Uncommon Descent Contest Question 10: Provide the Code for Dawkins’ WEASEL Program

_{Denyse O'Leary

August 26, 2009

Darwinism}

Share: Facebook; Twitter; LinkedIn; Flipboard; Print; Email

Special invitation for Richard Dawkins – but any civil person is entitled to enter.

There’s been some discussion here and elsewhere whether the the recent IEEE article by Dembski and Marks correctly characterizes Richard Dawkins’ famous METHINKS IT IS LIKE A WEASEL program.

Does the program ratchet correct letters or does it let them vary? One is a partitioned or stair-step search, the other a more realistic evolutionary search. From The Blind Watchmaker, where Dawkins describes the program, its performance suggests that it could be either of these options (though he doesn’t say).

On the other hand, from a (video-run of the program , go to 6:15), it seems to be the latter.

It’s easy enough to settle this question: Make the code for the program public. Perhaps Richard Dawkins himself or his friends at RichardDawkins.net can finally provide this code (apparently a program written in BASIC).

The prize is a copy of either Stephen Meyer’s new Signature in the Cell or Richard Dawkins’ soon-to-be-out The Greatest Show on Earth.

Should the winner choose the latter, I will ask Dawkins’s publicist to mail the copy. Given that at his site, he calls himself “the most formidable intellect in public discourse,” I would assume that if he signs the copy, it will be worth millions.

But wait. Let’s see that code first.

Comments

kf, you can I point you to where you ignore my question? You ignore it in all your thousands of words. Why not just answer it? Which algorithm from Atoms suite is Weasel, which one the partitioned search? Are they the same? Same active info? Same number of queries?Indium_{August 30, 2009
August
08
Aug
30
30
2009
05:19 AM
5
05
19
AM
PDT}

Kariosfocus
So, your mantra on “ignored” is rising to the level of reite3rated false allegations, i.e slander.
One could say the same about you. You have ignored Dawkins own comment on latching and Weasel. You have ignored comments that you are only seeing 0.047% of the total data and making assumptions based on that tiny fragment and applying the to the whole data set, as per comments #41, #78 and #149. You have ignored Indium's simple question - "Which one of Atoms algorithms is Weasel, which one the partitioned search? Are they the same?" You have ignored my question #190, asked several times elsewhere and answered with a flurry of strawmen but no comment on the substance of the question itself.
It is sadly evident from the above that we have much of straining at gnats while swallowing camels, as well as barking up wrong trees. Etc, etc.
Perhaps you would be better served by addressing the issues on their merits rather then trying to accuse everyone who disagrees with you of slander.Blue Lotus_{August 30, 2009
August
08
Aug
30
30
2009
05:17 AM
5
05
17
AM
PDT}

Oh, sorry for the doulbe post.Indium_{August 30, 2009
August
08
Aug
30
30
2009
05:17 AM
5
05
17
AM
PDT}

kf, anorther few thousand words ignoring my question, as expected. Too keep your evasion on display I will just repeat it: Which one of Atoms sehr algorithms is Weasel, which one the partitioned search? Are they the same? I will just briefly comment on a few of your points, there is too much noise in there to spend more time on it. Whenever you are ready to make a clear argument without all this obfuscation I will again be more exhaustive.
implicit latching is an observed phenomenon, one that answers to CRD’s enthusiastic description and showcased printoffs.
Of course it is: Nobody denies that evolution or Weasel are highly efficient searches. We are denying that that Weasel protects correct letters from mutating.
The primary fact is that Weasel is a confessed, targetted search
Confessed targeted search? What are you talking about? It´s written with big letters in the book, there is nothing to confess at all! And this is not the question at hand anyway. The question is: Doe D+M misrepresent it?
As such, Weasel c 1986 as presented is “fair game” for an analysis as-is, on the implications of the active information manifested by such cumulative, evidently ratcheted and latched search on mere proximity not relevant complex functionality.
This alone is enough to amke any further argument with you rather unnecessary. Weasel is completely misrepresented in the paper by D+M. It´s not fair game to just plug in a different algorithm that happens to also keep correct letters (which Weasel sometimes even doesn´t do) and say your discussing Weasel. Also, please notice that D+M explicitly cite TBW and D announced on this blog that his paper criticizes Weasel and therefore even evolution. They give very specific results for the active info, which are plainly wrong for Weasel. Eq 22 is in no way consistent with Weasel (independend from latching). They are discussing a completely different algorithm, it is not fair game at all. That you consider this to be fair game speaks volume about a few things...
how that latching is achieved is irrelevant to the point that active information is a key reason for the performance above unassisted random search, and to quantify the injected active information
Quantifying active info for different searches is the main point of the paper by D+M. And yet the results for the partitioned search and Weasel are completely different which you would notice if you would just fire up Atoms software and check it. Or you can verify that Eq 22 is simply wrong for Weasel (latching or not latching, doesn´t matter) and therefore also the calculation is wrong. That does not imply that the active info concept can´t be applied to Weasel, I am sure it can be done! But it´s done wrong in the paper because Weasel simply is no partitioned search. Ask Joseph. kf, honestly, these things are now so clear to see for everyone that you are just damaging your reputation. The discussion and formulas in the paper are not correct for Weasel because quite simply Weasel is no partitioned search. Which is again a good point to repeat my question: Which of the Search Algorithms in Atoms suite is Weasel and which one the partitioned search? Maybe asking it two times in one post helps you to focus. Oh, and the bonus question: Do they seem to have the same amount of active info?Indium_{August 30, 2009
August
08
Aug
30
30
2009
05:13 AM
5
05
13
AM
PDT}

Indium: You first. I have repeatedly pointed out to you that most of your relevant queries were long since answered, starting with the always linked. So, your mantra on "ignored" is rising to the level of reite3rated false allegations, i.e slander. Kindly stop it. Point out where my answers are INADEQUATE on the merits -- if you can -- then you may go on from there. GEM of TKIkairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
05:02 AM
5
05
02
AM
PDT}

kf, anorther few thousand words ignoring my question, as expected. Too keep your evasion on display I will just repeat it: Which one of Atoms sehr algorithms is Weasel, which one the partitioned search? Are they the same? I will just briefly comment on a few of your points, there is too much noise in there to spend more time on it. Whenever you are ready to make a clear argument without all this obfuscation I will again be more exhaustive.
implicit latching is an observed phenomenon, one that answers to CRD’s enthusiastic description and showcased printoffs.
Of course it is: Nobody denies that evolution or Weasel are highly efficient searches. We are denying that thatIndium_{August 30, 2009
August
08
Aug
30
30
2009
04:49 AM
4
04
49
AM
PDT}

KariosFocus, I ask you one more time. If I have two programs. Once always outputs data similar to Generation 1: XXXYYYXXXZZZ Generation 2: XXXYYXXXYZZY The other always outputs data similar to Generation 1: XXXYYYXXXZZZ Generation 2: REJSJTHVHASG Would any reasonable person claim that they were in fact the same program producing similar outputs? No. Yet given two programs where the outputs are Generation 1 Y*YVMQKZPFJXWVHGLAWFVCHQXYPY Generation 10 Y*YVMQKSPFTXWSHLIKEFV*HQYSPY and Generation 1: XYBPFPBMCLPGFONJYWFKXPFOVMKDX Generation 2: XPJSQJMGPRYPONAIRJSGXZRWQJQBX you continue to claim that they are the same program! Please note the latter set of results were generated via WeaselWare here http://www.evoinfo.org/WeaselGUI.html So they are valid results. The previous generation 1 and 10 are obviously from TBW. Please explain Kariosfocus how these two outputs are in fact being generated by the same program?Blue Lotus_{August 30, 2009
August
08
Aug
30
30
2009
04:26 AM
4
04
26
AM
PDT}

The first Dawkins quote should have been
Alas, I no longer have the original program. It seemed too trivial to be worth keeping. Obviously any half way decent programmer could knock it up in a trice. Fortunately, however, there is a film of the program in operation (see post #53) above, and you can clearly see from the film that there was no 'latching'.
http://scienceblogs.com/pharyngula/2009/08/but_why_would_dawkins_want_to.php#comment-1887052Blue Lotus_{August 30, 2009
August
08
Aug
30
30
2009
04:13 AM
4
04
13
AM
PDT}

Kariosfocus
–> on “forensically” reconstructing the algorithm to do that, two main approaches are possible: explicit latching and implicit where the pop per gen, mut per letter rate and filter interact to at least some of the time give runs that ratchet. Both have been demonstrated and are legitimate readings on the evidence of 1986. –> on subsequent statements (and possibly the 1987 video] the latter is the — on balance of evidence — best explanation for the observed published runs and descriptions c 1986.
Richard Dawkins has responded and has said
–> on “forensically” reconstructing the algorithm to do that, two main approaches are possible: explicit latching and implicit where the pop per gen, mut per letter rate and filter interact to at least some of the time give runs that ratchet. Both have been demonstrated and are legitimate readings on the evidence of 1986. –> on subsequent statements (and possibly the 1987 video] the latter is the — on balance of evidence — best explanation for the observed published runs and descriptions c 1986.
http://scienceblogs.com/pharyngula/2009/08/but_why_would_dawkins_want_to.php#comment-1887052 And
http://www.youtube.com/watch?v=5sUQIpFajsg About 6:10 seconds in, watch the 'Darwin' algorithm home in on 'METHINKS IT IS LIKE A WEASEL'. You will see (for instance at 6:16) that the program does not 'latch', because the W of weasel mutates and then comes back. It keeps winking on and off from W. Clearly no latching. However, as PZ and others have said, it really doesn't make a lot of difference whether the program 'latches' or not. These people are so unbelievably stupid. I claim my trip to the Bahamas. Richard
http://scienceblogs.com/pharyngula/2009/08/but_why_would_dawkins_want_to.php#comment-1886689 Note that Richard says "the" program not "This version of Weasel (which is different to the version on TBW)". Therefore according to Richard Dawkins there is only 1 version of Dawkins' Weasel and it behaves as per the video (no latching). Will you now withdraw your opinion or continue to say that Richard Dawkins does not know how Richard Dawkins Weasel operates?Blue Lotus_{August 30, 2009
August
08
Aug
30
30
2009
04:08 AM
4
04
08
AM
PDT}

PPS: I also happen to think his partnership with Dr Marks has enriched his work. And, that on a topic known for months to be controversial, there would have been significant cross-checking before publication. Now, compare that with the sort of srtrawmannising above, and it should be evident why I draw the conclusions I do on who is more likely to be correct in this case. then, multiply by the obvious didactic example context of the alleged algorithm that they are being castigated for. And mix in the fact that explicitly AND implicitly latched weasel programs have been demonstrated on actual runs -- programs sponsored on the web by the same EIL. After such factors are in evidence, what makes the best overall explanation? On what grounds?kairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
04:08 AM
4
04
08
AM
PDT}

PS: OS, If you rake time to look at my characteristic thought on ID [e.g through the always linked] in terms of functionally specific, complex information and its roots in thermodynamical and informational thinking (which trace back through Thaxton et al, not Dembski), as well as my related look at the Caputo case, you will see that my thought is significantly independent from that of Mr Dembski. I happen to think that Mr Dembski -- though finite, fallible and fallen as we all are -- has got some things right, things that are too often caricatured and wrenched by objectors to improperly dismiss them through strawman fallacies. And the habit of such strawmannising by denizens of Darwinland is I believe abundantly evident above.kairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
03:52 AM
3
03
52
AM
PDT}

i: you have redefined “query” to mean the determination of the next parent string. Everybody can look at Atoms GUI to see that you are wrong of course. --> The problem at root is in the absurdities thrown out by trying to read the didactic example as an algorithm. --> Where I do have what I think on further thought overnight is an error, is that I have taken Q to be number of generations at the first, in that context. Q is -- on second thought -- number of generations multiplied by size of generations. --> And, that makes no difference to the evident pattern of ratcheting from generation to generation in the line of champions [which is what the printoffs c 1986 show; and what it therefore the empirical foundation of all discussion] that is a key part of the analysis, or tot he comparison of effect of active information based vs random search, as Q is in any case consistent across the two. j: Somehow the paragraph/pictures/formulas in the D+M paper are only a pedagogical something that not criticizes Weasel directly but is somehow relevant anyway --> Utter misrepresentation, laced with ad hominems. (Advice: If you do not understand, ask, don't assert. Please.) --> A reading of 161 ff, for instance [not to mention all the way back to the always linked app 7] will easily and clearly show that I have said that the didactic example presented of what partitioning of a search looks like, is not a realistic representation of an algorithm, but a simple illustration of a behaviour of observed outputs: in ratcheted searches, once letters go correct they are preserved correct, and more and more letters go correct and are preserved until the target phrase is complete. [And, M & D say just about as much in pretty close words to what I have just said.] --> once that is seen, we can see the relevance and accuracy of the basic analysis of what such a ratcheted search looks like probabilistically, on the mere fact of ratcheting. (Which raises no commitments on what the ratcheting comes from, whether explicit or implicit.) --> And, I have said precisely nothing about the onward analysis in the IEEE paper, as this has not come up, i.e you are putting words in my mouth above, words that simply do not belong there. k: TA, 175: How about this then: Latching occurs when the probability of moving “up” the gradient is greater than the probability of moving down. This would again seem to include any search other than a blind random walk. --> Remember, we are starting form observing an evident o/p pattern, per showcased examples of "cumulative selection" in action. --> in those examples, for 200 cases were letters go correct, and can revert, the excerpts never show a reversion. And since listing every 10th generation's champion is unlikely to correlate with the search process, then we can infer that the description and the showcased runs coincide: there is ratcheted progress to target. --> After that, we look a the "occasional slips" case; one that is also observed on producing a program capable with certain parameters being matched, of latching. this makes sense,a s probabilistic barriers are not absolute. [All the oxygen molecules in the room where you sit can conceivably rush to one end, leaving you choking; no physical barrier absolutely forbids that. But, that is rather unlikely, and unobserved.] --> And a third case is possible, where there is no evident ratcheting. l: YD, 179: RD says he no longer has the original program but that it did not “latch”. Not that latching vs. nonlatching is that important in the grand scheme of things, but it’s helpful to get confirmation of this detail from the source. --> And how did CRD explain the showcased runs and gushing remarks on the wonderful power of cumulative selection c 1986? --> Other than, that he is claiming that he did not EXPLICITLY latch the program, which would be the same thing he is reported to have said c 2000. [In short, the issue of implicit latching is still very much on the table, and recall, such behaviour on "good runs" is DEMONSTRATED. If CRD's actual o/p's on the showcased runs did not latch, implicitly or explicitly, then to present them as if they did while gushing on the power of cumulative selection will require a bit of explaining on how the results and remarks were not presented in a misleading manner.] --> In any case we have it that no credible code will be forthcoming. Contest over, unless someone can dig up a credible copy from somewhere that has a reasonable chain of custody. m: OS, 180:Dembski, observing that something possible in execution of the algorithm was not evidenced in the sample, inferred that the program did not implement the stated algorithm --> Bold denial of stated facts on the record. --> the claim by Dawkins p 48 ff of BW, was that Weasel exhibited "cumulative [and targetted] selection," based on proximity to target, which conferred a major advantage over "single step selection." Where, cumulative NORMALLY means: Increasing or enlarging by successive addition. [1st meaning AmHD.] --> in support of this, he produced listings c 1986 in BW and NewScientist, that showed over 200 cases of letters going correct and then open to reversion, without a single reversion in evidence; on two runs, one of 40+ and one of 60+ generations. [That OS thinks there was only one published run shows that he has not investigated carefully before commenting adversely.] --> On such -- multiplying the two lines of evidence together -- it is evident beyond reasonable dispute that Weasel c 1986 generational champions ratcheted to target with associated latching of successful to date letters on "good" runs. --> on "forensically" reconstructing the algorithm to do that, two main approaches are possible: explicit latching and implicit where the pop per gen, mut per letter rate and filter interact to at least some of the time give runs that ratchet. Both have been demonstrated and are legitimate readings on the evidence of 1986. --> on subsequent statements (and possibly the 1987 video] the latter is the -- on balance of evidence -- best explanation for the observed published runs and descriptions c 1986. --> the rest of OS's case foes downhill from there, repeating a now familiar line of talking points. +++++++++++++ After taking time to go through the above, I am still shaking my head. GEM of TKIkairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
02:45 AM
2
02
45
AM
PDT}

3: Selected bloopers (too many to address one by one . . . : a: BB@ 168: An explicit, required, latching mechanism is the same as non-explicit, non-required, not-always-latching behaviour --> Not at all: Strawman. (This, buy one who has already stepped outside the pale of civil discourse.) --> For the purposes of the M & D analysis, how latching is achieved for showcased runs is irrelevant. (remember, the issue is to account for the SHOWCASED runs, which will not necessarily be typical. [THAT IS PART OF WHY CRD'S CODE C. 1986 WOULD BE HELPFUL.]) --> Once latching evidently exists, the analysis applies. --> And, let us recall: CRD in 1986 inadvertently admitted that the speedup was due to the targetting on proximity. That is, active information. b: A mutation rate that has to be between zero and one hundred percent is the same as a mutation rate that has to be either zero or one hundred percent --> Strawman, again. A caricature loaded with ad hominems is being set up. --> How latching is achieved is -- for the purposes of the actual analysis -- irrelevant to that it is achieved. --> This is now willful obtuseness, as well. [One who is involved in a civil discussion has a duty of care to seek to understand an interlocutor, not to twist words taken out of context to suit one's self. But then, sadly, BillB has long since demonstrated want of civility and a habit of twisting words self-servingly to accuse falsely.] c: A population of one, where no selection can occur is the same as a population of many from which one is selected --> Again, it is quite evident from p. 1055 as already analysed, that M & D gave a simplistic pedagogical example of what partitioning looks like. (Of course they did not reckon with the sort of word twisting rhetoric we are seeing in this thread and doubtless elsewhere. I am sure the IEEE engineers looking on are beginning to see what is wrong in the state of Darwinland.) --> A simple scrollup to 162 will suffice to show that a mutation of five letters going correct at once would only be plausible for a very large population indeed, with an aggressive selection filter. So, the notion of a population of one is read into the example, not drawn out of it and its context. --> And, the other half of this objection, that Q is number of mutants to date, then runs into the point that as discussed above, we would then have Q = G* S. --> the rest of BillB's analysis collapses due to strawman premises. d: WEASEL is an algorithm defined by Dawkins in The Blind Watchmaker. Many people have written software based on it, and others have made software inspired by it, but which employs different strategies. There is only one WEASEL though and it is described quite clearly by Dawkins. --> from the above, it is clear that there are many legitimate ALGORITHMS for Weasel that will fit with Dawkins' description c 1986. --> In short, this is mere caricature. e: Rob @ 170:That’s a nice theory, but it’s belied by the fact that both Dembski and Marks have stated that this algorithm is, in fact, Dawkins’ WEASEL algorithm. Dembski has been saying it for years, even after correction. The EIL website still says it. This is what The EIL page linked by Rob says: First, let's look at partitioned search used by Dr. Dawkins. Assuming uniformity, the probability of successfully identifying a specified letter with sample replacement at least once in Q queries is . . . [leading up to the same summarised math as appears in p. 1055 of the IEEE paper] --> Of course, the "belying" is based on forcing the observation of partitioning -- i.e evident cumulative, ratcheted search that advances to a target on proximity as showcased c 1986 -- into a particular algorithm that implements it; and algorithm that does not appear in the IEEE paper, nor for that matter in the EIL page as linked. --> AND, we note that it is demonstrated that implicit latching is possible, which will produce the same run of champions effect as seen in the 1986 showcased run excerpts. [And if the 1986 run excerpts were ATYPICAL BEHAVIOUR (which is what some above suggest, while claiming that he program did not explicitly latch), then, that raises questions on the integrity of the Weasel program as presented at that time; questions that should be answered by opening to public inspection credible code. Recall, Joseph and I have been trying to account for the observed behaviour c 1986, on the claim that CRD did not explicitly latch his program c 1986. If implicit latching is possible but atypical, that itself raises questions about what was going on in the showcased 1986 runs. ] --> In short, more word twisting and strawmanising. With some troubling possible implicaitons that call for credible code c 1986. f: mb, 171: In the Blind Watchmaker video… you do realize that you can see the “correct” letters briefly change, right? . . . Say what you will about the applicability or accuracy of Dawkin’s program, but it obviously doesn’t lock the letters once they’re right. --> this of course raises the issue of the apparent gap between the showcased runs c 1986 and the video c 1987. --> the first serious option is that the 1987 video is a detuned run of Weasel that shows unlatched behaviour due to the detuning from the matched pop size, mutation rate and filter in 1986. --> the second [suggested by an objector to the idea of implicit latching], which does account for the winking effect, is that we are looking not at generation champions, but at the raw members of the population. (This runs into the problem that if MB's observations are accurate, then mutation per letter rate is rather high -- known to lead to one form of detuning and non-latched behaviour.) g: OS, 176:You seem to think there’s huge value in the cultural war in presenting Dembski as an inerrant genius --> Where have I ever said or implied such? [I think I am on record that we are all finite, fallible, and indeed fallen.] --> I happen to think that in this case, the M 7 D analysis -- as opposed tot he caricatures presented above -- is reasonable, and have given my reasons. --> And, pardon my suspicions when I see talking points of the now all too familiar form "I am a supporter of X, but I think the supporters of X are idiots or worse . . . " --> BTW, Weasel c 1986 is admittedly targetted search that rewards mere proximity. h: Indium, 174: an algorithm that doesn´t protect correct letters in a search is the same as one that does (the famous implicit latching!) --> Notice how this strawman distortion [cf bloopers a - c supra] has now become a repeated mantra, to be taken as gospel truth on the power of sheer brassy repetition in the teeth of the facts. [I shudder to think of what is going on in Darwinland echo chambers on this . . . ] --> Indium, in case you don't recognise the tactic, this one is called the big lie, adn I need not list its well-known exponents -- who BTW, projected it unto their intended victims, instead of telling the truth that they were the ones using it [in short, turnabout tactics]. --> Please, don't be taken in by it. --> And as for the "lost in the laugh" remark above, that too is a well-known agit-prop tactic. Stop laughing and start reading more carefully to UNDERSTAND before you criticise, please. [ . . . ]kairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
02:42 AM
2
02
42
AM
PDT}

Onlookers: Shaking my head . . . sigh. It is sadly evident from the above that we have much of straining at gnats while swallowing camels, as well as barking up wrong trees. Etc, etc. But first, let's go back a moment to where I left off, on minor issues. I: More on Q One of the peculiarities of Weasel algors, is that they halt when they hit home; of course halting being a key property of algorithms. As a result, if they hit home before a generation number G, they do not get to G. Thus, if Q means number of mutants to date, then if size of generation is S, Q = G*S. Immediately, if a run completes [all L letters correct] at gen G, it cannot have been complete before G. So, the paper's p.1055 discussion would on this reading of Q would be of a ratcheted run that shows a march of champions with latching up to G, when it halts: completion AT G. (In an explicitly latched Weasel this would be automatic, in a version that on being tuned and giving a good run latches -- as observed [our first swallowed camel . . . ] -- this implies that when metric falls to distance to target = 0, there is a latching action imposed.) And there is also the second camel: debates over the meaning of Q do not affect the OBSERVED FACT of latching (regardless of other runs that may not do so -- remember, we are accounting not for typical or overall behaviour but for runs that are showcased c. 1986 that were showcased because of their cumulative progress to target). And that is camel no 3: implicit latching is an observed phenomenon, one that answers to CRD's enthusiastic description and showcased printoffs. (The description and printoffs that the objectors above are ever so eager to direct our attention away from.) 2: Camel no 4: distractions over code and algorithms The primary fact is that Weasel is a confessed, targetted search which makes cumulative progress to target, even through generations championed by "nonsense phrases." Something which by CRD's confession is "misleading" due to the long term targetting and associated artificial selection. Indeed, CRD also highlights that the targetting and artificial selection make a big difference to time to target: tha tis we have a case of active informaiton in action. This, we can see form BW, and it is apparently necessary to give it again, as last cited at 161: ___________________ >> It [Weasel] . . . begins by choosing a random sequence of 28 letters … it duplicates it repeatedly, but with a certain chance of random error – ‘mutation’ – in the copying. The computer [indirectly, the programmer!] examines the mutant nonsense phrases, the ‘progeny’ of the original phrase, and chooses the one which, however slightly, most resembles the target phrase, METHINKS IT IS LIKE A WEASEL . . . . What matters is the difference between the time taken by cumulative selection, and the time which the same computer, working flat out at the same rate, would take to reach the target phrase if it were forced to use the other procedure of single-step selection [i.e. "more than a million million million times as long as the universe has so far existed" -- this is where CRD admits that active information has played a key role in the speedup] . . . >> ___________________ As such, Weasel c 1986 as presented is "fair game" for an analysis as-is, on the implications of the active information manifested by such cumulative, evidently ratcheted and latched search on mere proximity not relevant complex functionality. And, since the analysis of ratcheted progress to target does not depend on whether the latching is or is not implicit or explicit -- these are mechanisms to get to the observations of evident latched, cumulative, ratcheting progress to target c. 1986 -- then, how that latching is achieved is irrelevant to the point that active information is a key reason for the performance above unassisted random search, and to quantify the injected active information. [ . . . ]kairosfocus_{August 30, 2009
August
08
Aug
30
30
2009
02:41 AM
2
02
41
AM
PDT}

I gave up reading the detailed arguments a while back, but I think I'm right in saying that that kf (verbose chap, isn't he?) accepts at this point that the original TBW program didn't contain any explicit latching but gives the appearance of it because only every 10th generation is shown. I've observed this in my own (non-latching) implementation too, with n(offspring)=100 and p(mutation)=0.01. What's interesting is that even if I print out the survivor for *every* generation, I still don't see any reversion of previously correct characters to incorrect ones. Obviously the mechanism of selecting the fittest offspring from a suitably large population with a small mutation rate is a very powerful one. Which is what RD was trying to illustrate? Since it's been shown over and over (and over) again that the algorithm he described produces exactly the results he printed in the book without any explicit latching, can someone explain to me why seeing the original source code is remotely important?nephmon_{August 30, 2009
August
08
Aug
30
30
2009
01:51 AM
1
01
51
AM
PDT}

I should also add I wrote a short BASIC program coded for the Algorithm Dr. Dawkins explains in the Blind Watchmaker. It isn't Dr. Dawkins' original code, but it works the same as the sample shown in the 1987 video. I wrote it for the MS-DOS 3.3 version of BASIC, not Apple ][, but the differences should be minor.sandlinjohnb_{August 30, 2009
August
08
Aug
30
30
2009
12:52 AM
12
12
52
AM
PDT}

[My entry, made polite.] Let’s step through this “Weasel” controversy. 1. Dawkins stated an algorithm. 2. Dawkins claimed to have implemented the algorithm in his Weasel program. 3. Dawkins provided a sample of the outputs from one run of the program. 4. Dembski, observing that something possible in execution of the algorithm was not evidenced in the sample, inferred that the program did not implement the stated algorithm. 5. Dembski went beyond this inference that the program did not correctly implement the stated algorithm to an inference of the algorithm the program actually did implement. 6. Dembski and Marks attributed the inferred algorithm to Dawkins, giving no indication whatsoever that it was not the algorithm Dawkins stated, and went on to analyze it. Step 4 is logically invalid. Steps 5 and 6 are poor scholarship. Regarding step 4, the program is randomized, and the fact that something that might have happened appears not to have happened is not sufficient evidence for concluding that the program is incorrect. At best, one may question the probability that the program correctly implements the stated algorithm. Regarding steps 5 and 6, when you doubt that a program correctly implements an algorithm, you do not jump to abductive inference of the algorithm that the program does correctly implement, and then make your guess the topic of discussion. Proper conduct for a scholar would be to a) work with the author of the algorithm and program to determine if the program is correct, and b) develop a correct program if the original program is incorrect. It is never appropriate to say, “Well, let’s just change the algorithm to match the program.” Programs implement algorithms, and algorithms are not cooked up after the fact to describe what programs happen to do. Whether or not Dawkins has provided the Weasel program is utterly irrelevant. Dembski and Marks analyze an algorithm, not a program. The correctness of a mathematical analysis of the algorithm is independent of the correctness of any implementation of the algorithm. The appropriate challenge here is to provide the email in which Dembski and Marks ask Dawkins to confirm that the algorithm they attribute to him is in fact his algorithm.Oatmeal Stout_{August 30, 2009
August
08
Aug
30
30
2009
12:06 AM
12
12
06
AM
PDT}

First, the program was not part of scientific research - it was a demonstration of function. Second, many thousands have replicated the method and results without coding in explicit latching. Third, the discussion about a demo program as if it were research is absurd. Fourth, the original code is presumed lost. The point was never about the program, it was about the algorithm and how it demonstrates the effect of selection pressure on random changes.sandlinjohnb_{August 30, 2009
August
08
Aug
30
30
2009
12:00 AM
12
12
00
AM
PDT}

If anyone's interested, Richard Dawkins himself has apparently commented on a recent thread PZ put up related to this contest. RD does comment on PZ's blog from time to time, so I think it's most likely really him. Anyway, RD says he no longer has the original program but that it did not "latch". Not that latching vs. nonlatching is that important in the grand scheme of things, but it's helpful to get confirmation of this detail from the source.yakky d_{August 29, 2009
August
08
Aug
29
29
2009
11:12 PM
11
11
12
PM
PDT}

Tomato Addict at 176: 1. There is no acknowledgement apart from publishing in the combox here. We assume that if you do, you have implicitly entered. If you win, you must send me a mailing address to claim the prize, but it need not be your home and will not be harvested anyway. 2. You can enter as many comments as you like, but you can only be one winner in a given contest. I have been known to declare two winners, in which case both get prizes. 3. I usually judge the entries and my task journal reminds me when. With many entries, there could be delay. When the subject is technical, as in this case, I will seek help, possibly introducing more delay. 4. We are all volunteers here. But yes, it IS a contest. There is a literal stack of books and DVDs right here in my office that will leave individually, when awarded.O'Leary_{August 29, 2009
August
08
Aug
29
29
2009
10:44 PM
10
10
44
PM
PDT}

There was an HTML failure in post #175: I was quoting kairosfocus at the beginning.Tomato Addict_{August 29, 2009
August
08
Aug
29
29
2009
09:29 PM
9
09
29
PM
PDT}

Some administrative questions related to this contest. I have received no acknowledgment of my entry into the contest. Who may I contact about this? When is the closing date for entries, and are multiple entries allowed? How will the winner be determined? It IS a contest, right?Tomato Addict_{August 29, 2009
August
08
Aug
29
29
2009
09:24 PM
9
09
24
PM
PDT}

2] What about quasi-latching? Since we are looking at showcased runs, it is the cases that show no reversions that become what we need to explain. The fact that — since, probabilistic barriers are inherently porous — other runs on the same parameters may show occasional reversions therefore makes no relevant difference. (I used the term quasi-latched to predict such cases, which have ALSO been observed.)
So let me see if I have this straight. Latching (quasi- if you like) is still considered to occur even with occasional reversions. Hence you do not accept my suggestion that a non-zero probability of reversion is non-latching. How about this then: Latching occurs when the probability of moving "up" the gradient is greater than the probability of moving down. This would again seem to include any search other than a blind random walk. ---and--- Joseph wrote> It is about directed vs undirected processes. Is there any search other than a blind random walk that is undirected? That doesn't seem like a useful distinction.Tomato Addict_{August 29, 2009
August
08
Aug
29
29
2009
09:06 PM
9
09
06
PM
PDT}

kf, as expected you have once again not answered my simple questions. But your arguments are becoming more and more entertaining, if this it at all possible! And I am still trying to read all your stuff, so please keep it up. A small history of your arguments: - an algorithm that doesn´t protect correct letters in a search is the same as one that does (the famous implicit latching!) - an algorithm that does a query by replacing every wrong letter with a new random one is the same as one that builds a population (n>1) by copying a parent string with mutations and selecting the best one - you have redefined "query" to mean the determination of the next parent string. Everybody can look at Atoms GUI to see that you are wrong of course. - Somehow the paragraph/pictures/formulas in the D+M paper are only a pedagogical something that not criticizes Weasel directly but is somehow relevant anyway (this argument I don´t really get, I guess I am not alone here...). This obviously contradicts what Demsbi has said on this very website. Dr Dembski started a thread here where he explicitly said that his article criticizes Weasel and therefore evolution. I hope every onlooker now understands that he did nothing of the sort. He discusses a partitioned search which is completely different from the search performed by Weasel. Neither the explanation by Demsbki nor Eq 22 can be mapped onto Weasel and your inability to recognize this or maybe to admit this rather obvious problem is not reflecting well on you. I think your friend Joseph has now understood the difference, I think he has enough patience to explain it to you! So, I have to ask again, just for fun: Which algorithm in Atoms software suite is Weasel and which one the partitioned search? Are they the same? - @Oatmeal Stout Yes while the D+M can be criticized for not representing Weasel correctly that does not mean that the active info concept can´t be applied to Weasel. I am sure that it can be done in a similar way as it is done in the "Random Mutations" paragraph. Therere, however, D+M at least implicitly acknowledge where evolution gets its feedback (or active information) from: From the environment of course. But this is another topic. - @ Joseph It´s not about evolution? Yes, I agree. -- I have a brilliant idea: Why doesn´t Dr. Dembski present *his* Weasel algorithm here or just points us to the relevant code in Atoms suite? Maybe he can even get the prize then?Indium_{August 29, 2009
August
08
Aug
29
29
2009
03:18 PM
3
03
18
PM
PDT}

Dembski and Marks actually do analyze something fairly close to the Weasel procedure in section III-F.2, "Optimization by Mutation." There are several differences from what Dawkins describes: 1. The target is specified over a binary alphabet. 2. The number of offspring is limited to 2. 3. The mutation rate is so low that in most generations both offspring are perfect copies of the parent. The difference in alphabet is trivial. The target is arbitrary, as in the Weasel problem, and may be set to a string of 1's to reveal that the optimization problem is the heavily studied ONEMAX. Solving ONEMAX is easy, yet it is clear in Figure 2 that D&M's simulation runs typically required a great many trials (the horizontal axis should be labeled "trials" or "generations," but the runs are very long in either case). The reason is that the mutation rate is 5 in 100,000 bits, when the length of bit strings is only 100. In other words, few offspring are mutants. From an engineering perspective, appropriate to an engineering journal, it is absurd to perform a stochastic search in which most trials generate no movement in the search space. The simulation does nothing but to give evidence that the simplifying assumptions in D&M's derivation of (28) do not introduce substantive errors, provided that the mutation rate is extremely low. It is worth noting in the caption of Figure 2 that D&M do acknowledge that the decreases in fitness they assume, in the appendix, do not occur actually do occur rarely -- even with an incredibly low mutation rate. It should be obvious, though no one has mentioned it here, that the probability of a decrease in fitness increases with fitness of the parent. When the parent matches the target, the condition for a decrease in fitness is that both offspring be mutants. I believe that D&M would have done better not to restrict the number of offspring to 2, and to have obtained a low probability of decrease in fitness not by setting the mutation rate extremely low, but through a combination of a moderately low mutation rate and a somewhat higher number of offspring.Oatmeal Stout_{August 29, 2009
August
08
Aug
29
29
2009
02:01 PM
2
02
01
PM
PDT}

kairosfocus, With friends like you, who needs enemies? Some of us ID proponents with genuine competence in engineering would like for you to stop with your obfuscation, just as we'd like for Dembski to admit once in a while that he made a mistake. You seem to think there's huge value in the cultural war in presenting Dembski as an inerrant genius. I'd remind you that even the original Isaac Newton spent at least as much of his time on alchemy as he did math and science. In scholarly circles, unlike political circles, owning up to errors is a key part of gaining credence. The truly sad aspect of this run-around is that Dembski and Marks actually do analyze something like Dawkins' procedure in their article. More on this in my next comment. But in the meantime I ask, why are you wasting so much energy on muddying the waters when there is more valuable work to address?Oatmeal Stout_{August 29, 2009
August
08
Aug
29
29
2009
01:09 PM
1
01
09
PM
PDT}

In the Blind Watchmaker video... you do realize that you can see the "correct" letters briefly change, right? In the linked video, around 6:24, you can see the K in METHINKS turn into an N, then an X, then an S, each for a fraction of a second. All of the letters flicker between correct and incorrect. Say what you will about the applicability or accuracy of Dawkin's program, but it obviously doesn't lock the letters once they're right.mindbleach_{August 29, 2009
August
08
Aug
29
29
2009
11:51 AM
11
11
51
AM
PDT}

kairosfocus:
So, once we respect context, we can easily see that the illustration at p 1055 in the IEEE paper is didactic rather than a description of a serious algorithm.
That's a nice theory, but it's belied by the fact that both Dembski and Marks have stated that this algorithm is, in fact, Dawkins' WEASEL algorithm. Dembski has been saying it for years, even after correction. The EIL website still says it.R0b_{August 29, 2009
August
08
Aug
29
29
2009
08:39 AM
8
08
39
AM
PDT}

Indium, The debate isn't about "evolution". It is about directed vs undirected processes. And even the most beneficial mutation has a better chance of becoming lost before it becomes fixed- especially in sexually reproducing populations. But anyway I have the book on order from the local library. I will have it next week and post again at that time.Joseph_{August 29, 2009
August
08
Aug
29
29
2009
07:26 AM
7
07
26
AM
PDT}

KF, I think we can now sum up your claims as follows: 1-> An explicit, required, latching mechanism is the same as non-explicit, non-required, not-always-latching behaviour 2-> A mutation rate that has to be between zero and one hundred percent is the same as a mutation rate that has to be either zero or one hundred percent 3-> A population of one, where no selection can occur is the same as a population of many from which one is selected Now:
many weasels are possible and they give various patterns of behaviour,
Nonsense, many types of search algorithm are possible, they can use many different strategies and mechanisms, and produce different results. But they are all different types of search algorithm. WEASEL is an algorithm defined by Dawkins in The Blind Watchmaker. Many people have written software based on it, and others have made software inspired by it, but which employs different strategies. There is only one WEASEL though and it is described quite clearly by Dawkins. Onlookers will have observed that you have expended a huge amount of text over several threads trying to rhetorically and semantically get around the basic fact that Dawkins algorithm is different that Dembski and Marks. One could be forgiven for thinking that your inability to admit error is pathological. Contrary to your allegations, those of us who keep arguing this point are doing so because it is a matter of fact, not because we are trying to confuse and poison the debate, or because we don't understand the issues, don't read your posts properly or are trying to avoid specifics. The irony of all this is it makes no difference to the content of D and M's paper if they simply removed the reference to The Blind Watchmaker, and replaced it with a correct reference to an actual peer-reviewed paper or an academic volume describing a partitioned random search, rather than referencing a crude pedagogical example in a popular science book. All this work just to avoid admitting that a Weasel is not the same as a Fox!BillB_{August 29, 2009
August
08
Aug
29
29
2009
07:16 AM
7
07
16
AM
PDT}

Prev 1 … 5 6 7 8 9 … 13 Next

You must be logged in to post a comment.

Leave a Reply