Dawkins’ WEASEL: Proximity Search With or Without Locking?

_{William Dembski
March 16, 2009

Darwinism, Evolution, Informatics

3}_{Categories
Darwinism
Evolution
Informatics}

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

On pp. 47-48 of THE BLIND WATCHMAKER, Richard Dawkins gives two runs of his WEASEL program (note that there were typos in both initial seeds — one had 27 characters, the other 29 whereas they should have 28; I’ve corrected that). Here are the two runs using the Courier typeface, which assigns equal width to each character (spaces are represented by asterisks):

WDL*MNLT*DTJBKWIRZREZLMQCO*P
WDLTMNLT*DTJBSWIRZREZLMQCO*P
MDLDMNLS*ITJISWHRZREZ*MECS*P
MELDINLS*IT*ISWPRKE*Z*WECSEL
METHINGS*IT*ISWLIKE*B*WECSEL
METHINKS*IT*IS*LIKE*I*WEASEL
METHINKS*IT*IS*LIKE*A*WEASEL

Y*YVMQKZPFJXWVHGLAWFVCHQXYPY
Y*YVMQKSPFTXWSHLIKEFV*HQYSPY
YETHINKSPITXISHLIKEFA*WQYSEY
METHINKS*IT*ISSLIKE*A*WEFSEY
METHINKS*IT*ISBLIKE*A*WEASES
METHINKS*IT*ISJLIKE*A*WEASEO
METHINKS*IT*IS*LIKE*A*WEASEP
METHINKS*IT*IS*LIKE*A*WEASEL

These runs are incomplete. The first, according to Dawkins, required 43 iterations to converge, the second 64 (Dawkins omitted the other iterates to save space).

As you can see, by using the Courier font, one can read up from the target sequence METHINKS*IT*IS*LIKE*A*WEASEL, as it were column by column, over each letter of the target sequence. From this it’s clear that once the right letter in the target sequence is latched on to, it locks on and never changes. In other words, in these examples of Dawkins’ WEASEL program as given in his book THE BLIND WATCHMAKER, it never happens (as far as we can tell) that some intermediate sequences achieves the corresponding letter in the target sequence, then loses it, and in the end regains it.

Thus, since Dawkins does not make explicit in THE BLIND WATCHMAKER just how his algorithm works, it is natural to conclude that it is a proximity search with locking (i.e., it locks on characters in the target sequence and never lets go).

Interestingly, when Dawkins did his 1987 BBC Horizons takeoff on his book, he ran the program in front of the film camera:

www.youtube.com/watch?v=5sUQIpFajsg (go to 6:15)

There you see that his WEASEL program does a proximity search without locking (letters in the target sequence appear, disappear, and then reappear).

That leads one to wonder whether the WEASEL program, as Dawkins had programmed and described it in his book, is the same as in the BBC Horizons documentary.

In any case, our chief programmer at the Evolutionary Informatics Lab (www.evoinfo.org) is expanding our WEASEL WARE software to model both these possibilities. Stay tuned.

Comments

kairosfocus, On second thought, I'll deal with the proper name issue now. There's a delay in my postings. (I have been in moderation for a testy comment I made to someone else. I hope I'll get removed from moderation soon.) With regard to ATBC: It's. Not. My. Board. I'm not going to ask Wes to do anything. I have never thought the proper name issue amounted to beans. I've removed that proper name from a blog I control, but that was on a forum I controlled and as a favor to the person who asked. I'm certainly not going to ask another person to change the practice on his board when I find the demand silly. Further, your justification does not hold water:
the reason I have asked that my name not be used: it causes spikes in email spamming
Wow: powerful spammers. When I send an email to a proper name, my emails get bounced back, unless I also use something called a . . . what's the term . . . oh yes: an email address. Emails go to addresses, not names. Further, most spam addresses are gathered by bots, not readers. They look for address-type things (like the @ sign), not for proper name-type things, which are useless. Finally, spam filters are available freely in most email programs and are very effective. I may get hundreds of spam messages every day; I see almost none of them, however, because my spam filter works. You should try one!David Kellogg_{March 21, 2009
March
03
Mar
21
21
2009
07:15 AM
7
07
15
AM
PDT}

Joseph
The way Dawkins describes cumulative selection and the way he uses it in the “weasel” program, cumulative selection is a ratcheting process.
As you have the book at hand, could you provide a quote that supports this?George L Farquhar_{March 21, 2009
March
03
Mar
21
21
2009
05:53 AM
5
05
53
AM
PDT}

nd, FYI a personal name is NOT public domain information: you nor anyone does not have a right to take it and use it as it if you were me, regardless of how you come by it, or to subject me to harassment or worse — that is called identity theft sir, or worse than mere identity theft.
I took a peak over there. No one is pretending to be you.crater_{March 21, 2009
March
03
Mar
21
21
2009
05:51 AM
5
05
51
AM
PDT}

hi, there seems to be a bug in the board, I post but nothing appears and no message about moderation either.George L Farquhar_{March 21, 2009
March
03
Mar
21
21
2009
05:49 AM
5
05
49
AM
PDT}

kairosfocus, I'm not going to deal with the proper name issue (which I think is your favorite red herring) until you deal with Mr. Elsberry's math. As for the rest:
4 –> We will readily see from such a thought expt., that reasonably uncorrelated sampling [i.e. credibly unbiased sampling] can give a reasonably good picture of a population or a trend, surprisingly quickly.
The runs you have seen are not, and have never been presented as, an unbiased sampling. They have been the best fit from an undetermined number of mutated versions. My example of the asexually reproducing weasel is more like the Weasel program than your appeal to dice-throwing.David Kellogg_{March 21, 2009
March
03
Mar
21
21
2009
05:48 AM
5
05
48
AM
PDT}

Now that I have a copy of "The Blind Watchmaker and have read the relevant chapters, I can say the following in full confidence: The way Dawkins describes cumulative selection and the way he uses it in the "weasel" program, cumulative selection is a ratcheting process. And it is ratcheting towards a specified target.Joseph_{March 21, 2009
March
03
Mar
21
21
2009
05:48 AM
5
05
48
AM
PDT}

the reason I have asked that my name not be used: it causes spikes in email spamming.
Could you explain exactly how this happens?
FYI a personal name is NOT public domain information:
What case law are you citing here?
nor anyone does not have a right to take it and use it as it if you were me
Nobody that I can see is "pretending to be you". In comment 135 you said
Mr Elsberry is grossly wrong
You used his name. He used your name. Why is it identity theft when he uses it but OK when you do?
So, kindly stop pretending to innocence and stop trying at rhetorical games to “justify” the indefensible.
If you explain how simply using your name is indefensible (when you feel free to use other peoples name!) and how it causes more spam (exactly how, not just a claim that it does) perhaps you'll get your way. Stamping your feet alone will never do so!George L Farquhar_{March 21, 2009
March
03
Mar
21
21
2009
05:37 AM
5
05
37
AM
PDT}

Seversky: And in so writing a few -- pardon -- weasel words -- the major issue ofer Weasel as yet another a misleading icon of ecvo mat is dodged: the BOOK, the BLIND Watchmaker is about the idea that we do not need design to get to complex info-rich bio-functionality. But a chief illustration in Ch 3 is . . . intelligent design at work. AND that in a context that Mr Dawkins explicitly confesses that if the realistic functionality threshold (even for a toy of a 28 letter phrase) had been imposed, Weasel could not have worked. So, nonsense -- non functional -- phrases are rewarded on mere proximity to target. Thus, Weasel is precisely a question- begging misleading strawman argument as I have pointed out. Weasel is not educational, it is indoctrinational and obfuscatory, begging the central question it suggests that it answers. Do you begin to understand why I despise rhetorical games? GEM of TKI PS: YOU KNOW THAT THERE IS A BIG DIFFERENCE BETWEEN INITIALS AND A NAME. THERE IS UTTERLY NO EXCUSE FOR THE SORT OF ATTEMPTED RETALIATORY OUTING AGAINST WHISTELBLOWING THAT ANTIEVO AND MR ELSBERRY HAVE INDULGED, PERIOD. (And, as a long time participant here, you KNOW or should know the reason I have asked that my name not be used: it causes spikes in email spamming. That my name may be accessed by search in a low traffic site of the Internet [for purposes of responsibility over authorship] is no excuse for putting it up in a high traffic site to be accessible without looking for it, and in effect inviting all and sundry to launch spammming attacks or worse. And, of course there is the obvious issue that evo mat advocates routinely resort to outing and persecution. And, FYI a personal name is NOT public domain information: you nor anyone does not have a right to take it and use it as it if you were me, regardless of how you come by it, or to subject me to harassment or worse -- that is called identity theft sir, or worse than mere identity theft. So, kindly stop pretending to innocence and stop trying at rhetorical games to "justify" the indefensible. That is enabling behaviour. SHAME ON YOU!)kairosfocus_{March 21, 2009
March
03
Mar
21
21
2009
03:47 AM
3
03
47
AM
PDT}

kairosfocus @ 135
In short, Weasel is plainly and self-confessedly a designed, targetted search that avoids the issue of complexity of first function in bio-systems, and so is precisely not a BLIND watchmaker in action.
How many times does this have to be repeated? We have the author's word that the WEASEL program was written for one purpose and one purpose only: as an illustration of the fact that cumulative selection can reach a target much, much faster than a purely random search. Nothing more, nothing less. As for the use of your name, you use your own initials in the sig on each of your posts and you yourself have placed your name in the public domain by publishing it on your website. It is absurd to expect others to pretend they don't know what you have made no attempt to hide.Seversky_{March 21, 2009
March
03
Mar
21
21
2009
02:34 AM
2
02
34
AM
PDT}

I have specifically requested that my proper name not be used online, to avoid spam. Thankfully, Mr Elsberry et al are in no position to damage my career by trying an “outing” game. And that is their plain and routine intent, let us not fool ourselves. SHAMELESS!
Haven't we been through this before? If you are concerned about your name being on the internet, perhaps you should consider that it is plastered all over your always linked.crater_{March 21, 2009
March
03
Mar
21
21
2009
01:28 AM
1
01
28
AM
PDT}

On Mr Kellogg, re 129: First, on personalities: Mr Elsberry has NOT got permission to use that proper name, and his abuse of that in the teeth of to all but moral certainty of knowing that he should not do so shows just how out of order, disruptive and uncivil the Anti Evo site is. I therefore request of you, Mr Kellogg, on basic duties of care, that you kindly inform Mr Elsberry et al of their misbehaviour and request that he take down the offending use of a personal name. If you do not do so, that would simply show enabling behaviour on your part. Especially, after you tried to ask me to go there to participate in that site's exchanges. As to the rhetorical assertion that 300 rolls of a loaded die are irrelevant to showing an underlying trend [my obvious point but tot he willfully obtuse], the above thought example at 133 just above, point 3, will show why Mr Elsberry is grossly wrong. But, too: on the evidence and presumption that he is a properly qualified biologist and/or has a duty to inform himself on relevant and reasonably accessible facts before making critical commentary, he is -- sorry to have to be so direct -- being deceitful by negligence or by intent. If he -- as I gather -- is a biologist, he MUST have done enough statistics to be familiar with the issue of obtaining sufficiently large but conveniently small samples; i.e the issue of the law of large numbers. In the case of Weasel circa 1986, the published 300 samples of letter behaviour are quite enough to show a very strong trend indeed to latch. Not only so, but there are two reasonable mechanism that would explain that: T2 -- explicit latching, and T3, implicit. On the latter, say 5% probability per letter per member of a generation [ = 0 - 2 changes, typical for 28 letters; 24% being at 0] and a sufficiently large but moderate generation size [so that multiple mutations -- way out in the skirts -- do not show up repeatedly and dominate behaviour] with letter by letter reward of targetting will produce just what is observed beyond reasonable dispute: (i) letters latch on going correct, and (ii) "no change" dominates over 50% of generations. Otherwise, (iii) the convergence to the target would be much faster than the reported 40+ and 60+ generations, which we can presume were for "good" runs. As well, (iv) we may note that 40+ and 60+ as "good" [= fast] runs are very consistent with the median number for latched searches, 98 generations. (Note, too: as pointed out from the previous thread, latching may be IMPLICIT, not just explicit.) In short, we have a very good model of what is going on, even on this secondary point. On the main issue, it stands clearly demonstrated from Mr Dawkins' own mouth in BW ch 3 [as cited in his attempted defense by Wiki . . . another inadvertent admission against interest], that Weasel is a question-begging strawman argument that avoids Hoyle's challenge to get TO shores of realistic functionality before playing with hill-climbing to optimality of function by RV + NS. Citing Wiki from BW ch 3:
I don't know who it was first pointed out that, given enough time, a monkey bashing away at random on a typewriter could produce all the works of Shakespeare. The operative phrase is, of course, given enough time. Let us limit the task facing our monkey somewhat. [Ducks the issue of complexity . . . observed bio-function starts out at FSCI of 600 - 1,000 k bits on DNA alone; config spaces that start at about 10^180,000 cells . . . ] Suppose that he has to produce, not the complete works of Shakespeare but just the short sentence 'Methinks it is like a weasel', and we shall make it relatively easy by giving him a typewriter with a restricted keyboard, one with just the 26 (capital) letters, and a space bar. How long will he take to write this one little sentence? . . . . [if he has to produce a toy scale fucntionality of the phrase above, he is looking at 27^28 ~ 10^40 configs; which as the EIL page on Weasel shows, effectively leads to endless non-functional wandering around in the letter space . . . i.e Dawkins credibly KNOWS that he real task is hopeless and instead of re-thinking, sets it to one side, begs the question posed by Hoyle and sets up a strawman instead] We again use our computer monkey, but with a crucial difference in its program. It again begins by choosing a random sequence of 28 letters, just as before ... it duplicates it repeatedly, but with a certain chance of random error – 'mutation' – in the copying. The computer examines the mutant nonsense phrases [non-functional configs are rewarde3d by passign on to the next manufactured genration of lemons], the 'progeny' of the original phrase, and chooses [i.e. design and decision] the one which, however slightly, most resembles the target phrase, [non-functional increments in proximity to target down to one letter are rewarded] METHINKS IT IS LIKE A WEASEL . . . . The exact time taken by the computer to reach the target doesn't matter. [Oh yes it does! A realistic toy search is infeasible. How much moreso the real challenge that Hoyle posed.] If you want to know, it completed the whole exercise for me, the first time, while I was out to lunch. It took about half an hour. (Computer enthusiasts may think this unduly slow. The reason is that the program was written in BASIC, a sort of computer baby-talk. When I rewrote it in Pascal, it took 11 seconds.) Computers are a bit faster at this kind of thing than monkeys, but the difference really isn't significant. 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 [telling word, again!] if it were forced to use the other procedure of single-step selection [i.e he knows he is ducking the challenge of threshold of complexity to get to initial function -- you have to land on the beach of an island of function before you can climb its hills to optimal performance]: about a million million million million million years. This is more than a million million million times as long as the universe has so far existed. [translation: a realistic case being probabilistically implausible, I have diverted to a simplistic case that seemingly supports my view, instead of revising my thoughts in light of what the probabilities are trying to tell me . . . ]
In short, Weasel is plainly and self-confessedly a designed, targetted search that avoids the issue of complexity of first function in bio-systems, and so is precisely not a BLIND watchmaker in action. All attempts to "justify" Weasel therefore amount to so much kicking up of sand into the eyes of onlookers, and to then leading naively trusting, blinded onlookers away from the truth. A more honest response would be to acknowledge the truth about how misleading an icon of evolution Weasel and its kin are, and to frankly apologise for having misled the public. That such is plainly not forthcoming is all too telling on the utter bankruptcy -- intellectual and moral -- of the current state of the evolutionary materialist paradigm. So, onlookers, let us draw our own conclusions for ourselves, and act in the defense of our true interests, before it is too sadly late for our civilisation. GEM of TKI PS: I have specifically requested that my proper name not be used online, to avoid spam. Thankfully, Mr Elsberry et al are in no position to damage my career by trying an "outing" game. And that is their plain and routine intent, let us not fool ourselves. SHAMELESS!kairosfocus_{March 21, 2009
March
03
Mar
21
21
2009
12:55 AM
12
12
55
AM
PDT}

8 --> That is, Mr Dawkins made an inadvertent declaration against interest, the strongest -- most likely to be true -- form of testimony in forensic situations. The fierce and sometimes outright uncivil effort put up by Darwinists ever since, including in recent days here at UD, to take this back or to cloud it, is telling on the force of the admission.] 9 --> Now, there is no good reason to believe that the Weasel type program, whether it is explicitly or implicitly latched or whether it uses more modern GA approaches, will be such that a sample grid of every tenth or so generation's champion will be systematically unrepresentative of the overall trend of the program. Indeed, it amply illustrates the steady march to the target that Mr Dawkins so triumphantly published in 1986. 10 --> Just a little too much so. For, the program circa 1986 also showed a strong tendency to "run" once a letter becomes correct, event hough it is not functional, i.e "nonsense phrases" are being blatantly rewarded for increased proximity to the target, so much so that a sample of 300 from "good" runs of the program, shows 200 examples of latching. 11 --> So, the sampled latching of the output, a secondary feature [and which can be done explicitly or implicitly], highlights the more troubling underlying problem: Weasel is not BLIND watchmaker search. 12 --> instead, it is targetted, directed search based on in effect broadcasting the location of the target by rewarding "warmer" search-points, even when they are not yet functional. And that is what I highlighted last December, which GLF quotemined to try to threadjack a previous thread that was heading towards some unpleasant truths on the implications of selective hyperskepticism:
[Unpredictability thread, slice from no 111:] Weasel sets a target sentence then once a letter is guessed it preserves it for future iterations of trials until the full target is met. That means it rewards partial but non-functional success, and is foresighted. Targetted search, not a proper RV + NS model. [Cf. from 107 this post referred:] . . . the problem with the fitness landscape [model] is that it is flooded by a vast sea of non-function, and the islands of function are far separated one from the other. So far in fact . . . that searches on the order of the quantum state capacity of our observed universe are hopelessly inadequate. Once you get to the shores of an island, you can climb away all you want using RV + NS as a hill climber or whatever model suits your fancy. But you have to get TO the shores first. THAT is the real, and too often utterly unaddressed or brushed aside, challenge. [NB: And, in fact, that was the challenge Sir Fred Hoyle had posed. A challenge that Weasel, from its outset, has ducked and distracted attention from. Weasel is -- and has always been -- a question-begging strawman fallacy.] And, I repeat, that [unmet cfhallenge] starts with both the metabolism first and the D/RNA first schools of thought on OOL. As indeed Shapiro and Orgel recently showed . . . . As for Weasel, you will note that I hardly bothered to take note of it once it was raised by JK, as it is so trivially irrelevant as a plainly DIRECTED, foresighted tartgetted search. It instantiates intelligent design, not the power of RV + NS. Even going up to Avida and the like, similar issues come up, as is highlighted under the issue of active information by Dembski and Marks.
____________ In short, right from the beginning, it should have been clear what the key problem with Weasel was. And, the law of large numbers shows just how the Weasel o/p circa 1986 is inadvertently telling on the underlying problem. And no amount of after the fact distractive selectively hyperskeptical rhetoric by evo mat advocates will change that fact. If you showcase an example meant to illustrate your point, it is entirely proper for others to analyse the case and show the inadvertent admission against interest that lurks therein. GEM of TKI PS: And, I repeat: GLF's attempt to play the "$100 K challenge not taken up" fallacious argument, simply shows the bankruptcy of the evo mat case on this.kairosfocus_{March 20, 2009
March
03
Mar
20
20
2009
11:53 PM
11
11
53
PM
PDT}

Mr Kellogg: Do you have any knowledge of the law of large numbers? In effect, there are many phenomena in the world that have indefinitely many possible observations. So . . . Q: How can we infer confidently to patterns in nature (or technical, or social situations), then, if we are expected to have a large fraction of the possible outcomes in hand? ("A large fraction" is plainly impossible and/or inaccessible for finite, & fallible investigators and/or decision- makers and/or investors constrained by opportunity costs or access etc. [E.g. the financial statements of firms are very carefully massaged indeed. Sometimes, just a little too carefully.]) A: Several points -- and points that tie directly to the under-appreciated significance of the explanatory filter that a leading ID researcher, namely, a certain Mr Wm A Dembski, has championed (and as has recently been incrementally improved by explicitly focussing on aspects): 1 --> First, things that are dominated by dynamical forces [i.e. expressible in differential and/or difference equation-based models, e.g. Newtonian translational and/or rotational dynamics, or electrical dynamics, or micro or macro economics, etc . . . ] will as a rule show regularities rather quickly and we will be able to see that the patterns characterised by relevant laws or models are reliable. (Of course there usually is some scatter, which we deal with as noise. |Graphical techniques, such as log-log or lot-linear or linear plots, time series plots [esp. three-sigma plots] box- whisker plots, etc are very often used to help in what is now often called exploratory data analysis.) 2 --> There are other systems that are highly contingent, i.e. outcomes vary significantly under similar initial circumstances; perhaps stochastically, perhaps under direction, perhaps in part by both. 3 --> Where the law of large numbers fits in is that samples that we have reason to believe are credibly uncorrelated to the system -- i.e. are unbiased -- soon enough tend to track the pattern of behaviour. For instance:
A bell chart test case: imagine a large bell-shaped curve on a chart on a sponge backing laid the ground that has been sliced into equal width strips, and put back together in order. Drop a dart more or less at random on it from a height repeatedly, so that the odds of landing at any particular point are more or less even. The holes will tend to be at fairly evenly and randomly scattered points, but will be more or less evenly distributed. After 1 drop, obviously this is not going to be evenly representative. After 10, we will probably begin to see a broad pattern of scatter. After 20 - 30, we usually will have a fairly good scatter, but it is unlikely that there will be a close enough gridding that we will see many points from the "tails" of the curve. After 300 the density of scattered points will be much finer, and we are likely to pick up more and more points in the tails. Thus, a count of the points by strips is a good representative of their area. [In fact, if you ever need to estimate a very irregular area, putting it on a rectangular chart of known overall area and doing the dartboard exercise then counting points in the area vs in the whole rectangle will be an excellent quick and dirty solution.]
4 --> We will readily see from such a thought expt., that reasonably uncorrelated sampling [i.e. credibly unbiased sampling] can give a reasonably good picture of a population or a trend, surprisingly quickly. 5 --> In the case in view, as I earlier described, we have two runs of a process, and we have no reason to think that the patterns int eh runs are materially different. indeed, i took a ration that showed a fairly stable ratio of the latched points to the points sampled that could change. 6 --> I therefore simply pooled, and we can see that there are 300+ points where letters can change, and of these 200+ show "correct" letters, mostly in non-functional configs, marching ever closer to the target. In NO case does a correct letter revert. Indeed, that trend of runs once a letter becomes correct is a dominant characteristic of the output. 7 --> Your objection now is in effect that he reports are on every tenth generation, not every generation and/or every member of every generation. This is of course the data we have in hand, which we can take to be Mr Dawkins' publication of what he thought circa 1986 were "good" runs. [NB: The fact that subsequently, from 1987 on, emphasis has been laid on how such Weasel-type outputs do not show such strong runs in every case, tells us that Mr Dawkins plainly misjudged the situation circa 1986. [ . . . ]kairosfocus_{March 20, 2009
March
03
Mar
20
20
2009
11:53 PM
11
11
53
PM
PDT}

ericB: I apologize for mischaracterizing your post in 122. The first two paragraphs looked very much like you were saying both the Dawkins and Zachriel algorithm suffered from combinatorial explosion, but I did no realize you only meant in a hypothetical sense. Then you wrote the following: *** If and when this can be done, that could indeed escape from the doom of the combinatorial explosion, provided that there are no discontinuities in function too large for the random variations to hop across in reasonable time. The greater the discontinuity in function, the less probable a random variation will discover the next island of functionality. *** So evidently you were allowing evidently that the Zachriel algorithm might work. Also your comment about proximity to distant targets was only hypothetical as well, and I did not pick up on that either. So either your post was ambiguous or I need to improve my reading skills. So again, I apologize. JTJT_{March 20, 2009
March
03
Mar
20
20
2009
03:04 PM
3
03
04
PM
PDT}

SteveB,
You’re describing animal husbandry, not blind watchmaker evolution.
I'm describing something that's pretty close to the weasel program: that is, a targeted search. In what ways is my example not like the Weasel program? And do those ways suggest that Weasel is locked or, rather, that kairosfocus is wrong?David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
01:40 PM
1
01
40
PM
PDT}

David:
Let’s try a thought experiment... let’s say I want to breed a weasel that has certain targeted features... (my emphasis)
Let’s try another thought experiment: how about we stop ascribing to natural selection anthropomorphic characteristics that it doesn’t have. You're describing animal husbandry, not blind watchmaker evolution.SteveB_{March 20, 2009
March
03
Mar
20
20
2009
12:46 PM
12
12
46
PM
PDT}

kairosfocus, I know you object to ATBC and the Panda's Thumb, but do you object to a mathematical explanation of the probabilities involved? Wes E. has provided one. Now, you may object to that post based on its use of a certain proper name, but do you object to the math?David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
10:12 AM
10
10
12
AM
PDT}

In the event of quibbles, let's say that the two targets are measured by 28 different variables, including (for claws) such things as differences between front and rear claws, sharpness, thickness, weight, hardness, ability to extend, etc., and (for fur) thickness, softness, length, glossiness, consistency on different parts of the body, color uniformity, etc. On all of these measures, the weasel closest to the target measures as good or better than the weasel ten generations (that is, 500 weasels) previous. Does this mean that the weasels never produce offspring that are farther from the target?David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
09:06 AM
9
09
06
AM
PDT}

Let's try a thought experiment. I have an actual weasel. It's ugly and undistinguished, but in one respect it's special: it can reproduce asexually. It will produce 50 offspring, some with little differences, that are capable of reproducing in the same way. Now, let's say I want to breed a weasel that has certain targeted features: a soft, glossy coat and short claws. I get my weasel to produce 50 offspring. I pick the one that's closest to the target and euthanize the other 49. I do this over and over again, over sixty times. Each time, I produce 50 offspring from the best weasel and kill off the 49 I don't want. By the 63rd generation, I have a pretty nice weasel: its coat is beautifully soft and glossy, and it has tiny, tiny claws. Along the way I did something else: every tenth generation, I stuffed the best weasel after he'd produced his offspring. I keep the original and these selections, in order, in glass cases, to show the overall progress of my breeding program. One day, my friend kairosfocus comes to my house and observes the cases. He says, "there's something odd here. From what I can see, the weasel offspring always have softer coats and shorter claws than their parents. In truth, your weasel will never produce offspring with rougher coats and longer claws." What's wrong with that conclusion?David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
08:40 AM
8
08
40
AM
PDT}

kf, actually you have two samples from two runs: the first sample has 196 characters, and the second has 224 characters. In that sample, I observe that correct letters in the first run have a total of 88 chances to flick back, and correct letters in the second run have a total of 135 chances. For a sample to be representative, it has to be random. These are nonrandom, because they represent the best fit. You are using highly nonrandom data to say what happens generally. Further, note that calling these "runs" is incorrect. They are usually 10 generations apart, and so you don't know what happens from one generation to the next. Moreoever, you don't know what happens with all the mutations that were not chosen to parent the next generation: you merely the best fit 10 generations down the line. Between generation 10 and 20, a "best fit" could involve one letter becoming incorrect at generations 11, 12, 13, etc., but then reverting back by generation 20. But you don't see any of this. How big is the population? If the population of "tries" in each generation is -- just a guess -- 50, and the total generations to target in the presented samples is (if I recall correctly) a little more than 60 in each case, then the total population of potentially changing characters would be 88,200 (28 characters x 50 tries x 60 generations). From that population you've got at most 88 or 135 characters that have the same correct letter when you see a nonrandom (best fit) sample several (in most cases, 10) generations apart. And from this you infer locking?David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
07:56 AM
7
07
56
AM
PDT}

EricB: I thought you were saying the weasel algorithm and Zachriel Algorithm suffered from combinatorial explosion (Let anyone else read the first two paragraphs of 122 and see if that was a fair conclusion). Evidently you were not saying that though. Your basic point was that Weasel and Zachriel were both cheating.JT_{March 20, 2009
March
03
Mar
20
20
2009
07:47 AM
7
07
47
AM
PDT}

The core issue there, as it is with WEASEL or Zachriel or any other such program, is that a runaway combinatorial explosion of possibilities will bury any hope of success. As the length grows, it quickly becomes unfeasible. So, the key is all about how to escape this combinatorial explosion.
The WEASEL algorithm doesn't suffer from combinatorial explosion either - how did you not know that?
The evolutionary claim is that this bullet can be dodged by selecting between randomly accessible variations on the basis of functional advantages that give preference for reproduction.
So what you seem to be saying is that both the weasel algorithm and the Zachriel algortihm sufffer from combinatorial explosion (not true for either of them), but that there is an evolutionary claim with no evidence that this combintaroial explision is avoided in nature, whereas it could not be in the algorithms above. You are just completely confused. Neither Zachriel or Weasel or combinatorially intractable, and Zachriel is an improvement over the Weasel Algorithm in that the former does not use proximity to a distant target, but rather only local data. Please try to learn something.JT_{March 20, 2009
March
03
Mar
20
20
2009
07:25 AM
7
07
25
AM
PDT}

ericB:
The core issue there, as it is with WEASEL or Zachriel or any other such program, is that a runaway combinatorial explosion of possibilities will bury any hope of success. As the length grows, it quickly becomes unfeasible. So, the key is all about how to escape this combinatorial explosion.
In [110] I asked you, BTW - it occurs to me to ask whether you’ve actually reviewed those websites I mentioned previously [57], since you’re discussing this. You didn't answer, but I'll assume its NO, judging from your remark above. You're just stating things for effect, with no knowledge or desire to learn anything. In a previous post I explained the following about the Zachriel "mutagenator" algortihm: For just a seven character word this incremental method is several thousand times faster than blind chance (because of regularities in the english language landscape). And incredibly the longer the word is, the more improvement there is over blind chance. It very quickly becomes 200,000 times faster for say a 10 character word than a 26^10 blind search.
This is why the issue of how one represents function is pivotal. To paraphrase kairofocus (114), it is illegitimate to reward on the basis of target proximity (i.e. similarity) when the string is not yet functional, i.e. not functioning in its current context.
How many times do I have to tell you, the Zachriel Algorithms do not reward proximity to a distant target. The mutagenator program starts with 0 letters and then find the shortest word of 1 letter then the find the next longest word randomly and then the next. and then the next. Only legal words are kept, and the word is continually lengthened this way. The phrasenator algorithm finds longer and longer phrases using the same method. There is no distant target that the algorithms are approximating. Just finding the next longest word randomly allows you to get increasingly longer words (or phrases) because of the nature of English. Why would you not even bother to go to the websites and spend at least 10-15 minutes carefully reading that material there.
It is not the ideas that are objectionable. (See my affirmations in this note.) It is the fact that an appearance of success is being generated by programs that are committing illegitimate moves. That is always objectionable. That is what makes it an illusion, i.e. not what it appears to be.
You said previously that the Zachriel algorithms suffered from "runaway combinatorial explosion". Now you're saying they're cheating and creating the illusion of success. You're just flatly contradicting yourself and don't even care.JT_{March 20, 2009
March
03
Mar
20
20
2009
07:10 AM
7
07
10
AM
PDT}

JT (111, 112):
"I personally want to write a program that attempts to reconstruct an entire text using only random incremental additions of words from that text. It seems obvious that should work though."
Depending on the particulars, if "work" means "finish before you have died" then it is not so obvious that it will work. The universe would suffer heat death before random typing would produce one sonnet of Shakespeare. So the devil is in the details. The core issue there, as it is with WEASEL or Zachriel or any other such program, is that a runaway combinatorial explosion of possibilities will bury any hope of success. As the length grows, it quickly becomes unfeasible. So, the key is all about how to escape this combinatorial explosion. The evolutionary claim is that this bullet can be dodged by selecting between randomly accessible variations on the basis of functional advantages that give preference for reproduction. *** If and when this can be done, that could indeed escape from the doom of the combinatorial explosion, provided that there are no discontinuities in function too large for the random variations to hop across in reasonable time. The greater the discontinuity in function, the less probable a random variation will discover the next island of functionality. *** This is why the issue of how one represents function is pivotal. To paraphrase kairofocus (114), it is illegitimate to reward on the basis of target proximity (i.e. similarity) when the string is not yet functional, i.e. not functioning in its current context.
JT: You seemed irritated or antagonistic towards these ideas, but to me they were at least interesting if not compelling.
It is not the ideas that are objectionable. (See my affirmations in this note.) It is the fact that an appearance of success is being generated by programs that are committing illegitimate moves. That is always objectionable. That is what makes it an illusion, i.e. not what it appears to be.ericB_{March 20, 2009
March
03
Mar
20
20
2009
06:12 AM
6
06
12
AM
PDT}

Onlookers: Re Mr Kellogg on sampling and runs. We must observe that the above 1986 Weasel shows a sample of 300, with 200 being in runs [thus a dominant feature of the output . . . and being of so large a number odf samples that if there is a reasonable chanve of a flick-back it should appear at least once], where never once a run appears do we see reversion away from the run. This contrasts with the 1987 outputs, which show frequent winks away from the correct letter. These observed stable characteristics of the processes -- whether explicitly or implicitly latched -- tell us a lot. So, again, we are simply looking at selective hyperskepticism. Perhaps, we should put it this way: Mr Kellogg you are in a dice game. Somehow the roll keeps on coming up 6's 2/3 the time. After 200+ out of 300+ rolls, are you going to say maybe this is not a loaded die? [If so, can we meet for a little dice game; I could do with a fatter bank account.] Onlookers, see the point? GEM of TKI PS: Joseph, it is not just large enough. If the generations were really big, the upper skirts would show up often enough that we would see very r5apid convergence as 2's and 3's etc of newly correct letters would win the test for best of generation. the factt hat good runs are hitting up at 40 - 60 tells us the generations are of a moderate not very large size. For, 0 changes is plainly winning half or more of the time per generation.kairosfocus_{March 20, 2009
March
03
Mar
20
20
2009
05:18 AM
5
05
18
AM
PDT}

I will be picking up a copy of "The Blind Watchmaker" today. However as I have already stated the program does indeed latch -input to output- via probabilities. That is given a target, a small enough mutation rate and and large enough sample size, the output will NEVER be farther away from mthe target than the parent. IOW latching, given the proper conditions, is inevitable. So why do David and George ignore that fact?Joseph_{March 20, 2009
March
03
Mar
20
20
2009
04:32 AM
4
04
32
AM
PDT}

kf,
The published 1986 Weasel output latches beyond reasonable dispute [testified to by the attempt to assert that 200 out of 300 letters with no counter instances is a small sample],
No. It is certainly not beyond reasonable dispute, the relevance of the sample size depends on the size of the total population, which you don't know.David Kellogg_{March 20, 2009
March
03
Mar
20
20
2009
04:02 AM
4
04
02
AM
PDT}

Onlookers: Re GLF's latest. 1] The published 1986 Weasel output latches beyond reasonable dispute [testified to by the attempt to assert that 200 out of 300 letters with no counter instances is a small sample], which is what I remarked on as a telling secondary feature, back in December. (Just look at the above in this thread's original post, and think through the 200 of 300 letters sample.) 2] When the issue of explicit vs implicit latching was raised, I proposed T2 and T3 as explicit vs implicit latching, and stated the grounds for preferring the former in the first instance. 3] On subsequent additional testimony, despite the onward issue of winking I am willng to accept that the best current explanation for the observed o/p latching behaviour is implicit latching and have again summarised how it its reasonably likely to be done. 4] GLF knows or should know that once we have any reasonably large sample that is not likely to be structurally unrepresentative, it will most likely reflect the population as a whole [i.e. the valid form of the layman's law of averages: typical behaviour of big enough samples will usually be close to typical behaviour of the whole population]. So oncewe are in the realm of law of large nuimbers, i.e 20 - 30+, we can be comfortable that a sample will give a good enough look at the population to be significant. of course getting bigger and bigger samples increases precision, and 300 is ten times the typical large enough numbers threshold. 5] Indeed, we know that Mr Dawkins put forth the showcased 1986 output as this he thought made a good case for him. he evidently did not realise the significance of the strong latching tendency. 6] It is therefore notewoerthy that within a year or so, the 1987 videotaped runs strongly show frequent winking not latching. This strongly suggests, per inference to best explanation again, that the parameters were detuned to get away from obvious output latching. I have no reason for apologies to be made and so make none. Those who have since the 1980's ROUTINELY used Weasel as if it were good evidence for the capacity of the blind watchmaker of natural selection have grossly misled large swathes of the public through yet another misleading icon of evolutionary materialism, and DO have apologies that are due -- indeed, long since over-due. None have been forthcoming, but distractive side-debates are being posed. No prizes for guessing why, onlookers. GEM of TKIkairosfocus_{March 20, 2009
March
03
Mar
20
20
2009
02:19 AM
2
02
19
AM
PDT}

Wesley R. Elsberry
The graph shown here plots the probability that a population will have at least one candidate that preserves all the correct characters from the parent string. The graph shows population sizes from 1 to 500, mutation probabilities from 0.0 to 1.0, and is done for the case where the number of correct characters in the parent is 27. Once the population is around fifty, increases in population size make very little difference.
http://www.antievolution.org/cgi-bin/ikonboard/ikonboard.cgi?act=ST;f=14;t=6034;st=#entry140135 What's your counter Kariosfocus? Why not jump onto that thread, tell Wesley why he is wrong? I've seen no such graphs here, it's not possible to reproduce them here. So that conversation cannot happen here.
I refuse to have anything to do with these notoriously disruptive and destructive fever-swamp of evo mat sites
Is this the same reason you refuse to enter the "peer review game"? What could be disruptive or destructive about you proving somebody wrong with math? I'm sure the good folks at ATBC will be nice to you. If you can defend your arguments with data and logic anyway.George L Farquhar_{March 20, 2009
March
03
Mar
20
20
2009
02:00 AM
2
02
00
AM
PDT}

KariosFocus
Mr Dawkins did not EXPLICITLY latch.
You were wrong. I was right. Was that so hard to say? Of course, "EXPLICITLY" gives the lie to your "apology". It's really a "I was still right". Yet when I mention that the probability of latching behaviour being seen is a topic of dicussion at the links I gave you don't want to know! You'd rather proclaim victory without having fought any wars!
If there is any merit in their arguments, GLF must present them here — without the nasty ad hominems, strawmen and red herrings
I would note that this thread http://www.antievolution.org/cgi-bin/ikonboard/ikonboard.cgi?s=49c357ba0cc9c9ba;act=ST;f=14;t=6034 Has been created for discussion of Evolutionary Computation and Evolutionary Computation alone. Any ad hominems, strawmen and red herrings will no doubt be sent to the "bathroom wall" where off topic posts get sent. Therefore, you have nothing to worry about, you delicate sensibilities will remain unbruised. The fact of the matter is that this blog is not the best place to discuss such matters. How can graphs be posted? How can formulas reliably be shown? This format does not lend itself to such a dicussion.
GLF’s turnabout attempt backed up by refusal to address the brute fact of a sample size of 300+ letters with 200+ showing the secondary phenomenon he has chosen to debate over in order to get away from the implications of his wager argument attempt, fails.
Do you often draw hard and fast conclusions based on 5-10% of a sample size? Conclusions that have caused the creation of mutiple threads on mutiple blogs about this issue? I'm not "gettin away" from any implication of my "wager argument". I asked for a quote from Dawkins where he says latching was used. You still have not been able to provide such. Therefore you lose. Is that so hard to understand? Kariosfocus
GLF is refusing to accept that a sample size of 300+ in a conte4xct of “good runs” being published, is not a small one by the law of large numbers standard.
If we agree, for the sake of argument, that the sample size published was 300+, could you tell me what that was from? I.E. 300+ samples out of a possible N? What is N? 301? 3001? 30001? If you are not too scared of engaging with people who know what they are talking about, we can contine this at the ATBC thread linked to above. I look forwards to seeing you there. Why not prove everybody there wrong (after all, this is math and you can prove things in maths! If you are right, show your working!) and you can come back here and crow about it? About how you beat the Darwinists on their own patch.George L Farquhar_{March 20, 2009
March
03
Mar
20
20
2009
01:53 AM
1
01
53
AM
PDT}

Prev 1 … 6 7 8 9 10 … 12 Next