligent" than what they model, because it is too hard to
    incorporate *every* significant aspect of what is being modeled into
    the computer program. Thus, the algorithm is no more intelligent
    than gathering together suitable physical materials and attempting to
    recreate the *unintelligent* conditions of early Earth to see if life
    will arise.

    6. We don't even *have* to have the algorithm. We can simulate a
    "universe" with "particles" that have a "physics" and that "exist" in
    an "environment" of other "particles," with "energy flows"
    ("thermodynamic disequilibrium"), but that is otherwise as random
    and dumb as a Young Earth Creationist. If the conditions thus
    determined happen to fall within certain ranges, the evolutionary
    algorithm will arise without further intervention. Why? Precisely
    *because* it doesn't need any intelligence, and it's not really an
    algorithm but rather merely a dumb *physical* process that we
    reproduce in a computer by necessarily algorithmic means because
    we are reproducing or modeling it in a *computer*.

    7. One of the reasons we use such processes in computers is that we
    are *not* intelligent enough to solve certain problems, and therefore
    are not smart enough to tell a computer how to solve them. So,
    instead, we institute a totally mindless search mechanism that
    randomly generates variations and then, when it runs out of memory
    and disk space (if not sooner), begins selecting out those that do not
    produce results that have some similarity to the results our successful
    program will ultimately need to have. This may be done by nothing
    more sophisticated (i.e., nothing more "intelligent") than
    mechanically comparing lists of numbers.

    The point is to bypass the need for intelligence by using the dumb
    *speed* of the computer to try many bits and pieces of prospective
    solutions, filtering out the attempts that seem least likely to work.

    Design is *pre-*selection. A fully intelligent designer does not need
    *ever* to produce useless variations, because it is intelligent and
    knowledgeable enough to know beforehand what will work. In
    intelligent design, there is no massive amount of variation *because*
    it's intelligent.

    The evolutionary algorithm is random variation followed by *post-
    *selection. It is only needed when sufficiently intelligent *design* is
    not available. The evolutionary algorithm is *hugely* wasteful in
    design terms, precisely because it doesn't know what it's doing. And
    it doesn't know because it's not intelligent.

    If we grant that the evolutionary algorithm is intelligence by proxy,
    ID theorists would be right in this one case at the cost of their entire
    real position. This is because, if we grant that such a mindless
    process as the evolutionary algorithm is intelligent, then blind,
    mindless *Nature* is "intelligent" in the same sense, and thus does
    not need God or any other alleged non-natural designer to direct it to
    produce functionally complex living things.

    That is, if the "designer" is merely something at the level of this
    mindless "algorithm," then the essentials of ID theory are being
    conceded as needless, because it is precisely this kind of mindless
    algorithmic process that is the essence of evolutionary theory. No
    God, no demi-gods, no aliens, no minds, etc., are needed for this kind
    of "designer." If *this* is all ID theorists are claiming, then there
    really isn't any substantive dispute at all.

    But, it's not. Their criticism of the algorithm is a red herring. The
    idea is to get people to focus on the intelligence of the designer of
    the computer program, while totally ignoring the relevant issue,
    which is the intelligence, if any, of the algorithm *itself*.

    What they are counting on is that many people will not bother to
    think even one step further, that people won't realize that it doesn't
    matter *how* intelligent the designer of the algorithm is, if the
    algorithm itself does not plan, does not have foresight, does not *pre-
    *select components or other aspects of the result, does not *think*,
    and does not have a *goal* in mind, then it *isn't* intelligent design.

    But, what if they are underestimating their audience? What if they
    are forgetting that some of us naturalistic evolutionists and even
    other *theistic* evolutionists may see through their attempt at
    equating the powers of a mindless algorithm with the super-
    intelligence of their omnipotent God, and realize that, *if* the
    equation holds, it actually *refutes* their main lines of claims thus
    far? What if we point out to people that this algorithm, by proxy or
    not, is not an intelligent designer, because it doesn't *do* any
    designing?

    I don't know, but, to me, it makes those ID theorists who make this
    argument look pretty silly and imprudent, because it takes so *little*
    to show that equating this mindless algorithm with either or both of
    intelligence and design essentially wipes out their position by
    effectively *redefining* intelligent design to mean or to include
    mindless mechanical algorithms such as the very evolutionary
    process we naturalistic evolutionists have been arguing for all along.

    What happened to super-intelligence? What happened to
    omniscience? What happened to foresight? What happened to
    planning? What happened to achieving a *goal*?

    Conclusions:

    1. The intelligence-by-proxy type of argument is unsound unless the
    proxy itself is intelligent or at least *functions* intelligently (like a
    chess-playing program).

    2. The present argument confuses the intelligence of the programmer
    (if there is one) with the intelligence (if any) of the *program*. Since
    the evolutionary algorithm is not intelligent, it cannot be argued that
    the intelligence of the programmer is transmitted via the program to
    the results of the program. The algorithm itself has the intelligence of
    the pile of rocks mentioned at the beginning. If the pile of rocks
    happens to have been made by an intelligent person, it doesn't make
    it more intelligent than it would be if the rocks just happened to end
    up in a pile at the conclusion of a rock-slide.

    3. If we accept the claim, it reduces ID to the level of claiming the
    existence of a totally mindless algorithm that produces complex
    results. This would be tantamount to abandoning the whole position
    of ID theory, and accepting naturalistic evolution as, at most, the way
    God chose as His means of producing functionally complex things.

    4. ID theorists are *often* way too eager to use any argument that
    seems superficially to support their position, without bothering with
    soundness. This tends to discredit them and their position as one
    argument after another is shown to be counterfactual, or non-factual,
    The "Evolution Algorithm" as Intelligence by Proxy

    Some ID theorists have been claiming that the "evolutionary
    algorithm" that is used to simulate natural evolution in computers is a
    case of "intelligence by proxy" because the algorithm is written by
    intelligent humans.

    Can this be true? Can it be that this fact alone makes the algorithm a
    proxy for an intelligent designer? As a hint to my main conclusion,
    ask yourself: Is a pile of rocks intelligence by proxy if it happens to
    have been built by an intelligent person?

    But, before you decide, consider some problems with this claim:

    1. The algorithm is a simple, highly mechanical process, and the
    production of variations can be completely randomized for each
    "generation" of the algorithm, and the process can still work to
    produce *any* degree or kind of complexity as long as that degree or
    kind of complexity is possible to the medium (computer-stored bit
    strings, genes, etc.).

    2. The process works by *blindly* generating variations. This is
    *not* intelligence at work. Intelligence design, by ID theorists' own
    claims, is *pre-selection* of results, as when a person intelligently
    writes what he wants to say rather than merely selecting bits and
    pieces as they are produced by some random word-generator in a
    computer.

    3. Even selection may be purely random and the process will *still*
    work, in a large class of cases, to produce any finite degree or type
    of complexity.

    4. Nature, fed with energy and beginning with material that can be
    formed into autocatalytic sets of molecules, can perform this
    "algorithm" without any intelligence at all.

    5. The algorithm is used by human researchers as a means of
    modeling processes in Nature, and is therefore no more "intelligent"
    than what it models. Generally, computer models are *less*
    "intelligent" than what they model, because it is too hard to
    incorporate *every* significant aspect of what is being modeled into
    the computer program. Thus, the algorithm is no more intelligent
    than gathering together suitable physical materials and attempting to
    recreate the *unintelligent* conditions of early Earth to see if life
    will arise.

    6. We don't even *have* to have the algorithm. We can simulate a
    "universe" with "particles" that have a "physics" and that "exist" in
    an "environment" of other "particles," with "energy flows"
    ("thermodynamic disequilibrium"), but that is otherwise as random
    and dumb as a Young Earth Creationist. If the conditions thus
    determined happen to fall within certain ranges, the evolutionary
    algorithm will arise without further intervention. Why? Precisely
    *because* it doesn't need any intelligence, and it's not really an
    algorithm but rather merely a dumb *physical* process that we
    reproduce in a computer by necessarily algorithmic means because
    we are reproducing or modeling it in a *computer*.

    7. One of the reasons we use such processes in computers is that we
    are *not* intelligent enough to solve certain problems, and therefore
    are not smart enough to tell a computer how to solve them. So,
    instead, we institute a totally mindless search mechanism that
    randomly generates variations and then, when it runs out of memory
    and disk space (if not sooner), begins selecting out those that do not
    produce results that have some similarity to the results our successful
    program will ultimately need to have. This may be done by nothing
    more sophisticated (i.e., nothing more "intelligent") than
    mechanically comparing lists of numbers.

    The point is to bypass the need for intelligence by using the dumb
    *speed* of the computer to try many bits and pieces of prospective
    solutions, filtering out the attempts that seem least likely to work.

    Design is *pre-*selection. A fully intelligent designer does not need
    *ever* to produce useless variations, because it is intelligent and
    knowledgeable enough to know beforehand what will work. In
    intelligent design, there is no massive amount of variation *because*
    it's intelligent.

    The evolutionary algorithm is random variation followed by *post-
    *selection. It is only needed when sufficiently intelligent *design* is
    not available. The evolutionary algorithm is *hugely* wasteful in
    design terms, precisely because it doesn't know what it's doing. And
    it doesn't know because it's not intelligent.

    If we grant that the evolutionary algorithm is intelligence by proxy,
    ID theorists would be right in this one case at the cost of their entire
    real position. This is because, if we grant that such a mindless
    process as the evolutionary algorithm is intelligent, then blind,
    mindless *Nature* is "intelligent" in the same sense, and thus does
    not need God or any other alleged non-natural designer to direct it to
    produce functionally complex living things.

    That is, if the "designer" is merely something at the level of this
    mindless "algorithm," then the essentials of ID theory are being
    conceded as needless, because it is precisely this kind of mindless
    algorithmic process that is the essence of evolutionary theory. No
    God, no demi-gods, no aliens, no minds, etc., are needed for this kind
    of "designer." If *this* is all ID theorists are claiming, then there
    really isn't any substantive dispute at all.

    But, it's not. Their criticism of the algorithm is a red herring. The
    idea is to get people to focus on the intelligence of the designer of
    the computer program, while totally ignoring the relevant issue,
    which is the intelligence, if any, of the algorithm *itself*.

    What they are counting on is that many people will not bother to
    think even one step further, that people won't realize that it doesn't
    matter *how* intelligent the designer of the algorithm is, if the
    algorithm itself does not plan, does not have foresight, does not *pre-
    *select components or other aspects of the result, does not *think*,
    and does not have a *goal* in mind, then it *isn't* intelligent design.

    But, what if they are underestimating their audience? What if they
    are forgetting that some of us naturalistic evolutionists and even
    other *theistic* evolutionists may see through their attempt at
    equating the powers of a mindless algorithm with the super-
    intelligence of their omnipotent God, and realize that, *if* the
    equation holds, it actually *refutes* their main lines of claims thus
    far? What if we point out to people that this algorithm, by proxy or
    not, is not an intelligent designer, because it doesn't *do* any
    designing?

    I don't know, but, to me, it makes those ID theorists who make this
    argument look pretty silly and imprudent, because it takes so *little*
    to show that equating this mindless algorithm with either or both of
    intelligence and design essentially wipes out their position by
    effectively *redefining* intelligent design to mean or to include
    mindless mechanical algorithms such as the very evolutionary
    process we naturalistic evolutionists have been arguing for all along.

    What happened to super-intelligence? What happened to
    omniscience? What happened to foresight? What happened to
    planning? What happened to achieving a *goal*?

    Conclusions:

    1. The intelligence-by-proxy type of argument is unsound unless the
    proxy itself is intelligent or at least *functions* intelligently (like a
    chess-playing program).

    2. The present argument confuses the intelligence of the programmer
    (if there is one) with the intelligence (if any) of the *program*. Since
    the evolutionary algorithm is not intelligent, it cannot be argued that
    the intelligence of the programmer is transmitted via the program to
    the results of the program. The algorithm itself has the intelligence of
    the pile of rocks mentioned at the beginning. If the pile of rocks
    happens to have been made by an intelligent person, it doesn't make
    it more intelligent than it would be if the rocks just happened to end
    up in a pile at the conclusion of a rock-slide.

    3. If we accept the claim, it reduces ID to the level of claiming the
    existence of a totally mindless algorithm that produces complex
    results. This would be tantamount to abandoning the whole position
    of ID theory, and accepting naturalistic evolution as, at most, the way
    God chose as His means of producing functionally complex things.

    4. ID theorists are *often* way too eager to use any argument that
    seems superficially to support their position, without bothering with
    soundness. This tends to discredit them and their position as one
    argument after another is shown to be counterfactual, or non-factual,
    self-refuting, or just plain invalid.

    --Chris Cogan



    This archive was generated by hypermail 2b29 : Sun Oct 01 2000 - 21:51:50 EDT