# Re: Information request re: Dawkins' "weasel" algorithm

From: Richard Wein (rwein@lineone.net)
Date: Tue Oct 10 2000 - 13:22:59 EDT

• Next message: Chris Cogan: "Re: (non-flame post) good chess programs intelligent?"

Oh dear. Maximum embarrassment. ;-)

I've just taken a trip down to the public library to re-read the relevant
section of TBW. (They only had a reference copy in stock, so I couldn't
bring it home with me.) It turns out my memory of the Weasel model was
faulty. Wesley was right. Both Dembski and I were wrong.

My aplogies to Wesley and all whose time I've wasted.

Richard Wein (Tich)
--------------------------------
"Do the calculation. Take the numbers seriously. See if the underlying
probabilities really are small enough to yield design."
-- W. A. Dembski, who has never presented any calculation to back up his

-----Original Message-----
From: Richard Wein <rwein@lineone.net>
To: Wesley R. Elsberry <welsberr@inia.cls.org>; dembski@discovery.org
<dembski@discovery.org>
Cc: evolution@calvin.edu <evolution@calvin.edu>; welsberr@inia.cls.org
<welsberr@inia.cls.org>
Date: 10 October 2000 16:05
Subject: Re: Information request re: Dawkins' "weasel" algorithm

>From: Richard Wein <rwein@lineone.net>
>
>>From: Wesley R. Elsberry <welsberr@inia.cls.org>
>>
>>>Information request to William Dembski:
>>>
>>>[Quote]
>>>
>>>He starts with a target sequence taken from Shakespeares
>>>Hamlet, namely, METHINKS IT IS LIKE A WEASEL. If we tried to
>>>attain this sequence by pure chance (for example, by randomly
>>>shaking out scrabble pieces), the probability of getting it on
>>>the first try would be around 1 in 10^40, and correspondingly
>>>it would take on average about 10^40 tries to stand a better
>>>than even chance of getting it.12 Thus, if we depended on pure
>>>chance to attain this target sequence, we would in all
>>>likelihood be unsuccessful. As a problem for pure chance,
>>>attaining Dawkinss target sequence is an exercise in
>>>generating specified complexity, and it becomes clear that
>>>pure chance simply is not up to the task.
>>>
>>>But consider next Dawkins' reframing of the problem. In place
>>>of pure chance, he considers the following evolutionary
>>>capital Roman letters and spaces (thats the length of METHINKS
>>>IT IS LIKE A WEASEL); (2) randomly alter all the letters and
>>>spaces in the current sequence that do not agree with the
>>>target sequence; (3) whenever an alteration happens to match a
>>>corresponding letter in the target sequence, leave it and
>>>randomly alter only those remaining letters that still differ
>>>from the target sequence. In very short order this algorithm
>>>converges to Dawkinss target sequence. In The Blind
>>>Watchmaker, Dawkins recounts a computer simulation of this
>>>algorithm that converges in 43 steps.13 In place of 10^40
>>>tries on average for pure chance to generate the target
>>>sequence, it now takes on average only 40 tries to generate it
>>>via an evolutionary algorithm.
>>>
>>>[End Quote - WA Dembski, "Can Evolutionary Algorithms Generate
>>>Specified Complexity", "Nature of Nature" conference, Baylor
>>>University]
>>>
>>>There are several issues that this text brings up. Of the three
>>>steps listed as comprising Dawkins' algorithm, only step (1) has
>>>anything like it in the pages of "The Blind Watchmaker". Steps
>>>(2) and (3) appear to be inventions rather than descriptions.
>>>What is the basis for claiming that steps (2) and (3) represent
>>>Dawkins' "weasel" algorithm?
>>>
>>>Further on, the issue of "tries" it takes to find a solution
>>>is raised. For "pure chance", a figure of ~10^40 "tries" is
>>>given, which would correspond to individual candidate
>>>solutions tested. For "weasel", though, only ~40 "tries" are
>>>given, but in this case the number 40 derives from the number
>>>of generations taken by the "weasel" algorithm rather than the
>>>number of candidate solutions examined. It seems to me that
>>>for the purpose of comparison, a "try" ought to mean the same
>>>thing for both approaches. I would like to see a restatement
>>>of the section concerning "tries" that takes this into
>>>account.
>>
>>It's been a while since I read TBW, but I'm almost certain you're wrong
>>here, Wesley. Dembski's description above of Dawkins' weasel algorithm
>seems
>>OK to me (except that I wouldn't call the weasel model an "evolutionary
>>algorithm", because it has a built-in target, and I don't think Dawkins
>>calls it one.)
>
>On re-reading, I see that Dembski's description of the weasel model is less
>clear than I first thought. But it can just about be reconciled with
>Dawkins' original.
>
>Correctly described, each randomization of the remaining unmatched
>characters is considered one step, and proceeds whether or not any new
match
>was achieved at the last step. Dembski has the algorithm randomizing the
>remaining unmatched characters when, and only when, a new match is made.
>This creates following potential problems:
>(a) One has to assume that each randomization (of all remaining unmatched
>characters) is completed before proceeding to check whether any new matches
>have occurred, but this is not clear from Dembski's account.
>(b) Dembski's account implies that if, at any stage, randomizing the
>remaining unmatched characters fails to produce another match, then the
>algorithm ceases; this is clearly wrong, but one can assume it's not what
>Dembski intended.
>(c) It's not clear from Dembski's account what constitutes a "step".
>
>I think the issue is one of poor writing on Dembski's part rather than a
>difference in interpretation of Dawkins' model. If one knows what Dawkins'
>really wrote and interprets Dembski generously, then there should be no
>problem.
>
>Richard Wein (Tich)
>--------------------------------
>"Do the calculation. Take the numbers seriously. See if the underlying
>probabilities really are small enough to yield design."
> -- W. A. Dembski, who has never presented any calculation to back up his