The Simulation Wars

_{William Dembski
March 26, 2009

Intelligent Design

1}_{Categories
Intelligent Design}

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

I’m currently writing an essay on computational vs. biological evolution. The applicability of computational evolution to biological evolution tends to be suspect because one can cook the simulations to obtain any desired result. Still, some of these evolutionary simulations seem more faithful to biological reality than others. Christoph Adami’s AVIDA, Tom Schneider’s ev, and Tom Ray’s Tierra fall on the “less than faithful” side of this divide. On the “reasonably faithful” side I would place the following three:

Mendel’s Accountant: mendelsaccount.sourceforge.net

MutationWorks: www.mutationworks.com

MESA: www.iscid.org/mesa

Comments

(notes cont...) Does the fitness function we use matter? ================================== There is no way to code for all the possible Reward Matrices created by all the different possible fitness functions, so the above are just a tiny sampling. Some (like the Proximity Reward matrix) will help a search. Others will not, or may even hinder it. For a string of length n, where each position can have one of m characters, the possible number of string permutations is m ^ n. If we assign each of those permutations a value between 0 and F (in other words, give it a "fitness" value), we get (F + 1) ^ (m ^ n) distinct matrices, which can be generated by as many different fitness functions. To put real numbers to this, assume we have a binary string of three bits that we are searching for. (Let's say we're looking for the string 011.) Since we can have two possible characters (0 or 1) and our string is three bits, we have 2 ^ 3 = 8 possible permutations we'd have to search through. (This is our Search Space.) If we now want to use a fitness function to help find the string, we can decide to limit ourselves to fitness functions that assign values to the three bit strings that are between 0 and 3. Therefore, we now have 4 ^ (2 ^ 3) = 65,536 unique fitness/reward matrices and as many fitness functions that output a unique reward matrix for the string permutations. We must therefore choose our fitness function wisely from the 65,536 available, since not every fitness function will help us in our search. Since Weasel Ware allows you to try a variety of different fitness functions you can see for yourself that it isn't the power of selection that allows a difficult search to be successful, but the choosing of an information-rich fitness matrix that aligns closely enough with your target. Using just any fitness reward matrix (such as one that is neutral to the target) will by no means guarantee success, even if mutation, reproduction and selection are available, as they are in the Proximity Neutral Search. USAGE NOTES: ============ "Offspring" in the Proximity Reward and Proximity Neutral controls how many strings per generation are generated based on the current best string. Mutation Rate is the per-letter mutation rate for offspring strings. The default of 4.0% will change roughly one letter in a 28-letter string each replication, on average. Mutli Run Mode will allow you to test a given search strategy over a fixed number of runs. The "View Results" button will then show the results in both HTML display and CSV display, for pasting into a spread-sheet analysis program such as Excel. Atom's Notes: ========== Any errors in interpretation or understanding in the version 2.0 release are strictly mine, are unintentional, and should not reflect negatively on the EIL or EIL staff. If you do find any errors in the implementation, please let me know by contacting me via my website (atomthaimmortal.com, the contact form.) If you do choose to contact me, please be as friendly as you would in person.Atom_{April 8, 2009
April
04
Apr
8
08
2009
12:02 PM
12
12
02
PM
PDT}

Here goes the notes, since browsers may not word-wrap the text file... Changes in Version 2.0: ================= - Two new Searches: Proximity Reward Search and Proximity Neutral Search - Multi-Runs Mode: Useful for running the GUI many times in a row, for gathering data on average perfromance. - Sortable Searches: You can now reorder the searches, based on which searches you're interested in viewing. Just grab the search by its label and move up or down. - Disabling of Searches: You can disable searches, so that you spend your CPU cycles on only the searches you are interested in. - Proximity Neutral Search - New Reward Functions: Simple Sum, CRC32, Wave Interference, Partially Neutral: CRC32 Based, Partially Neutral: Anagram, Custom Fitness Function mode (where you can code your own fitness function to run in the GUI.) Weasel Ware 2.0 Highlights: ===================== To latch or not to latch?: Version 2.0 contains both a Partitioned Search, that uses information about the target to freeze correct letters in place, as well as a Proximity Reward Search, that does not freeze letters, but uses information about the target to assign values to candidate strings. Weasel Ware now allows you to run both types of searches side by side. Proximity Neutral Search: Since choosing a reward function rich in active information, such as a simple Proximity Reward function, is unrealistic for biological purposes, we have a Proximity Neutral Search available. This search can use fitness functions with varying levels of active information. There are fitness functions that hardly reference the target string at all, other than using information about the target length (such as CRC32, Simple Sum and Wave Interference) and there are those that use some information about the target string to further narrow the search space. These functions (prefixed with "Partially Neutral: ") do narrow the relevant search space, but not to the extent of a Proximity Reward function. For example, the Anagram fitness function ranks all Anagrams of the target string (which by definition will include the target string) as the highest rank, and will rank all other strings based on how many letters they contain that the target does not, or vice versa. (In other words, based on their distance from the Anagram subset.) In this way, it quickly converges on a subspace of only n! options from the full set of m^n options, where n is the string length and m is the number of letters in the currently used alphabet. For an english word ten characters long, we reduce the relevant search space from 205,891,132,094,649 options (27^10, the 27th "letter" being the space symbol) to only 3,628,800 options (10!), which we'd then have to search through using random methods. (Though even here we can further intelligently tailor our mutation rate and population size to aid our search.) The available Reward (Fitness) Functions for Proximity Neutral Search are: - Simple Sum: This function creates a reward mapping based on the ASCII sum of the characters of the string. It plugs this sum into a sin() function and takes the absolute value to get a number between 0.0 and 1.0, which it then multiplies by the string length. This value becomes the number of "errors" in the string. The string with the lowest number of errors is then chosen. - CRC32: This function uses a CRC32 checksum of the string to get a 32-bit integer, which it then plugs into a sin() function and multiplies against the string length, similar to the Simple Sum method. - Wave Interference: This function uses the interference pattern created by two sin waves of different wavelength and phase to get the value for each string. The lowest value represents the best string. - Partially Neutral CRC32: This function uses the distance between the CRC32 checksum of the string compared to the target string. It then selects the string with the smallest CRC32 distance from that target. - Partially Neutral Anagram: See discussion above. Selects based on distance from being an anagram of the target string. - Custom Fitness Function: In addition to the above mentioned functions, users can test any other fitness function they can come up with by using the Custom Fitness Function mode. A button labeled "Edit Custom Code" appears when that mode is selected, where users can enter valid javascript to assign two values used for comparison: aError and bError. Whichever has the lowest value becomes the "fittest" string, and is selected. The user has access to any standard javascript function and can also code their own functions, as the default examples show. (The default code loaded is a javascript snippet for a Proximity Reward function.) The two strings the users will compare are available as variables "a" and "b" in the javascript code, and the target string is simply the variable "target". The user's code will be evaluated using eval() and if an exception occurs it will be trapped and a message displayed to the user in red text. (On exceptions, aError and bError are both assigned a fitness value of 0.) (to be cont...)Atom_{April 8, 2009
April
04
Apr
8
08
2009
11:59 AM
11
11
59
AM
PDT}

Gentlemen, Weasel Ware 2.0 is now available for a sneak preview here: http://eil.digitalrelics.com/weasel It now features both latching and unlatched versions (Partitioned Search and Proximity Reward Search), custom fitness functions via the Proximity Neutral Search, re-orderable searches, multi-run mode (for doing large scale experiments), search disabling (to focus only on certain searches), and much, much more. Please take a look and relate any errors you notice. KF, You'll be pleased to know that your "grouped letters" mode is also available via the Custom Fitness Functions, as the "Advanced" Example. (You need to click on the "Tips for Creating Custom Fitness Functions", then "Examples" to find it...you can load it into the editor by clicking the button provided, then hitting "Update".) Your intuition was correct, search performance drastically degrades when grouping 2, 3 or more letters. An overview of the changes is available here: Changes in version 2.0 (textfile) This will be put on on the EIL site soon, so feel free to test and run some experiments in the meantime. Again, if you notice any errors, please let me know. AtomAtom_{April 8, 2009
April
04
Apr
8
08
2009
11:55 AM
11
11
55
AM
PDT}

kairosfocus, I edited the post I forwarded so your name doesn't appear. That's the limit of what I might do. I'm not going to edit everything for your delicate sensibility on the off chance you might get the vapors. Buy some smelling salts. The bulk of your response to Dr. Elsberry was non-responsive. I would have thought you would have attacked his math with fewer words and more calculations -- more than none, at least.David Kellogg_{April 8, 2009
April
04
Apr
8
08
2009
05:39 AM
5
05
39
AM
PDT}

Joseph said: 'The only thing the “weasel” program demonstrates is that Dawkins can write a computer program. It doesn’t do anything else.' It riles an awful lot of people :)faded_Glory_{April 8, 2009
April
04
Apr
8
08
2009
05:14 AM
5
05
14
AM
PDT}

I don't know if there is a direct quote or not, and I grant you that initially I myself more or less assumed that letters were fixed once found. However, upon reflection and considering the context of the entire argument in TBW, it is much more reasonable to let all letters mutate regardless of the fitness level of the phrase. Only in that way will the software mimic 'mutations random wrt fitness'. Published examples of Weasel programs that operate that way show clearly that the target phrase is reached quite quickly, making the point that Dawkins intended all along - cumulative selection vastly outperforms random selection. At this stage I really am unsure of what the discussion is all about. Everybody seems to agree that 'explicit latching', i.e. a partitioned search, is not required for the process to work. So why are we arguing?faded_Glory_{April 8, 2009
April
04
Apr
8
08
2009
05:09 AM
5
05
09
AM
PDT}

Joseph @219
As I said according to Dawkins once something is found the search for it is over.
Please provide a quote from the Weasel discussion in The Blind Watchmaker that supports this claim. If you read Dawkins' actual explanation of the Weasel program, there is nothing to suggest that letters are not subject to mutation once correct. In fact, the term "random mutation" is used. Either defend your statement with reference to what the Dawkins actually wrote or have the integrity to retract it. JJJayM_{April 8, 2009
April
04
Apr
8
08
2009
05:04 AM
5
05
04
AM
PDT}

The only thing the "weasel" program demonstrates is that Dawkins can write a computer program. It doesn't do anything else.Joseph_{April 8, 2009
April
04
Apr
8
08
2009
04:57 AM
4
04
57
AM
PDT}

kf writes,
...And, tha tis compounded by a slanderous conflation of design thought and Creationism — which DK I specifically requested that you refrain form using here.
I'm not sure that your request means much, kf. I could request that you never again write about how Weasel is faulty because the fitness function is merely comparing to a phrase and not anything functional, because we all know that and have acknowledged that to death, but you are under no obligation to honor my request. This is an internet discussion forum, and people just take what other people have the time and inclination to give.hazel_{April 8, 2009
April
04
Apr
8
08
2009
04:47 AM
4
04
47
AM
PDT}

faded_glory:
What is found and kept and improved upon is the phenotype, i.e. the phrases.
The phrase is the phenotype.
Not the genes, the letters.
There aren't any genes, only letters in the "weasel" program.
Those keep mutating happily away in the background, oblivious to how close the winning phenotype of each generation is to the target.
Please provide the relevant quote or quotes from TBW. This discussion is in reference to TBW only.Joseph_{April 8, 2009
April
04
Apr
8
08
2009
04:42 AM
4
04
42
AM
PDT}

Alan Fox:
There was nothiong in Dawkins’ “Weasel” program that fixed correct letters.
One more time for the willfully ignorant: 1- Given a target 2- Given a survival qualification of "closest to the target" 3- Given a small enough mutation rate AND 4- Given a large enough sample size The output will NEVER be less than the input. That means the letters are fixed just because of the programming. And that appears to be the whole point behind cumulative selection- that you don't keep looking for what has already been found. Now if you can provide the quotes from TBW that refutes that premise.Joseph_{April 8, 2009
April
04
Apr
8
08
2009
04:39 AM
4
04
39
AM
PDT}

Joseph, What is found and kept and improved upon is the phenotype, i.e. the phrases. Not the genes, the letters. Those keep mutating happily away in the background, oblivious to how close the winning phenotype of each generation is to the target.faded_Glory_{April 8, 2009
April
04
Apr
8
08
2009
04:38 AM
4
04
38
AM
PDT}

Alan Fox:
I asked Douglas Axe if he had successfully applied the explanatory filter in a biological context. He said “No”. What was wrong with the question?
For starters Dr Axe stated:
I have in fact confirmed that these papers add to the evidence for ID. I concluded in the 2000 JMB paper that enzymatic catalysis entails "severe sequence constraints". The more severe these constraints are, the less likely it is that they can be met by chance. So, yes, that finding is very relevant to the question of the adequacy of chance, which is very relevant to the case for design. In the 2004 paper I reported experimental data used to put a number on the rarity of sequences expected to form working enzymes. The reported figure is less than one in a trillion trillion trillion trillion trillion trillion. Again, yes, this finding does seem to call into question the adequacy of chance, and that certainly adds to the case for intelligent design.--Douglas Axe
(for the original Evolution News and Views article go HERE) Therefor you should have asked: "Dr Axe according to this quote (provide him the quote and the context) you said these papers add to the evidence for ID because you have determined that chance cannot account for the sequence specificity. How did you make that determination?" Ya see Alan the EF first eliminates chance and necessity, then if specificity is met, along with complexity, design is inferred. So just going by his quote if he didn't use the EF explicitly, he certainly used something very EF-like. However I don't expect you to understand that.Joseph_{April 8, 2009
April
04
Apr
8
08
2009
04:35 AM
4
04
35
AM
PDT}

Ummm in TBW Dawkins implies that not only does what is found, kept, it is also improved upon, however slightly. Alan Fox chimes in with:
The reason letters tend to stay largely unchanged is because the closest match to target is selected at each generation.
IOW Alan you don't understand my point. But anyway I challenge anyone to find the quote or quotes from TBW which would demonstrate that cumulative selection is a lost and found mechanism. As I said according to Dawkins once something is found the search for it is over. THAT is the whole premise behind cumulative selection. And THAT is how Dawkins portrayed it in TBW. So if you can provide the relevant quote or quotes that refute that.Joseph_{April 8, 2009
April
04
Apr
8
08
2009
04:30 AM
4
04
30
AM
PDT}

It was intended to demonstrate the power of cumulative selection as opposed to random selection
Really, Alan, you are rather late to the party. This has been pointed out numerous times to KF over the several weeks this latching/non-latching discussion has been going on. His insistence in not conceding this point, along with his continued use of Mr. when referring to Doctor Elsberry, has led me to the conclusion that he is not going to acknowledge any error on his part. Onlookers, even sympathetic ones such as me, have long since noted.crater_{April 8, 2009
April
04
Apr
8
08
2009
03:51 AM
3
03
51
AM
PDT}

Sure, Weasel is not an example of unguided evolution...
Which Dawkins himself makes very clear from the outset. It was intended to demonstrate the power of cumulative selection as opposed to random selection. "It is really a bit of a cheat" -Richard Dawkins.Alan Fox_{April 8, 2009
April
04
Apr
8
08
2009
02:52 AM
2
02
52
AM
PDT}

Mr M, There was nothiong in Dawkins' "Weasel" program that fixed correct letters. The reason letters tend to stay largely unchanged is because the closest match to target is selected at each generation. This is a simple matter of fact. "Weasel" did not fix correct letters. Will you accept Professor Dawkins' confirmation that this is so? By any stretch of the imagination, Dr Elsberry (and other, see links) have also demonstrated that latching is unnecessary. "Failed to make his case" is farcically denying the obvious.Alan Fox_{April 8, 2009
April
04
Apr
8
08
2009
02:49 AM
2
02
49
AM
PDT}

gpuccio, Sure, Weasel is not an example of unguided evolution. We can repeat this endlessly but I'm not sure why, because it was never presented as such, least of all by Dawkins who explicitly stated that it 'is a bit of a cheat'. I am at a loss why some people spend so much time and so many words to erect and tear down this strawman of their own making. The other point about needing prior knowledge has actually been addressed many times. Evolution has no goal, but that does not prevent it from generating solutions that are adapted to their environment. The main flaw in the tornado argument is that it ignores descent with modification in the presence of selection pressures. As a criticism of Darwinian evolution it misses its target by miles. Small stepwise tornadoes will indeed not result in an airplane. Do you really think anyone believes they do?faded_Glory_{April 8, 2009
April
04
Apr
8
08
2009
02:45 AM
2
02
45
AM
PDT}

faded_Glory: Ah, I forgot... And anyway, the weasel in no way "counters the ‘tornado in a junkyard’ argument against evolution". The tornado is not supposed to know in advance what it has to build in the junkyard. That remains by far the main difference which has to be countered. Even small, stepwise tornadoes in a junkyard, given a very long time, would never build up an airplane without any prior knowledge of what an airplane is. When has Dawkins, or anybody else, countered this argument?gpuccio_{April 8, 2009
April
04
Apr
8
08
2009
01:12 AM
1
01
12
AM
PDT}

faded_Glory: I have not followed the discussion about latching because scarcely interested in it. I think hazel's summary at 205 pretty well summarizes the substance of the matter. Least of all I am interested in showing is Dawkins was smart or not when he wrote those words in TBW: I already think that he is not smart at all, but for much more serious reasons. My idea is that the weasel argument in itself is completely silly as even generically related to the problem of biological evolution, unless Dawkin's intent was to show that intelligent, designed selection is a very powerful tool. Which, I believe, was not probably his purpose. So, his bringing about that argument, and using it for so long, has anyway been an act of cognitive dissonance, even if I am glad of how much that helped us in ID. In the same way, to be fair, I must say that I don't understand fully the necessity of my ID friends to concentrate so much on the "latching" issue. Finally, even if I have some minor problems with your terminology, I could well accept you statement that: "Weasel demonstrates that a process of steady-state random mutation of ‘genes’ (letters) coupled with selection of offspring ‘organisms’ (phrases) is far more likely to hit the target (a particular pre-defined phrase) than random selection of all letters in one generation." provided we add the following: "Weasel demonstrates that a process of steady-state random mutation of ‘genes’ (letters) coupled with intelligent selection of offspring ‘organisms’ (phrases) by means of a previous knowledge of the final target is far more likely to hit the target (a particular pre-defined phrase) than random selection of all letters in one generation.gpuccio_{April 8, 2009
April
04
Apr
8
08
2009
01:06 AM
1
01
06
AM
PDT}

3] Increasing N, the population size, does not make it more likely that the best candidate in a generation will show a loss of a correct base from its parent, it makes it less likely. On the contrary, when per letter mutation rate, population size and nearest- to- target filtering [with no reference to functionality -- NB: "nonsense phrases"] interact, under certain circumstances, it will be possible for letter substitutions and letter substitutions with advances in a third letter to occur. The relevant circumstances are that sufficent mutated phrases -- samples from the overall population of 10^40 configs -- are sampled for a generation that the far skirt/tail becomes a material factor in the generations, sufficiently so that enough double and triple mutation cases happen that the cases just outlined become reasonably likely to be sampled by the mutation mechanism. A threshold for that is that Np --> 1 or so, i.e. we are now in a sufficiently large sample population [N] that low probability cases as described are reasonably likely to be observed. [Thus, the force of my darts and charts illustration, where of course area of stripes in the bell or reverse J curve are proportional to odds of getting hit. The peak is far more likely to be hit than the tail. But if enough "drops" hit the chart and scatter more or less evenly, eventually you will see hits on the tail as a reasonable expectation.] And, the underlying assumption here is that EXPLICIT latching of the correct letters undergoing mutation -- i.e. in the previous champion -- is forbidden. So, ANY and all of the 28 letters faces the same probability of mutation, but one that is low enough that no-chance will be material, and that 1-change, 2-chance, 3-change etc will peak then tail off. Then, when the filter acts, it selects the closest to target -- not a random process but a hamming digital space distance metric or the like. (For ease of insight, let us say that an incorrect letter is a 1, and a correct one is a 0, so that the metric is a number that runs form 28 to 0.) In such a situation, most 1-letter changes will be the same or farther back than the previous champion. A relative few -- based on the 1 in 27 odds of letters being correct on random change -- will possibly advance. So, if the population is in a moderate range and the rate is small enough, we will see no change winning fairly frequently, and single steps forward dominating teh rest; which fits with the 1/2 or so cases being no change and the rest being dominated by single steps to target so that 40+ - 60+ gens are "good" showcased runs. Under these circumstances, double mutations of the previous champion will be rarer, and triples rarer still. but as the number in the generation of mutants rises, far-skirt cases will begin to appear, and eventually we will see substitutions [which would cause reversions while preserving distance to target] and substitutions with an advance [which would cause a step forward beyond where the previous champion was]. Such changes would then appear in the champions march, and we would see letters reverting and the like -- i.e first quasi-latching [rare shifts] then complete breakdown of latching -- especially int he earlier stages. Once we are close to target and MOST letters are correct, the population dynamics will shift significantly, as most single letter mutations would be reversions and substitutions and substitutions with steps forward will be harder to achieve. In short, Mr Elsberry has failed to make his case. ________________ GEM of TKIkairosfocus_{April 8, 2009
April
04
Apr
8
08
2009
01:06 AM
1
01
06
AM
PDT}

Onlookers: It must first be pointed out that the evo mat advocates at the often linked site have inadvertently told us a lot about how such would use power in institutions and the general community -- and, what we see is a clear warning to our civlisation of the peril it is in at such hands. We have been warned; if we are paying attention. Second, I must -- again -- thank Joseph for his kind intervention. he is of course right, but it is clear that all we can do is make the evident truth on the merits a matter of record for onlookers. Now, there are some points that are worth remarking on: 1] fG, 211: [Weasel] counters the ‘tornado in a junkyard’ argument against evolution . . . Actually, it simply begs the question of the need to achieve a minimal threshold of complex functionality [~ 600 - 1,000 bits worth of DNA; MUCH more than is in the Weasel sentence] before cell based life can emerge and before novel major body plans [~ 10's - 100's of mega bits] can emerge. Weasel works by INTELLIGENT design, carrying out a targeted search that makes no reference to the need for adequate functionality before hill climbing can be a reasonable strategy. That is acknowledged by Mr Dawkins in the qualifications -- or, is that: weasel words -- he makes Ch 3 of TBW, but the fact that he still used Weasel plainly reflected his judgement that it would be rhetorically effective nevertheless in distracting attention from the cogency of the Hoylean objection on the credible threshold of complex function before hill-climbing mechanisms such as natural selection is held to be, can be reasonable. It also illustrates how a confessedly "misleading" [cf 88 above] simulation can be all too rhetorically effective. In short Weasel is yet another misleading icon of evolutionary materialism. So, let us be warned. 2] Mr Elsberry, as cited DK, 202: the misuse of “latching” FYI Mr Elsberry, I (among others) first used this term and similar terms such as "ratcheting," specifically, to refer to the way the o/p of Weasel 1986 [not 1987, not various neo-weasels since and not the various quasi-weasels] acts: across 200+ of 300+ letters that could in principle vary, once a letter becomes correct, it stays that way on the o/p. You don't have any right to twist this around to assert that we are "misusing" the terms that describe what was evidently happening with the o/p of Weasel in 1986. That is a strawman fallacy, with ad hominem component. And, tha tis compounded by a slanderous conflation of design thought and Creationism -- which DK I specifically requested that you refrain form using here. On the contrary, you have a duty of care to represent those you wish to criticise fairly and accurately. This -- in the teeth of easily accessible information to the contrary -- you have failed to do. Back on the substantial issue: as I pointed out at 183 above, it is utterly implausible for the o/p of Weasel 1986 to be such that reversions occur but are always offstage, so that the observed o/p latching is only apparent, and artifact of sampling. For, the required patterns of behaviour are simply utterly probabilistically implausible. The simplest, best explanation of the o/p -- the march of generational champions -- as published by Mr Dawkins at that time, is that it is latched. Thus, the real question is the mechanism to explain it, per empirically based inference to best explanation. Weeks ago now, once this issue was on the table, I put forth T2 -- explicit, and T3 -- implicit, mechanisms. Of these, implicit mechanisms based on interaction of the proximity-only filter with the mutation rate and population size. Specifically, the number of generations to target indicates that no-change wins the generation championship about 1/2 the time in Weasel 1986, and that the rest of the time 1 step changes predominate. These are pretty direct inferences from the printouts. they entail that the double-step substitution mutation where a correct letter reverts and another substitutes, or the triple mutations where a substitution plus an advance where another letter goes correct are excessively far-skirt to materially affect the march of champions to the target, as published. All of this is consistent with the pattern of interaction between filter, populaiton per gneration and per letter mutation rates as described, under the rubric: implicit latching as a mechanism for the observed o/p latching. There is no misuse of terms -- at least, on my part. [ . . . ]kairosfocus_{April 8, 2009
April
04
Apr
8
08
2009
01:05 AM
1
01
05
AM
PDT}

Weasel demonstrates that a process of steady-state random mutation of 'genes' (letters) coupled with selection of offspring 'organisms' (phrases) is far more likely to hit the target (a particular pre-defined phrase) than random selection of all letters in one generation. It counters the 'tornado in a junkyard' argument against evolution. That is all it does, no more and no less. Why does anybody actually have a problem with that?faded_Glory_{April 7, 2009
April
04
Apr
7
07
2009
10:59 PM
10
10
59
PM
PDT}

Looking back, I now regret going along with "implicit" and "quasi-" latching. It would have been better to have used "latching" for just one thing - the situation where an explicit rule keeps a correct letter from ever be subject to mutation again - and found a different word to refer to the relatively steady progress towards the target. I'll know better next time, if there is a next time.hazel_{April 7, 2009
April
04
Apr
7
07
2009
05:00 PM
5
05
00
PM
PDT}

Link broken in #208 Link to Zachriel's "weasel" program. The site JayM links to in 207 also gives a lucid explanation.Alan Fox_{April 7, 2009
April
04
Apr
7
07
2009
02:33 PM
2
02
33
PM
PDT}

Dawkins' "weasel" didn't latch. We'll see if Professor Dawkins wishes to confirm this yet again. "Weasel" does not need to latch. This is amply demonstrated by Zachriel here. @ Joe I asked Douglas Axe if he had successfully applied the explanatory filter in a biological context. He said "No". What was wrong with the question?Alan Fox_{April 7, 2009
April
04
Apr
7
07
2009
12:56 PM
12
12
56
PM
PDT}

Joseph @206
n the original sense Dembski et al. referenced “The Blind Watchmaker” for this “latching”. And going by TBW there isn’t anything to indicate otherwise. Cumulative selection as described and illustrated by the “weasel” program- in TBW- is a latching process.
This has been repeatedly shown to not be the case. David Kellogg has posted copious, clear refutations of this claim. He even posted a link to a site that goes through the text of The Blind Watchmaker and creates the Weasel program from Dawkins own words. Rather than simply repeat your baseless assertions, why don't you go through the same exercise and show how it is even remotely possible to come to your erroneous conclusion? JJJayM_{April 7, 2009
April
04
Apr
7
07
2009
12:47 PM
12
12
47
PM
PDT}

hazel:
In the original Dembski sense, latching meant explicitly fixing correct letters so they were no longer subject to mutation, and that is the meaning we should have stuck with.
In the original sense Dembski et al. referenced "The Blind Watchmaker" for this "latching". And going by TBW there isn't anything to indicate otherwise. Cumulative selection as described and illustrated by the "weasel" program- in TBW- is a latching process. No need to go looking for what you have already found. Had he said cumulative selection was a lost and found mechanism then that wouldn't quite illustrate how nature can be a designer. "Here is nature, not only blind but also clumsy." wouldn't quite grab the readers' attention nor would it help make a case for TBW. So by reading the referenced material Dembski et al.'s inference of latching looks OK.Joseph_{April 7, 2009
April
04
Apr
7
07
2009
09:57 AM
9
09
57
AM
PDT}

I concur with most (maybe all) of what faded glory has said. Much of the confusion has been because we have been unable to settling on some definitions. In the beginning, "latching" meant that the program contained a rule that said that once a letter was correct in a parent, it would never change - it would not even be subject to the possibility of mutating. Once correct, the letter was fixed, or latched. Dembski's explanation in the paper mentioned way above assumed Weasel used latching in this sense of the word. The alternative to this is non-latching: correct letters can indeed mutate. Everyone agrees that in the long run letters that are correct stay correct, not because they are latched but because the selection process moves us slowly towards the target - which is the idea the program was written to illustrate. Because of this, kairosfocus added the phrases implicit latching or quasi-latching to describe this second idea. That is: non-latching at the mutation level leads to implicit latching at the selection level. Everyone, even kairosfocus, accepts that this is true. However confusion about the terminology just keeps muddying the waters. Arguing about what is "really" latching is fruitless. In the original Dembski sense, latching meant explicitly fixing correct letters so they were no longer subject to mutation, and that is the meaning we should have stuck with.hazel_{April 7, 2009
April
04
Apr
7
07
2009
09:07 AM
9
09
07
AM
PDT}

Alan Fox, You are not even addressing my argument. I take it that means you don't understand it. IOW you think that your lack of understanding is some sort of refutation. IOW you are right. Trying to have a discussion with you is a waste of time because you cannot think. But anyway:
Weasel did/does not latch because it does not need to latch.
If a target is given, AND the qualification for the survivor is "closest to the target", AND the mutation rate is small enough, AND if the sample size is large enough, the output will NEVER be less than the input. IOW latching happens because it was designed into the program given target, survivor qualification, small mutation rate AND large sample size. And yes I remember Doug Axe. Last I knew he didn't respond unequivocally and you didn't ask the proper questions.Joseph_{April 7, 2009
April
04
Apr
7
07
2009
08:35 AM
8
08
35
AM
PDT}

Prev 1 … 4 5 6 7 8 … 13 Next

You must be logged in to post a comment.

Leave a Reply