Dawkins’ WEASEL: Proximity Search With or Without Locking?

Is it possible that the code is the same, but other parameters, such as the population size and mutation rate, were changed between the book and the TV show? Right at the end of the video clip showing the simulation, you can see that the generation count was 2485. That is a very diferent result that less than 100 generations in the runs summarised in the book.

William Dembski:

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.

They both work, so what does it matter?

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.

Why not just ask him?

Stay tuned? You bet…

Dr. Dembski:

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).

Actually, Dawkins described his algorithm very clearly. What he didn’t tell us is the parameters he used, namely mutation rate and population size.

When I coded Dawkins’ algorithm, I chose a mutation rate of 5% and a population size of 50, just because the values struck me as reasonable (although 5% would be quite high in a biological context.) It turned out that these values rendered the reversion of correct letters highly improbable. The math to bear this out would be ugly but doable.

Why, then, is it natural to conclude that Dawkins implemented latching, didn’t mention it in his description of the algorithm in TBW, and then removed it before the 1987 video? Doesn’t it seem more likely that Dawkins’ parameters were such that the reversion of correct letters was highly improbable, and that he used a different set of parameters for the video?

Actually, on looking at the video, I don’t think that Dawkins necessarily changed his parameters. It appears that the screen in the video is cycling through the whole population, not just showing the winner. In that case, reversion of correct letters is occurring in the sense that correct letters get mutated, but not necessarily in the sense that selected winners contain reverted letters.

In summary, there is no evidence that Dawkins used a latching mechanism in his 1986 algorithm, and the 1987 video constitutes evidence that he did not.

I am still looking for the book “The Blind Watchmaker”- I have it on order through the local library so until I read it again I can’t say what Dawkins did in the book.

That said I have offered my opinion on why the “locking” is only apparent-

That is in each generation the only “survivor” is the one that is closest to the target.

Before the next generation the parent is the closest.

Now given that the mutation rate is 4% (about 1 letter change per offspring per generation), this means there is also a 96% chance there won’t be any change.

So once a match is found and deemed “closest to the target” then all surviving offspring should be at the minimum equal to the parent and at best some degree closer to the target.

Therefor to see if any letters are “locked” one then has to look at ALL of the REJECTED offspring.

That is if one cannot get a hold of the ORIGINAL code that Dawkins used in BW.

I disagree with R0b on whether Dawkins changed parameters. The number of generations is the evidence he did. I think the video parameters are a smaller pop and perhaps higher mutation. I think the result is a more “videogenic” simulation, since you see lots of visual change on the screen.

Another way of saying is that the letters are “locked into place” by the laws of probability.

That is given some “correct” mutation rate and the “correct” number of tries per generation to choose from, the odds would favor at least one offspring per generation being equal to the parent- ie no change.

The other 99 are competeing against that one to get “displayed”, and then becoming the “parent” of the next generation.

Picture “The Price is Right” wheel with every letter in the alphabet (and a space)- except that once a parent becomes established 96 out of 100 spaces are then that letter and the other 4 are any letter but that one- wildcards.

Then spin away- 100 spins per wheel, with 28? wheels trying to get “MeTHINKs…”

And each time a new parent is choosen the wheels change to match it.

Wesley Elsberry has just posted the following at his own forum:

I already corresponded with Dawkins back in 2000. There was no locking of characters in any implementation he did, nor was there any description of locking in anything he said.

If Dawkins is tuning the parameters differently for the program as described in the book and for it as exhibited in the BBC documentary, isn’t he in effect using a different program?

Re-running the same program with different data or settings is equivalent to using a different program? I don’t see that.

Dr. Dembski asks a very interesting follow-on question. Since Dawkins isn’t posting comments here yet, can we get Atom’s opinion about the same question re Weaselware or Dr. Dembski’s himself re MESA? All these parameter driven programs can give vastly different results depending on how the parameters are set.

I don’t care if he used a different program or not, if he monkeyed with the parameters to acheive a desired result, then he is demonstrating design not chance and/or necessity.

I also find it interesting that he admits in the video to aiming for a specified target but that evolution doesn’t do this,but then just glosses over this distinction as if it isn’t all that important. I’ve never understood why this ‘weasel’ program tells us anything about how evolution is supposed to work.

DonaldM, all sorts of things could be changed between versions of a program run: the mutation rate, the population size (number of mutated versions generated from each best fit), the speed of display. These wouldn’t make the program work signficantly differently, though they might change the time it took to reach the target, the number of generations, and the way the running program looked on film.

I don’t know as much about programming as many here, but I’m pretty sure this is right.

DonaldM:

I don’t care if he used a different program or not, if he monkeyed with the parameters to acheive a desired result, then he is demonstrating design not chance and/or necessity.

I was the one who suggested that he changed his parameters, but I retracted that suggestion in [6]. On looking at the video, I see no evidence that he changed his parameters.

R0b, Look at the generation number. There has to be a param change, probably lowering the population size significantly. That would lead to the fallbacks in good letters and a faster update of the screen, which I think was the reason for changing the params – better visuals!

Pendulum, I don’t think that “tries” refers to generations — I think it refers to instances of the sequence, ie “organisms”. In the video, it succeeded in 2485 tries. If the population was 50, that would be 50 generations (50 * 50 = 2500 — apparently it doesn’t finish out the current generation when it finds the target). That’s in line with the number of generations reported in TBW.

ROb, good point. I think some people have misunderstood what happens in the program:

Each “generation” contains a number of “tries” (these can be adjusted to see how it works with changing variables). The best “try” becomes the parent for the next set of “tries” (each of which is free to mutate at any letter, correct or not). One could change the number of letters that mutated and the number of “tries” per generation without changing anything foundational about the program.

I don’t know why people don’t seem to understand this. I don’t know beans about programming, and I understand it.

It would be interesting to change the target while the program is running. What happens then. What if

METHINKS IT IS LIKE A WEASEL

became

THINK YOU IT IS LIKE A WEASEL

and then

I THINK YOU LOOK LIKE A BEAGLE

in the middle of a run, but with no other changes?

I bet there would be no latching anywhere because the program never had or needed latching.

When I coded Dawkins’ algorithm…

In other words, a set of requirements/business rules was articulated by one intelligent entity and implemented by another. The programmer selected design parameters based on what he thought was reasonable, and used an integrated set of known-to-have-been-designed infrastructure (programming language, operating system, hardware…) to code to those requirements. There were probably bugs in the early iterations (as there always are), but he knew the “correct letters” (something evolution does not and cannot know, according to the mainstream theory) and was able to trouble shoot until the desired state was obtained.

Question: given the fact that none of the makers of this particular watch can justifiably be called blind, what conclusion are we expected to draw again?

Clive, Dawkins points out the same disanalogy that you do, namely that WEASEL has a target while biological evolution does not (as far as science can tell).

The point of WEASEL is to illustrate the contrast between cumulative selection and random sampling. Of course, WEASEL doesn’t demonstrate that the conditions necessary for cumulative selection exist in biology, and the point of irreducible complexity is to show that these conditions don’t exist in some biological cases.

Can someone help me understand this:

What exactly does this model purport to simulate? What exactly (specifically) does the METHINKS*IT*IS*LIKE*A*WEASEL represent?

If it represents a “completed” evolution of an organism, wouldn’t each successive generation have to have to have some natural advantage to the previous one? An advantage so great that the previous generation genetic makeup eventually dies out? If not, why does that generation keep going and not the others? Without these steps the model simulates design, not randomness. But just how big does an advantage have to be that it is to the detriment to rest of the gene pool? Isn’t that what each step represents?

This is also something I haven’t quite understood: does the model account for the variability of survival that has nothing to do with this natural advantage? In other words, the chance the organism will die before it is able to even take advantage of the mutation?

As you can tell, I am not a scientist, but I am having trouble getting my head around these things.

“Cumulative sampling” (to a statistical researcher, that is an odd term) is no more effective at coordinating functionality within separate organizations using meta information than “random sampling” is.

Yet, this is exactly what is called for.

Clive Hayden:

What if the goal were “any valid english sentence [of length n].” So of all the individuals on a given iteration, it would keep the closest phrase to any valid sentence. That would seem much easier to attain actually, than “me thinks it is a weasel”. It would also be general, not specific, but would result in very complex meaningful sentences.

And would it be comparable to, “Any viable biological organism.”

Clive Hayden [25]:

Apparently, evolution doesn’t have anything in “mind”, not even a “sentence”.

Well, apparently it has in mind “any viable biological organism”. And it has in mind “increasingly viable biological organisms”. Increasing viability would imply increasing information in an organism. adaptedness would mean the wherewithal to handle more and more situations in the environment, which would imply more and more complex organisms.

So in the context of “Any valid english sentence”, presumably there should be a goal of longer and longer valid english sentences.

I was discussing evolution with Kris on another thread, and he said that there was no “goal” to evolution.

I believe it would be contended that the solutions are in the landscape and not the selection process, but this distinction may not be relevant.

What is the shortest English sentence supposed to show? that evolution can do such a thing by keeping parts of the sentence until all of the sentence is reached?

Well, in the Zachriel algorithm (which is actually a more refined and thus more relevant version of the weasel program) Only valid english phrases are kept as intermediary forms. Increasing fitness is measured as increasing length of a valid english phrase (or sentence).

Clive @25 Doesn’t that mean that evolution is familiar with “sentences”?
Or at least English grammar. The analogy in biology would be “evolution is aware of protein chemistry”.

Switching from magenta to yellow, bees and dandelions don’t know what we call yellow, they are just agreeing on a signal that one can make and one can look for. The dandelion blindly paints itself a color, and the bee blindly wires its brain to look for a color. If they agree, they both survive and pass down their respective recipes. over much time, they evolve a fit as close as lock and key.

R0b @18, my bad. I didn’t see the “tries” wrapped on the next line. I now agree that it probably is showing every member of the population on the screen.

Clive:

“What is the analogy trying to show?
What would showing it prove?”

To use Dawkins’ other analogy, it shows that you can find your way up the natural staircase at the rear of Mount Improbable rather than clearing the sheer front face in a single bound. It doesn’t address the standard ID arguments that 1) there is no such smooth path requiring only manageably short leaps or 2) if there is, then that’s an improbable enough contrivance to indicate some kind of design in itself. But then, in fairness, it isn’t meant to. It’s not a sufficient case for “RM + NS”, but it is a necessary early step in one.

(But all this is IIRC as I don’t have the book to hand.)

Clive Hayden:

Do you mean to say that just by virtue of living and reproducing, mutating and evolving, that the English sentence is being crafted? one evolutionary step at a time? Wouldn’t that mean that there were some fixed property that the biological viability was approximating to in it’s “iterations”?

Fixed? No.

If the goal were “any legal C program” for example, that would be a dynamic goal, not fixed. And a context-free parser (something that could identify a C Program) is not in principle a complex program.

(BTW, If your indirect point is that there is teleology inherent in the process, I would agree with you, but it is just extremely difficult to get people to see that if they are not so inclined – just an observation.)

How would biological changes amount to anything “readable” over the course of the changes?–even in their own “language”? We already know when an English sentence is being approximated to, how do we have this equivalent from biological changes–how are the changed amounting to something that we already know and can read and pass judgment on their proximity to it?

I think you would have to agree that there would be potential objective standard for gauging comparitive fitness of organisms in an environment. So that if you compare two otherwise identical organisms and one has better vision due to specific refinements in its eyes. So however those refinements came about, it would be a true fact that the organism possessing them was more fit.

The Zachriel mutagenator program can generate a third of the english dictionary in a few hours. I don’t know even how it took that long – given that in my experience it was generated what looked like a hundred words in a few seconds. But all the intermediary words are legal words themselves. And then longer and longer valid words are formed strictly through random mutations. 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.

So now imagine you have a bunch of individual words floating around, and those that can hook up gramatically start doing that to form phrases. So then the challenge is to form phrases and sentences by combining words randomly, burt that’s no different than combining letters to form words. And then as phrases reach a certain length a period becomes an option. And then certain sentences that make sense together hook up.

And how would we get say a children’s book “Bob’s New Hat” out of this, for example? Well presumably there would be a lot of copies of Bob and Hat floating around and maybe it becomes sort of like Mad Lib and we get a whimsical children’s book out of it.

I go sidetracked here, but I will post this as is.

What I’m still not understanding is the analogy between when something has been, if you will, attained, or captured, such as biological viability, and how that’s analogous to sentences.
…
What would this be? Is the analogy saying that “biological viability” (which I presume means just living, and maybe reproducing) is akin to a valid English sentence in some way? How would the analogy follow?

Well language seems a very apt model for biological organisms. Languges are extremely complex and have themselves coincidentally evolved in a very haphazard incremental fashion. No one’s saying that it is a perfect analogy, but what would be a better one?

But if we’re define fitness as an increase in biological functional complexity, the relevance of an increase in meaningful sentence or word length should be apparent.

Word and sentence juggling as “support” for the plausibility of biological evolution is confused and fatally flawed. There are multiple problems.

To take one such problem as an illustration, Darwinian evolution does not evaluate by anticipating what will be useful someday. Selection is ruthlessly oriented to the present competition. However, if we imagine a world of functional sentences, the idea that isolated words can be preserved because they are part of English immediately breaks the analogy with biology.

In English, we might mutate “of” to “off”, knowing that both are part of the “landscape” of English — i.e. subsets of the already existing fully functional language, even though by themselves they are not yet functioning as sentences. But that is not the position evolution is in. Biology has no already compiled “dictionary” to know what incomplete sentence parts might be useful to a fully functional future sentence.

If you want to seriously attempt to perform some analogous thought experiment or computer simulation, you have to get rid of the idea that the non-functional parts (e.g. words in English) are predefined, knowable, recognizable — and therefore preservable.

The real function of such exercises has been to bolster hope that would normally be dashed against hard reality. Behe’s The Edge of Evolution takes a serious step toward uncovering how hard that reality really is.

Clive [20], I’m not sure how seriously you’re posing this question. But all languages, including English, do evolve, one step at a time. All languages have developed by means strikingly akin to biological evolution: They mutate and acquire new function beyond the control of any individual or group. While sentences are designed, languages (with the exception of oddities like Esperanto) are wholly natural.

Here is the first poem in English:

Nu we sculon herigean / heofonrices weard,
meotodes meahte / ond his modgeþanc,
weorc wuldorfæder, / swa he wundra gehwæs,
ece drihten, / or onstealde.
He ærest sceop / eorðan bearnum
heofon to hrofe, / halig scyppend;
þa middangeard / moncynnes weard,
ece drihten, / æfter teode
firum foldan, / frea ælmihtig.

(The slashes represent half-line marks in Old English poetry.)

Stuff like this evolved into modern English. How? Reproduction, variation, selection. Rinse, repeat.

When? Over time, one step at a time. Middle English — after the Norman Conquest — is a “transitional form” between Old English and modern English. The “body plan” of modern English stabilized with the introduction of print and the normalization of spelling that was codified — but not caused — by the introduction of dictionaries.

p.s. To complete the analogy, imagine someone trying to explain the existence of proteins unique to the bacterial flagellum by saying that those proteins are part of the landscape of bacterial flagellum proteins. Once mutation finds one of them, it can preserve it until the rest come along in similar fashion. Eventually a complete functioning flagellum can be constructed (after some trial and error combining the parts).

This would be obvious nonsense. No one would seriously propose this. Yet when we substitute words and sentences, the audience can be taken in by the illusion.

Such illusions work in part because, as users of English (or other languages), we take the language for granted. In its entirety, we are prepared to treat it as real and as natural as the physical landscape around us. We forget that without language translators (e.g. us) meaningful symbolic language could not exist.

The target search string is contained in the code. Therefore, no search is required.

That’s the end of that. Why would a computer programmer search for something that he included in his code?

I’ll tell you the answer: So he can dupe people into thinking that dumb, random searches can produce interesting, innovative results. Which they can’t.

Gil, if it’s so easily dismissed, why are people blathering on about partitioned search and latching and the like?

ericB

Darwinian evolution does not evaluate by anticipating what will be useful someday. Selection is ruthlessly oriented to the present competition

So is the Zachriel scheme.

If you randomly find a two letter word and from that randomly find a three letter word and from that randomly find a four letter word and eventually find a 10 letter word (and accomplish all this 200,000 times faster than could be accomplished via random generation of in tact 10 letter combinations), its because there’s regularity in the fitness landscape. When you preserved that two letter word it wasn’t with a mind to finding the 10 letter word you eventually hit upon by chance. But randomly selecting that two letter word (and then the three letter word and so on) did focus you in on a 10 letter word, but it wasn’t by your own design.

I.D. cannot require that the fitness landscape for evolution be random. If it is random, that has to be demonstrated. However, The fact that the human languages are structured to facilitate this type of random search (and also considering that human languages themselves evolved via a haphazard a process.) seems strongly suggestive as to the potential nature of biology.

Clive Hayden:

I don’t see the analogy between biological complexity and sentence complexity. The only similarity is that they’re complex. But it alludes to an implication that biology is approximating to something already in existence, such as a meaningful statement, already known.

The statement “I see” has meaning on its own. It doesn’t have meaning only on the basis of some other sentence it could potentially be a part of, but not yet in existence. But by combining it with other words gives the phrase increasing specificity (i.e. complexity) and increase its fitness for a specific function. “I see dead people”.

biological changes cannot be “read” and understood.

It seems they are read and understood by reality itself.

Consider an eye, but not the means by which it may or may not have emerged. Are you saying that reality cannot be allowed to determine how specific attributes of an eye’s physical configuration dictate how the eye functions and how this function bestows on the eye its highly selective advantage.

The issue isn’t whether the Zachriel program is a perfect model of nature. Clearly however, its a vast improvent on the Weasel program, in considering the issue of intermediate functionality among many other things. Why would the Dawkins program be continually attacked for a deficiency in it that its author acknowledged from the beginning, and at the same time the Zachriel program be merely ignored.

Gil:

Why would a computer programmer search for something that he included in his code?

To illustrate cumulative selection. That’s a no-brainer for anyone who has read the relevant section of TBW.

I’ll tell you the answer: So he can dupe people into thinking that dumb, random searches can produce interesting, innovative results. Which they can’t.

Apparently it’s okay on this board to divine nefarious motives without offering a speck of evidence.

And I don’t know what you mean by “random searches”, but genetic algorithms most certainly produce interesting and innovative results. Dawkins’s algorithm, of course, produces no useful results, as the target is defined in the domain of the objective function. Useful genetic algorithms, on the other hand, find elements in the domain that meet desired criteria in the codomain.

[41]:

I can certainly see the potential for teleological arguments concerning the probability of getting a fitness landscape so structured. In fact, any such argument would reduce to the observation that if f(x) = y, then f(x) cannot be more probable than y.

H’mm:

Seems I am late to the party.

A few observations, but first, a clip 9courtesy Wiki) on what Dawkins said he was trying to do, and did:

[Citation from BW ch 3 by Dawkins, on Weasel:] We again use our computer monkey [NB: to select a specific target functional case form 28 27 state elements identifies to 1 in "only"~ 1.2 * 10^40, well below the credible challenge to get FSCI for OOL and body plan level biodiversity, which is what Hoyle -- who Dawkins was trying to rebut -- had raised], 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, 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 [but just such single step to gain functionality is actually what is necessary for natural selection to compare on differential functionality]: 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.

1 –> Let us highlight key admission no 1: The computer 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.

2 –> Admisison no 2: This [the time the PC would take to select at random something on the order of 1 in 10^40] is more than a million million million times as long as the universe has so far existed.

3 –> So, to escape the implications of the long odds against arriving at functionality by chance [and remember, these odds are far, far, far better than those for observed bio-functionality information . . . ], Dawkins illustrates a purportedly BLIND watchmaker, by intelligently designed, targetted search that rewards detected proximity to a target of non-functional, nonsense phrases.

3 –> So, Weasel illustrates, not the power of a BLIND watchmaker, but the capacity of an intelligent designer to find a target by a partially random search process. “Improved” versions could doubtless go all the way to GA’s that define a functionality parameter and seek to get it to peak.

4 –> But in neitehr case would there be any proper dismissal of the key insight: Weasel-style programs instantiate intelligent design in action, not he power of some alleged BLIND watchmaker.

5 –> Therefore, let this major point be settled in our minds, before we turn to a secondary issue: the observed lockstep of Weasel circa 1986, where letters in the output latch and we see that 200+ out of 300+ letters in the program’s output, show an evident pattern of o/p latching with no counter instances. (Note, it is the absence of counter instances that led tot he inference, far and wide that the program in question is a letter- by- letter search that rewards “success” on a letter by letter basis, which is also a very plausible reading of Mr Dawkins’ words. For,t he minimal step forward is obviously one letter’s step forward.)

[ . . . ]

6 –> What best accounts for it? Candidate no 1 is partitioned search with explicit latching [my model T2 from previous]. No 2 is an implicit quasi-latching [my model T3]. However, it is to be noted that Apollos showed by code example that explicitly latched programs can show quasi latching if that is written in.

7 –> So, only code will be explicitly decisive, if demonstrative proof is required; which obviously will not be forthcoming. But, we can proceed on inference to best explanation. Ont he first look, the most credible explanation is that Weasel circa 1986 latched, but was improved by 1987 to not latch.

8 –> On charity, we will however take Mr Dawkins at his reported word, that he did not ever EXPLICITLY latch Weasel. So, the particular balance of parameters used, implicitly latched in the 1986 cases, and quasi-latched with fairly frequent flickbacks in the 1987 video. (The suggestion above that parameter tweaking for visuality advantage is an interesting suggested context.)

9 –> On number of generations to target, we have explicit statements in the 1986 cases that the program took 40+ and 60+. By sight, especially the rate at which flickbacks reverted in the topline display of what appears to be current generation [the bottom one that runs too fast to follow by eye is obviously current mutant), we are running at a generation per ~ 1/10 second or so. So, we probably have a much higher number of generations in 1987, which runs a lot longer than 40 - 100 gens at one per ~ 1/10 sec. [A generation per 1/10 second is also consistent with the report that the Pascal version ran in 11 seconds, which would run out in ~ 100 generations at that rate.]

______________

CONCLUSION: Weasel does not address the Hoyle challenge, on getting the increments of functionality required to be on the shores odf Dr Moreau’s Island Improbable, for which hill-climbing algorithms can in principle climb, step by step up the easy slope side to peak performance at the crest of Mt Improbable. Instead, it rewards closeness to target of non-functional configs of nonsense phrases. Thus, it begs the question Hoyle raised and acts as a distractive strawman from the issue of FSCI/CSI; that is, it is one of the unfortunately long line of fundamentally misleading icons of evolution. Now, charity per Dawkins’ reported statement, we can accept that the Weasel program does not explicitly latch, but instead creates output latching as seen in the 1986 printoff samples, by parameter balancing etc, i.e. quasi-latching with significant incidence of flickbacks in at least some cases. [Algor model T3 not T2 from 346 - 7 in the previous thread].

All said and done, that leaves the bottomline issue clear: Weasel does not answer to the FSCI challenge,as it rewards non-function on proximity to target. And that we canboth see and can understand from Mr Dawkins’ direct published statements.

Weasel — sadly — in the end only manages to be a distractive, only rhetorically persuasive irrelevancy. On substance, we still have only one credible explanation for CSI/FSCI: design.

GEM of TKI

There isn’t any such thing as “cumulative selection” in nature.

Cumulative selection requires a target.

In nature the “target” does not (yet) exist.

And as I said earlier the “locking” of the letters is a product of probability.

That is given a small enough mutation rate and a large enough sampling size (number of “offspring” per generation), letters will appear to locked in place when only viewing the chosen one- that which is closest to the target.

That is just the way it is.

Of course that would change if Dawkins’ original program has the parent copied unaltered in each generation, as well as others that can be altered.

Challenge to evolutionists: Can you salvage Dawkins’ WEASEL words (or similar attempts) from the War Games fallacy?

At a dramatic point in the movie War Games, a computer endeavors to crack a password code. After working some time at blinding speed, it manages to solve one character of the password. Some time later, it has solved another character, and so on.

It does not take much reflection to realize this is nonsense. If the password is unknown, it cannot possibly determine it one character at a time, preserving and accumulating partial successes until it reaches a working password. A password works as a whole, or it doesn’t work at all. It is all or nothing.

How are these attempts by Dawkins or others (e.g. Zachriel) at letter and word manipulation not equally nonsensical as illustrations for biological evolution?

For instance, in the second example series of the opening post, the sequence “LIKE” appears very early and is preserved, even before it is functioning as a word, let alone a word in a functional sentence.

The same type of question would apply to “the Zachriel algorithm (which is actually a more refined and thus more relevant version of the weasel program)” according to JT’s description.

What is the biologically equivalent justification for preferentially preserving partial progress toward function even before function is reached (e.g. before we reach a functional sentence)?

I have noticed that in the WEASEL runs printed above, there is no instance of the letter U. Nowhere, not in almost 450 characters and spaces.

From that observation, I propose that the WEASEL program will never change another letter to a U. The program must be designed to avoid Us.

Silly, you say? A pointless distraction? Dawkins mentioned nothing about it?

I. Don’t. Care.

I don’t see a U in these partial prints of runs. Therefore, the program can allow no Us at all.

But, you say, there are Us in the 1987 video.

Ahh. 1987. That’s a different matter.

If a U shows up in 1987, it must be a different program.

David,

No one cares what an English professor has to say about computer programming.

No one.

Joseph, how is the allegation of locking any more evidence-based than my allegation about the letter U?

David,

No one cares what an English professor has to say about a computer program unless tat program was written by thta English professor.

And if all you can do is act like a little child about it, then why bother?

Don’t like the message? Dismiss the messenger.

Your qualifications are precisely what, Joseph? Please be specific.

The message was derived by the messenger. And it was derived just to prove the messenger can act like a little baby.

My qualifications:

30 years of working with computers, writing and editing codes.

Also, as evidenced by responses to this question, I understand perfectly what is going on with this “weasel” program.

Mr Kellogg:

before I leave for the day.

1 –> you will see above that there is some fresh evidence, albeit of course hearsay, that Mr Dawkins states that he has not used explicit locking. Fine, this is not a courtroom: on charity, I have taken that.

2 –> the import of that is that as Joseph says, the program implicitly latches in the case in view.

3 –> That no “U” happens to appear is irrelevant to the observed fact that with 200+ out of 300+ letters in play, no correct letter reverts, regardless of want of functionality. (Indeed, the absence of a U is easily accounted for by the rapidity of convergence by which 2/3 of the available slots are effectively out of play. Wrong letters are being rather rapidly squeezed out by the force of the proximity rule, probably multiplied by a modest per letter “quasi-functional” mutation rate — 5% [which is utterly huge for realistic mutations in bio-forms!] means 1 – 2 letters typically vary per population member. I use “quasi-functional” as in effect any mutation is allowed into the population to compete. the nearest tot he target — regardless of want of overall linguistic funciton — is promoted, which means that the trend is to converge to the target. (The circumstances would bias outcomes sharply towards those that have 1 – 2 functional additions, and do not derange the existing ones. These circumstances are utterly unrealistic, of course. Thus, implicit locking [and with suitable shift of parameters, quasi-locking] as I discussed at T3 in my previously linked 346 – 7 in the earlier thread.)

4 –> Thus, the targetted, proximity based search implicated by the behaviour and by Mr Dawkins’ direct statements as cited, is irrelevant to the capacity of a BLIND watchmaker such as RV + NS. Hoyle’s FSCI challenge still stands unmet.

5 –> Similarly, as a third level matter, in the 1987 video, the program’s observed behaviour, with fairly frequent flickbacks, is significantly different from the case of the 1986 printoffs, which we have good reason to see as representative of the overall pattern of the program.

6 –> that suggests at minimum, that probabilities and populations have been shifted, which is a material difference to the program. Such an adjustment does not obviate the force of the point at 4 just above.

GEM of TKI

Joseph, you certainly understand better than kairosfocus!

Kairosfocus, I find your writing opaque as always, but I think you may be notpologizing for your silly accusation that the program fixed letters in the first place — which people have been telling you for weeks!

If that is out of the way, it might be possible to correct your other misunderstandings, but the return on investment is very low, so I’m not going to make that effort.

EricB [48]:

The same type of question would apply to “the Zachriel algorithm …

What is the biologically equivalent justification for preferentially preserving partial progress toward function even before function is reached (e.g. before we reach a functional sentence)?

The Weasel algorithm has not definition of functionality other than one particular sentence.

The Zachriel scheme is completely different. Here is what it defines as functional:

1)Any English word.

2)Any group of English words that can legally appear contiguously.

(And the longer a legal word or legal group of words is, the more adaptive it is.)

At no state of the search is anything preserved that does not meet the above criteria of functionality.

———————–

A legal sentence is an example of a legal group of words. Any contiguous subset of words in a legal sentence is also a legal group of words.

If you accept a sentence as being functional (which evidently you do judging from your above statements ), then you have no rational basis for denying that a smaller valid set of contiguous words is also functional (albeit to a lesser degree).

If you accept a word as functional, you have no rational basis for denying a set of words that can legally appear contiguously together in English is also functional (and functional to a greater degree than a single word).

As far as a “biologically equivalent justification” specifically – I personally would say that the parallel to English is valid to the extent that there are “grammatical” rules governing how biological components can be configured together, and to the extent there is a hierarchical and modular organization of biological components, being built up from reusable subcomponents and the like, and to the extent that there is semantic meaning arising from the arrangement of biological components which determine what larger biological components can be associated together.

And to reiterate, it would be reality itself that determines if a biological organism is functional. It is reality itself that parses the chemical and physical syntax and semantics of some biological configuration, and determines whether that configuration confers some selective advantage.

I don’t really feel inclined to debate this a lot longer, though.

Here are the websites:

http://www.zachriel.com/mutagenation
http://www.zachriel.com/phrasenation

—————————

Someone made statements previously to the effect that the reason that English is amenable to this type of incremental search (i.e. an efficient imcremental cumulative search that hits intermediary islands of functionality ) is because English is intelligently designed. Well then if you say biology is intelligently designed, then you are forced to to admit that it also is amenable to the same type of search.

David Kellogg,

If you want to get technical the program does lock in letters because of the probability thingy and the target.

Those two factors, working together, will most certainly produce “survivors” that have at least the match that the parent had and you will never see any reverses- that is given a small enough mutation rate and a large enough sampling size.

Take away the target and see what happens.

Monkeys pounding away on a keyboard do not have a target- not that they know of anyway.

JT:

Someone made statements previously to the effect that the reason that English is amenable to this type of incremental search (i.e. an efficient imcremental cumulative search that hits intermediary islands of functionality ) is because English is intelligently designed. Well then if you say biology is intelligently designed, then you are forced to to admit that it also is amenable to the same type of search.

Wrong. Not all that is intelligently designed has to amendable in the same way.

However with ID biology does have a target and the resources to help reach that target.

JT, explain to me the justification for calling the lone word “of” functional, to use one simple illustration. What is its function, apart from other words in a meaningful sentence?

Obviously, if one claims that it is functional by mere definition, one can define anything at all to be functional.

One might also “define” a lone protein as functional, despite the fact that its function will be future in the context of other proteins.

To be selectable, it needs current function, not future function. The use of arbitrary definitions does not salvage an inappropriate and non-functioning analogy to biology.

ericB:

Can you salvage Dawkins’ WEASEL words (or similar attempts) from the War Games fallacy?

I believe I addressed this in #32 above. AFAICR, the Weasel assumes the existence of a gradual path from source to target; it’s not meant to prove the existence of one.

Put another way:

Given

1- A Target

2- A small enough mutation rate

3- A large enough sample size

The output will never be less than the input if the output is “closest to the target”.

IOW there doesn’t have to be coded statement that locks the matching letters.

The locking is a byproduct of the program.

ericB [60]:

Obviously, if one claims that it is functional by mere definition, one can define anything at all to be functional.
…
The use of arbitrary definitions does not salvage an inappropriate and non-functioning analogy to biology.

As best I can tell, any analogy would be inappropriate to you.

Its obvious you don’t have any definition in mind yourself as to what would constitute functionality within the context of an English Sentence.

A computer program is a series of words too. Those words are generally speaking actual functions. If the target were a C statement as opposed to a sentence, what would be the functions in that. It hardly seems “arbitrary” to identify words and groups of words in a sentence as carrying the functional information of that sentence.

If you have no concept in mind yourself as to what constitutes functionality in a sentence, then any objection to the weasel program for preserving non-functional states is meaningless, as to you, any intermediate state is non functional.

JT, explain to me the justification for calling the lone word “of” functional, to use one simple illustration. What is its function, apart from other words in a meaningful sentence?
What about the word “apple”? Any problem with that term being defined to carry functional information? If you personally want to derive a exacting set of exceptions, for example ruling out single connector words like “of” as legitimate targets by themselves then fine. Why don’t you just imagine the model with all necessary revisions that would render it sensible in your mind. Or is your potential list of objections actually endless.

Oops- the locking is a consequence of the programming.

IOW it is an inevitable outcome- relatively speaking input to output- given the proper parameters.

There isn’t any need for a specific coding sequence to lock in any letters.

JT et al:

If we’re going to do simulations, let’s at least make them conform to the theory.

Consider the following string of letters (which happens to be a scrambled English clause—underscores represent spaces to give you a head start if you like puzzles). This string represents some kind of biological proto-structure.

wsoldin_je_d_swodx

Now, consider the following “mutations” to the original, A and B.

A wsocdin_je_d_swodx
B wsoldin_je_d_srodx

You, representing natural selection, do not know the target—in fact, you don’t even know what a target is, as you have no goals. Additionally, you cannot predict, cannot reason, and certainly cannot write computer code; you don’t have any lifelines and you can’t buy a vowel. All you can do is evaluate the “fitness” of each variation in the here and now. Your options are: select A, B, both, or neither.

Which do you choose and why.

ericB [60]:

ericB:

I think I see what you’re getting at now.

Consider the phrase “Tomorrow, I think I will wash the car.” The sentence fragment “Tomorrow, I think I will” is certainly generally functional, and millions upon millions of sentences can start that way. In fact, while writing “Tomorrow I think I will…” I could have a change of heart and instead of writing “Tomorrow I think I will wash the car” I could write, “Tomorrow I think I will go to the beach”. So, we can imagine an eye being developed incrementally and some component that was being developed could certainly be associated with innumerable other potential organs – not just an eye.

But your point would be that “Tomorrow I think I will” could not exist on its own – would not be a valid functional target on its own – there would have to be something with it, it could not be a target in isolation. But in what sense is the sentence “Methinks it is a weasel” functional in isolation? Could you write a post that only contained the phrase “Methinks it is a weasel?” Everyone would certainly consider that an unacceptable post and incomplete. But you implied such a sentence would be a complete function.

And furthermore, even though the “Tomorrow, I think I will” ostensibly cannot exist in isolation, the fact is it does. Until I finish the sentence the phrase exists in isolation for a discrete interval of time. Its not as if the phrase “Tomorrow, I think I will wash the car.” materializes instantaneously. And nor am I locked into one target merely by beginning a sentence “Tomorrow, I think I will”

As far as individual words, they do exist in isolation as well, in my mind (and I would say my brain – not some platonic dimension).

Anyway, I have more intuition as to what you were implying now, and have a sense how I want answer it, but just don’t have any more time at the moment.

JT:

So, we can imagine an eye being developed incrementally and some component that was being developed could certainly be associated with innumerable other potential organs – not just an eye.

Only if it was designed to do so.

That is an eye being developed incrementally.

JT, when you next have even a spare minute, SteveB has posed a fine question at 65. I would also be interested in your response.

As a speaker of English, you are falling into the trap of unjustifiably assuming too much, based on your existing fluency in English as a whole. Without realizing the unfair edge this gives you, you are backtracking from the meaningfulness of the whole to a subset. For example, you wrote (with my emphasis added):

63: It hardly seems “arbitrary” to identify words and groups of words in a sentence as carrying the functional information of that sentence.

and

66: Consider the phrase “Tomorrow, I think I will wash the car.” The sentence fragment “Tomorrow, I think I will” is certainly generally functional, and millions upon millions of sentences can start that way.

and

66: So, we can imagine an eye being developed incrementally and some component that was being developed could certainly be associated with innumerable other potential organs – not just an eye.

and

66: As far as individual words, they do exist in isolation as well, in my mind (and I would say my brain – not some platonic dimension).

But of course, mindless Darwinian evolution cannot use a mind and it cannot select based on potential organs, just as it would be illegitimate to defend the preservation of a word or phrase in isolation based on the meaning it would have in a sentence (that doesn’t yet exist).

You cannot legitimately have selection take out a loan against future meaning or future functionality. Yet, without realizing it or thinking about it, that is exactly what you as an English speaker with a mind are doing when you consider fragments to have meaning/function based on potential completed sentences.

But you implied such a sentence would be a complete function.

I am allowing, uncritically and for the sake of argument, that a complete and correct English sentence is considered to have function, so that we can examine whether Darwinian sentence building programs show us anything worthwhile — even if we grant that assumption. I am supposing, hypothetically, that a sentence might be considered to be something like a working bacterial flagellum. The question then is clearly this: How does one justify preserving all of the needed words (proteins) while we have not yet reached a working sentence (flagellum)?

Notice that we cannot assume we have a working flagellum (sentence) and then backtrack to the idea that all the parts have function. That is illegitimate. Going in a forward direction, one must contend with finding a basis for selection when the isolated sentence parts have never yet made any functioning sentence at all.

If such programs were made realistically comparable to real world requirements for undirected change, they would become just as feeble as Darwinian mechanisms are in practice (cf. Behe’s The Edge of Evolution).

Mr Kellogg:

Re :your silly accusation that the program fixed letters in the first place

Mr Kellogg, all that this cited statement reveals is that you are still unwilling to look at the evidence that is staring you in the face, not least at the head of this thread. (I will ignore the ad hominemistic tone.)

1 –> Please, first, LOOK. And, COUNT.

(You don’t have to take my counts, do your own. Having counted, could you let us see your results: (i) how many letters appear after the first line in each run as published? (ii) of these how many are “correct”? and (iii) how many of the correct letters revert to incorrect status? And, (iv) are you able to comprehend the just stated point and questions — if not, why not?)

2 –> I believe, you will see that in the 1986 program outputs, there are some 300 places in which letters might change.

3 –> As pointed out, of these 200+ go correct and once they are correct, NONE reverts in the sample space. A sample space of 300 instances in which we see 200 of a phenomenon, and none of the phenomenon that would point away, is not “no evidence” or a “silly accusation” or the like. It is in fact, well within the law of large numbers, and we have no reason to believe that sampling every tenth line will correlate with the program’s behaviour such that thew results will be strongly atypical. [Do you understand what the adverted to "law odf large numbers" speaks about? Do you understand what "typical" in a sampling context is about? Do you understand that scientists will routinely publish their "best" results?)

4 --> Thus, as I noted from December on, we can see that the output plainly has "latched" letters, and indeed, it does so in a context that Mr Dawkins admits, rewards non-functional increments towards the target. [NB: As I have pointed out since Dec last, the fact of marching towards a target without reference to functionality is enough to show that Weasel is not an example of a BLIND watchmaker -- the thesis it allegedly supports -- but of in fact intelligent, foresighted, goal-oriented DESIGN.]

5 –> The only issue of consequence on is what mechanism gives rise to the observed latching of the output. On that question being raised as an issue in the previous thread, at 346 – 7 I put forth three alternative candidate reverse engineered algorithms:

T1 — monkeys typing at random with hit/miss;

T2 — explicitly letter-latched search with warmer/colder;

T3 — implicitly letter-latched search with warmer/colder.

6 –> Of these, T1 is by Mr Dawkins own words exactly what he is trying to provide an alternative to. this, because as he states, it will not find the target on a PC within the available search resources, most notoriously. [But, observe: the increment in informational functionality to make the Weasel sentence is far smaller than that for origin of life or of body plan level biodiversity, which are plainly beyond the search capacity of the observed cosmos. This is of course being studiously evaded and ignored.]

7 –> As it turns out, BOTH T2 and T3 can supply the observed results, the difference being simply if one is explicit or implicit. Indeed, it is also the case that for the different 1987 behaviour, BOTH T2 and T3 can again fulfill the observed behaviour, as Apollos showed by writing a code that explicitly latches but occasionally reverts.

8 –> Thus, strictly, only inspection of credible code would be demonstrative. But, on the recent report above that Mr Dawkins has stated that he did not explicitly latch, on the charitability principle of accepting testimony unless there is reason to specifically reject it, I am willing to agree to T3 as the current best explanation. Never mind, that I suspect that such an algorithm is probably going to struggle to get the sort of “typical” outputs as are published in the 1986 case, if the 1987 program output is also typical of full runs for the 1986 cases.

9 –> To illustrate how implicit latching could appear, with 28 letters and a 5% per letter “mutation” rate [the parameters are unreaalistically favourable relative to the bio-world, of course . . . ], typically 0, 1 or 2 letters will change from one generation to the next: e.g. there is about a 24% chance that no letters will change for any given member in a given genration. So for a smallish to modest populations, the program will normally preserve currently correct letters or will advance a step or two. (With a bigger population and that sort of per letter mutation rate, with the closest to the target selected for, the program would probably find a case with two or even more further advancing letters in each generation, and would jump to the solution rather quickly, maybe even better than the deterministic, explicitly letter-latched search that is guaranteed to hit the target in 27 generations. [The 1987 version does not seem to do that sort of rapid convergence, just the opposite.])

10 –> By shifting parameters a bit to slow down approach to the target, the program will be more likely to show advances but with occasional flickbacks, e.g. if double-letter changes pull one back and push one forward. But the problem here is that the 1987 flick backs then typically rather rapidly flick forward to correct state. (Hence the flickering, letter winking effect.)

11 –> But, strictly on the evidence of the 1986 output, the simplest — most natural — explanation would be explicit latching, as we can reasonably be assured that Mr Dawkins published samples are TYPICAL of “good” runs of the program at that time. In short, latching without reversion was the typical “good” output behaviour of Weasel at that time. (All the emphasis on how Weasel does not latch is based on subsequent developments. indeed, the 1987 video of outputs that fairly frequently flick back on correct letters, as already noted, was on the order of months to a year after the 1986 published data. There are no good grounds to equate the 1986 and the 1987 outputs.)

12 –> Algorithms implemented for subsequent versions of Weasel are irrelevant to the behaviour of the program as at 1986, as published.

_____________

Onlookers, we continue to see the damaging impact of selective hyperskepticism.

GEM of TKI

Joseph, re 62:

To get implicit latching without reversions, the population as well as mutation rate actually need to be mutually tuned, so that you get steady advances without letter substitutions [one flicks back while another moves ahead].

In short, for a given mutation rate, there is an “optimal range” of population size in which the number of mutations per member of a generation is sufficient that advance will happen, but not so large that it will converge over-rapidly relative to the sort of run lengths we see as probably good runs: 40+ and 60+.

Also, the population has to be small enough and the rate for mutation small enough that we are unlikely to get the sort of double-mutations that would revert one while advancing another. (Observe, in 200 possible places for that, we never see it happen in the 1986 outputs.)

Explicit latching is probably significantly easier to achieve than such tuning that achieves implicit latching or quasi-latching.

But, if Mr Dawkins said that he did not latch explicitly, then I conclude that he most likely did some tweaking to get what he then thought was “good” results.

And, that boils down to implicit letter latching or quasi-latching.

GEM of TKI

PS: One apparent feature of the 1987 output is the “winking” on the reverting correct letters in what I take to be the “generational champions” string. That is, they come back on target fairly quickly once they wander off. On average ADDITIONAL letters are going correct once every 2 – 3 or even 4 generations, to get runs of 40+ and 60+, which means that often — 1/2 or more ot the time, it is the 0-mutations case that wins the contest. At 5% rate, there will be 24% of the population in that state. (That would be suggestive of such flicked back letters being specially treated.)

If it represents a “completed” evolution of an organism, wouldn’t each successive generation have to have to have some natural advantage to the previous one?

You would think so, yes.

David Kellogg –But all languages, including English, do evolve, one step at a time.

That is a very interesting point, and you are correct. But seriously, how could language evolve without intelligence i.e. understanding and agreement that particular sounds depict particular objects?

And of course all alphabets are created.

Something else to ponder: Why does Dawkins program stop with METHINKS IT IS LIKE A WEASEL

Because that was the goal of the program.

Goal reached program stops.

Here is a pair of house building illustrations of the War Games fallacy that evolutionary sentence building programs repeatedly fall into.

Suppose one writes a program to simulate the construction of houses or other buildings, perhaps out of discovered/mutated building blocks, rather than the construction of English sentences.

War Games Fallacy, Version #1

In this version, the program is designed to eventually produce a specific target house. The program cheats by comparing the work in progress against the complete target house. So if it discovers that a roof piece matches at such and such distance from the ground, it can preferentially preserve it there, floating in midair, even if the supporting structure is not yet in place.

The Dawkins WEASEL program commits this version of the fallacy.

War Games Fallacy, Version #2

In this version, there is no specific target house. Rather, the program has access to a generalized understanding of complete houses. (This is comparable to a program operating not according to a specific English sentence, but rather according to a dictionary of English words or phrases and an understanding of what is required to form a complete valid English sentence.)

Nevertheless, the program is still cheating by taking advantage of this knowledge of the complete future picture. If it discovers it is appropriate in general that a roof piece should exist at such and such distance from the ground, it can still preferentially preserve it there, floating in midair, even if the supporting structure is not yet in place.

If a program, such as Zachriel (cf. JT’s comments above) or others like it, preferentially preserves words such as “of” because they are in the dictionary and will eventually become useful in meaningful sentences (even though they are currently useless for meaningful sentence construction), then that program is committing the second version of the same fallacy.

Notice that the distinction between pursuing a specific predefined target or a generalized predefined target makes no difference to the error. The problem is that selection is allowed to take advantage of knowing future utility.

It is exactly like a movie in which a computer can supposedly crack a password by cumulatively selecting one character at a time — something that would be impossible to do without some kind of access to the final answer.

Because that was the goal of the program.

IOW, it was the design of the program to show how things can come about without design.

IOW, evo–taken to the anti-ID level — is irrefutably irrational.

It is exactly like a movie in which a computer can supposedly crack a password by cumulatively selecting one character at a time — something that would be impossible to do without some kind of access to the final answer.

There is no reward for guessing 99 percent of a safe’s combination.

tribune7:

IOW, it was the design of the program to show how things can come about without design.

The purpose of the program is simply to show that iterative randomness + selection can come to a target faster than randomness alone.

Zachriel’s program has been discussed to a limited extent before:

http://www.uncommondescent.com.....ment-90063

http://www.uncommondescent.com.....gen-daily/

http://www.uncommondescent.com.....ent-145598

And since we’re discussing English-based examples I’ll copy over my previous thoughts on this subject:

the Explanatory Filter can take multiple types of inputs (which also makes it susceptible to GIGO and thus falsification). Two are (a) the encoded digital object and (b) hypothetical indirect pathways that lead to said objects. My name “Patrick” is 56 informational bits as an object[each letter is represented by 8 bits]. My name can be generated via an indirect pathway in a GA. An indirect pathway in a word-generating GA is likely composed of steps ranging from 8 to 24 informational bits.

Let’s say you take this same GA and have it tackle a word like “Pseudopseudohypoparathyroidism” which is 30 letters or 240 informational bits. It can be broken down into functional components like “pseudo” (48 informational bits) and “hypo” (32 informational bits). Start with “thyroid” (56 informational bits). For this example I’m not going to check if these are actual words, but add “ism”, then “para”, and then “hypo”. “hypoparathyroidism” is a functional intermediate in the pathway. The next step is “pseudohypoparathyroidism”, which adds 48 informational bits. Then one more duplication of “pseudo” for the target.

That may be doable for this GA but what about “Pneumonoultramicroscopicsilicovolcanoconiosis” (360 informational bits) or, better yet since it’s more relevant Dembski’s work (UPB), the word “Lopado­temakho­selakho­galeo­kranio­leipsano­drim­hypo­trimmato­silphio­karabo­melito­katakekhy­meno­kikhl­epi­kossypho­phatto­perister­alektryon­opto­kephallio­kigklo­peleio­lag?io­siraio­baph?­tragano­pterýg?n” (1464 informational bits). I’m not going to even try and look for functional intermediates.

And I’d add that none exist, although the entire word consists of functional components. So someone could argue that an indirect pathway could duplicate all of them from other words and somehow assemble them into a coherent whole.

Hoki,

The purpose of the program is simply to show that iterative randomness + selection can come to a target faster than randomness alone.

Of course. The problem is natural selection knows nothing of targets; therefore, the whole exercise is invalid. See my question @65—which choice would you make?

Patrick and Clive,

Zachriel is about the most polite commenter in the blogosphere that I have ever come across. Why is he banned here?

Hoki–The purpose of the program is simply to show that iterative randomness + selection can come to a target faster than randomness alone.

So evolution requires a target? How would that target be determined?

StephenB:

Of course. The problem is natural selection knows nothing of targets; therefore, the whole exercise is invalid. See my question @65—which choice would you make?

Neither obviously.

…

Oh, you want a more exhaustive answer??? OK, you’re too teleological.

How about this for an example that BETTER represents natural selection: Look at the results from the weasel program in the opposite order. I.e. start with the string “METHINKS IT IS LIKE WEASEL” (the modern “evolved” string we have found) and work your way towards a (to us) totally random string (it’s distant ancestor), as such:

1. METHINKS IT IS LIKE A WEASEL
2. METHINKS IT WS LIKE A WEASEL
3. METHINKS IT WS WIKE A WEASEL
…
x. HKADHFKLUWHEKJLHWEHUREKJEWHR

Each string represents a sequence of letters that were more fit than the sequence below it under it’s CURRENT selective pressure. The trick is to realize that “METHINKS IT IS LIKE A WEASEL” was only the target (i.e. more fit) when the sequence was on the second line. The target when on the third line was “METHINKS IT WS LIKE A WEASEL” and the target when on the fourth line was “METHINKS IT WS WIKE A WEASEL”.

In this way, the original random sequence never strived for a distant “METHINKS IT IS LIKE A WEASEL”

Note: this is an ANALOGY, so it will be far from a perfect representation of natural selection (and evolution in general). We could always quibble about whether or not natural selection (or other evolutionary processes) actually could create a, e.g,. a human from something prokaryotish, but that is missing the point: you created a strawman version of how natural selection supposedly works.

Tribune7 (#82):

My post in 83 deals with your question. There is no real target as such (at least not one in the distant future). For natural selection there are current selective pressures.

Arthur Smith (#81):

“Zachriel is about the most polite commenter in the blogosphere that I have ever come across. Why is he banned here?”

Can I join in asking that Zachriel may post here? (If he wants, obviously). I had a long exchange with him on another blog, and I agree that he is polite, but the most important thing is that he is very intelligent. And there is nothing better than an intelligent adversary.

Regarding the discussion here about the weasel algorithm, I must admit that I cannot find it really interesting. Basically, I agree with Gil: an algorithm which already knows its target is of no interest for our debate here. I have always found Dawkins’Weasel a silly issue, however the algorithm may work in the details. And anyway, if we really have to discuss the algorithm, we should know it exactly (I can’t see why Dawkins has not made it publicly available). Frankly, trying to recreate an unknown algorithm which anyway does not demonstrate anything seems not a very fruitful task (although I must commend Atom for having the patience and goodwill to try that just the same).

Zachriel’s word games are another thing. Unfortunately, I have never had the time to study them well, but I am sure that they demostrate something. They probably demonstrate that it is possible to easily explore a semantic space with an algorithm well designed to explore semantic spaces, but I am not really sure of that. But I will give only two comments here about that:

1) The protein functional space is not a semantic space: function, in proteins, is not linked to a symbolic meaning, but to biochemical activities.

2) Although Zachriel’s algorithm can explore the space of words and even phrases, it still cannot really generate original language and original meanings. Language and meaning are still the product of conscious beings. Machines can only copy them, remix them, and recognize them only passively. In other words, a program can recognize that a phrase has a meaning only if it already “knows” (IOW, it has been inputted in it) that it has. The program, being a machine, has no concept of meaning. It can only match something with what it already “knows” in its code. Conscious intelligent beings are all another thing (although AI enthusiasts very much like to believe differently). So, after all, even Zachriel’s programs are probably in Dawkin’s category (they already know the answer they are searching), even if they certainly are much smarter.

Hoki–The purpose of the program is simply to show that iterative randomness + selection can come to a target faster than randomness alone.

IOW, it is sort of like someone taking the time to write a program illustrating that a man will take fewer steps walking 10 feet than he would 10 miles.

The only think definitively illustrated by Dawkins’ program is its pointlessness.

tribune7:

The only think definitively illustrated by Dawkins’ program is its pointlessness.

Remember that an old creationist argument was that evolution was equal to a hurricane blowing though a junkyard to produce a 747. Considering how many people fail to understand the difference iterative selection can make to randomness, it is anything but pointless.

Hoki — Remember that an old creationist argument was that evolution was equal to a hurricane blowing though a junkyard to produce a 747.

Actually, that’s an atheist argument. At least Sir Fred Hoyle was an atheist at the time he made it

tribune7, good point regarding Hoyle. Even though he never accepted that the intelligence was God, he and Wickramasinghe were pioneers in pointing out the absurdity and the mathematical unreasonableness of undirected evolutionary explanations for biological life.

It is an inconvenient truth that inference to intelligent agency, even among the pioneers in our time, was not confined to theists.

Hoki:

Considering how many people fail to understand the difference iterative selection can make to randomness, it is anything but pointless.

How many more people fail to understand the difference it can make for a program to illegitimately “assist” its selections by looking at future utility?

You are right to point out that actual selection works according to current pressures. Did you think that anyone doubts this? Or that anyone doubts that this type of selection is real?

The problem is that programs such as Zachriel (or WEASEL) create an illusion of heightened effectiveness that does not work within the confines of true natural selection.

But the fact that real natural selection helps (when there are incremental advantages to select) is not questioned.

A key aspect of Weasel, et al. is that they are population based searchers. If you dial down the population size to 1, Weasel is the same as random search. The benefits of selection only appear when the population is set higher.

What is surprising about Weasel is not that it reaches the goal, but how fast it reaches the goal, even for very small population sizes.

Pendulum

Actually, if Weasel were to bump up mutation rate and population size, within the performance capacity of the PC, it would outrace deterministic search by getting double, triple and so on finds in each generation. (That is because configs that would be way out in the skirts of the population would begin to show up in “random” samples, and by the distance to target metric, would consistently win and advance.)

That is, Weasel is utterly unconnected to the real world, where the issue is that targeted search that rewards non-functional configs based on mere increments in proximity to a target, have nothing to do with the alleged designing capacities of chance variation plus natural selection across competing FUNCTIONAL sub populations.

GEM of TKI

hoki:

There is no real target as such (at least not one in the distant future). For natural selection there are current selective pressures.

No target = no search. Just the “necessity” of survival.

Pendulum: “The benefits of selection only appear when the population is set higher.

What is surprising about Weasel is not that it reaches the goal, but how fast it reaches the goal, even for very small population sizes.”

On the contrary, it is not surprising at all once one realizes that WEASEL selection is not based on true selection according to current function.

In the second run in the original post, notice that even by the second entry in the sample we are given, the sequence “LIKE” is already being preferentially preserved, even though it is not yet functioning as a word.

The WEASEL selection is cheating by taking advantage of recognizing and selecting according to the function it will eventually have. This is possible only through access to knowledge about the ultimate solution.

Given that it can peek at the future utility, of course it is able to quickly converge.

Instead of using a dictionary of English words, a program such as Zachriel could have used a dictionary of proteins found in the human body represented as character sequences. However, the fundamental flaw in the approach is the same.

Supposing that it discovered a sequence that matched one of the entries in the dictionary of proteins. Then, since this is one of the “words” in the dictionary, Zachriel would preferentially preserve it.

However, proteins have function in the human body in the context of cooperation with other proteins. Many of these proteins would be useless if taken in isolation, without the cooperating proteins they are interdependent with.

By selecting for any word in the dictionary, Zachriel is selecting according to potential value — value as it will be in the context of the future whole, once one has the complete language or the whole body. It is not selecting according to current utility or real advantage conferred with regard to current pressures.

EricB:

Did you think that anyone doubts this?

Yes. Did you read SteveB’s comment to which I responded? (my apologies to StephenB whom I originally thought had written that post)

Lots of ink has been spilled here about the goal/target of the weasel program. Make no mistake about it: the program would not work without a future target specified by Dawkins or his programmers.

Consider this fact in light of the quotations below (my emphasis). IMO, what appears to be happening here is that goals, targets and the like are expressly not allowed when evolution/NS is discussed at the 30,000 ft-level, but then teleology gets smuggled into the bowels of the weasel program–by Dawkins, not by me–when it suits his needs.

Adopting this view of the world means accepting not only the processes of evolution, but also the view that the living world is constantly evolving, and that evolutionary change occurs without any ‘goals.’ The idea that evolution is not directed towards a final goal state has been more difficult for many people to accept than the process of evolution itself. (Life: The Science of Biology by William K. Purves, David Sadava, Gordon H. Orians, & H. Craig Keller, (6th ed., Sinauer; W.H. Freeman and Co., 2001), pg. 3.)

The ‘blind’ watchmaker is natural selection. Natural selection is totally blind to the future… Humans are fundamentally not exceptional because we came from the same evolutionary source as every other species. It is natural selection of selfish genes that has given us our bodies and brains … Natural selection is a bewilderingly simple idea. And yet what it explains is the whole of life, the diversity of life, the apparent design of life.(Richard Dawkins quoted in Biology by Neil A. Campbell, Jane B. Reese. & Lawrence G. Mitchell (5th ed., Addison Wesley Longman, 1999), pgs. 412-413.)

Nothing consciously chooses what is selected. Nature is not a conscious agent who chooses what will be selected… There is no long term goal, for nothing is involved that could conceive of a goal.(Evolution: An Introduction by Stephen C. Stearns & Rolf F. Hoeckstra, pg. 30 (2nd ed., Oxford University Press, 2005).)

Dawkins developed Weasel 30 years ago to demonstrate the difference between entirely random and cumulative selection. He said himself “it was a bit of a cheat” and, unlike Weasel, evolution is an untargetted process, blind to the future.

It is a piece of history. Evolutionary theory does not rely on Weasel for support. Pointing out the obvious, that it is not an analogy for natural selection, will not have any impact on the validity of evolutionary theory.

@ Gpuccio

I agree with your assessment of Zachriel’s intellect. I should have mentioned it.

KariosFocus

That is, Weasel is utterly unconnected to the real world, where the issue is that targeted search that rewards non-functional configs based on mere increments in proximity to a target, have nothing to do with the alleged designing capacities of chance variation plus natural selection across competing FUNCTIONAL sub populations.

Still creating strawmen to destroy?

Perhaps it would have been better for you, overall, if you had just said way back when “why yes, George. The letters do appear to latch but that’s because of the small sample size I’m looking at and not any explicit latching code”

I mean. There is now a thread at Pandas thumb talking about how Dr Dembski is finally going to get round to publishing a “correct” version of Weasel, finally, on evoinfo.

And of course we have the upcoming paper to look forwards to which it’s suspected uses the “incorrect” (or latching) version of Weasel in some way.

I notice you have not made an apperance at the thread discussing the actual math behind Weasel either at PT or ATBC

http://www.antievolution.org/c.....=14;t=6034

http://pandasthumb.org/archive.....-para.html

I look forward to seeing you at either of those places Kariosfocus. If you want to discuss the math further, go there. If you want to repeat yourself ad nauseum, stay here. Under discussion is the probablity of a correct letter appearing and not changing. If you are a true believer in knowledge, in taking the chance that you might be wrong, you might learn something.

And on the subject of ad nauseum, FSCI, instead of talking about it and typing tens of thousands of words defending it to people asking the same questions over and over where you give the same answers over and over, why don’t you do something with it instead? You know, use it?

Put your money where your mouth is.

Publish.

The Panda’s Thumb post is interesting. Here’s what Ian Musgrave says:

I’ve gone back and done a head to head comparison myself between a program with no “locking” (all letters in any given string have a chance to be mutated) and one with “locking” (where the matching letters are preserved against mutation). Trying to implement “locking” al la Dembski proved too hard. You have to keep indices of the letter locations and keep updating them. It is such a pain in the bottom to try and do this that I cannot imagine Dawkins even wanting to try and program a “Locking” implementation in GBASIC. Remember, Dawkins weasel was a quick and dirty program bashed out in a short time. To implement “locking” I just kept a copy of the parent string unmutated (after all, in the real world not every offspring has mutations in genes of interest).

His conclusions?

“Locked” runs finished earlier, on average [sic] but most of the trajectory of the run was determined by mutation supply. As you can see, runs done with locked and unlocked versions fell within the error bars of each other, for runs that set the Offspring number at either 100 or 30.

He says some other things, but I’ll leave it to others to go there.

Does Ian Musgrave consider that locking takes place regardless of any specific coding given a target, a small enough mutation rate and a large enough sampling size?

That alone ensures that at least one exact copy of the “parent” will be present in each generation.

I have already gone over this. So what is so special about Ian?

Arthur Smith:

Pointing out the obvious, that it is not an analogy for natural selection, will not have any impact on the validity of evolutionary theory.

No but pointing out that “evolution is an untargetted process, blind to the future“, does have an impact on its validity.

And that is why no one can model “evolution”.

It is an untestable concept- “untargeted process, blind to teh future”.

It may make for interesting philosphy but it doesn’t belong in a science classroom.

You rock Arthur! Thank you.

SteveB [65]:

wsoldin_je_d_swodx
Now, consider the following “mutations” to the original, A and B.
A wsocdin_je_d_swodx
B wsoldin_je_d_srodx
You, representing natural selection, do not know the target—in fact, you don’t even know what a target is, as you have no goals. Additionally, you cannot predict, cannot reason, and certainly cannot write computer code; you don’t have any lifelines and you can’t buy a vowel. All you can do is evaluate the “fitness” of each variation in the here and now. Your options are: select A, B, both, or neither.
Which do you choose and why.

Let’s say the original “wsoldin_je_d_swodx” is an eye.

A) distorts the lens incrementally resulting in incrermental increase in image resolution.,
B) results in a corrosive acid being introduced into the eye fluids.

————————–

Patrick [79]:

That may be doable for this GA but what about “Pneumonoultramicroscopicsilicovolcanoconiosis” (360 informational bits) or, better yet since it’s more relevant Dembski’s work (UPB), the word “Lopado­temakho­selakho­galeo­kranio­leipsano­drim­hypo­trimmato­silphio­karabo­melito­katakekhy­meno­kikhl­epi­kossypho­phatto­perister­alektryon­opto­kephallio­kigklo­peleio­lag?io­siraio­baph?­tragano­pterýg?n” (1464 informational bits). I’m not going to even try and look for functional intermediates.
And I’d add that none exist, although the entire word consists of functional components. So someone could argue that an indirect pathway could duplicate all of them from other words and somehow assemble them into a coherent whole.

Those words above would never be specifically targetted themselves. There’s only incremental increases in word length while maintaining legal words. Any conceviable word of the length above would be just as amazing as any other. And actually the way words of that length can be directly parsed and understood from their latin components to discern their meaning, a lot of the words randomly formed from such roots would be acceptable as descriptive of something, whether or not anyone had actually coined them yet or not (IMO). So when you get to a word of that length – its really like a sentence only with much less strict syntax.

———————-

EricB:

Let’s start just with english (and not consider biology for the moment): Do words like “of” have meaning on their own? Clearly they do:
If… – movie title; “One” – song “Yes” – group. Think of album titles, advertizing slogans and campaigns and so on. Phrases and sentence fragments are used all the time to convey meaning. Infants and toddlers also use words and sentence fragments to convey meaning and are understood. In normal conversation, people use fragments as well. Poets will slice and dice language. It is really an arbitary convention that legal sentences have to have a subject and verb for example. Groups of words and even single words most assuredly do have meaning in English.

And complex words and sentences can be built up at random very easily from these smaller components in English – with no planning involved for future use. This results from the fact that smaller words and phrases can be used in millions of different ways, so that utilizing one phrase isn’t locking us in to only one future target. (And just to reiterate, all the individual words and small phrases are useful on their own as well).

So the applicablity to nature would be is if it as well had elementary components that were functional on their own and were functional in many other contexts as well. If this implies nature would have to be much like a human language itself, then there’s your teleological argument right there, and so be it.

Let’s start just with english (and not consider biology for the moment): Do words like “of” have meaning on their own? Clearly they do:
If… – movie title; “One” – song “Yes” – group.

(I once heard some old-timer mention the above titles from some bygone era.)

JT [105], I’ll add to that list “A” — the title of an important long (800 page) poem by Louis Zukofsky. Also, Zukofsky’s collected critical essays are divided into three sections, entitled “for,” “with,” and “about.”

JT, I’ve never claimed there are no single words that might be justified as having functional value in isolation. The problem is that Zachriel presumes to treat any word it finds in the dictionary as having functional value in isolation.

Dictionaries are constructed from usage within full blown English taken as a whole. This means that Zachriel is illegitimately basing selection upon a word’s value in the context of the entire English language. That is cheating by looking at future utility, whereas as Arthur Smith said, “evolution is an untargeted process, blind to the future”.

You wrote:

“Let’s start just with english (and not consider biology for the moment)…”

If you think Zachriel has any relevance for biological evolution, why not consider biology? That is where its error becomes the most obvious.

I expect you know that proteins that have function in the whole human body may nevertheless be useless when taken in isolation, apart from the support of certain cooperating proteins. It is plain that if Zachriel selected one of these in isolation, based merely on its appearance in a dictionary of functional protiens, that this would be cheating as an analogy to biological evolution. It selects based on future function in the eventual whole body.

Do you disagree with that analysis? If so, where is the biological case mistaken? Yet this is exactly what Zachriel does with English words within the English language.

The fact that Zachriel uses not just one but a whole dictionary of many available selectable targets does not save it from the plain fact that its relevance to biological evolution is an illusion. The problem is not removed by having more targets to select. One of its core problems, just as with WEASEL, is the fact that it makes no discrimination between present and future utility.

David Kellog:
JT [105], I’ll add to that list “A” — the title of an important long (800 page) poem by Louis Zukofsky.

What about the flavor enhancer, “Accent”.

That’s not even a character.

Arthur Smith, it seems at least that JT is not yet convinced. I suspect there are others who do not realize that these programmed word games are irrelevant as an analogy to biological evolution.

But if we sincerely only wanted to show that cumulative selection that is unrelated to biological evolution is more effective than random selection, we wouldn’t need a program for that.

To find a password of N letters by chance is to choose from 26^N possibilities (or 52^N if case sensitive).

If instead I can discover them one at a time in isolation and cumulatively select them, even a brute force search need only consider 26*N possibilities (or 52*N for case insensitive) to completely examine all possibilities.

As N grows, the difference becomes huge, and as you know proteins are quite long. The advantage of non-biological, cumulative selection (where present function is not a concern) is beyond dispute.

However, that is irrelevant to evolution of proteins, due to the nasty problem of needing present selectable function on the whole protein, combined with the predominance of amino acid sequences that are not functional as proteins.

ericB [107]:
The problem is that Zachriel presumes to treat any word it finds in the dictionary as having functional value in isolation.

What are you saying? That supposing some hypothetical word is difficult to interpret as “functional” that invalidates the entire illustration?

Dictionaries are constructed from usage within full blown English taken as a whole.

I honestly don’t know what you mean. Some people communicate quite adequately with a vocabulary of a couple of thousand words or so (I suspect – maybe less) – children and toddlers with even far less. Also are you saying there was never a time when the vocabulary of English was not far far less than it is today? Did words have any meaning back then? What about the limited vocabuluary in the vocalizations of animals. Meaningless as well?

I expect you know that proteins that have function in the whole human body may nevertheless be useless when taken in isolation, apart from the support of certain cooperating proteins.

Obviously subcomponents have to be functional enough to survive before being incorporated into larger functions. I would tend to think in early stages of life there wouldn’t have been a lot of competition, so a lot of niches. A lot of neutral genetic info squirelled away in various corners not hurting anybody and slowly multplying.

It is plain that if Zachriel selected one of these in isolation, based merely on its appearance in a dictionary of functional protiens, that this would be cheating as an analogy to biological evolution. It selects based on future function in the eventual whole body.

I don’t know anywhere Zachriel goes to a dictionary of functional proteins and tries to establish a strict parallel with the English Language. If you’re just stating that its not proven yet nature is like human language no one’s said otherwise. But the idea that human language would be more powerful and more expressive and more flexible than nature itself seems hard to believe.

The fact that Zachriel uses not just one but a whole dictionary of many available selectable targets does not save it from the plain fact that its relevance to biological evolution is an illusion

So now you’re just asserting that its relevance to biology is an illusion. That’s a type of argument from ignorance as well – “Until conclusively proven I will assert its an illusion.”

One of its core problems, just as with WEASEL, is the fact that it makes no discrimination between present and future utility.

This is just categorically false. English words do have utility in isolation – and utility in many sorts of combinations. No one has purported to prove as of yet that nature is identical to human language, however.

BTW – it occurs to me to ask whether you’ve actually reviewed those websites I mentioned previously [57], since you’re discussing this.

ericB – I don’t want to be an unequivocal advocate until having more time to review all this. 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.

You seemed irritated or antagonistic towards these ideas, but to me they were at least interesting if not compelling.

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.

That would be a complete waste of time- obviously it would work.

Onlookers:

Re, GLF,100 : Still creating strawmen to destroy?Perhaps it would have been better for you, overall, if you had just said way back when “why yes, George. The letters do appear to latch but that’s because of the small sample size I’m looking at and not any explicit latching code”

Selective hyperskepticism, turnabout accusation form. And ecvidently he is broadcassting a twisted wversion of whart has happened here at UD to his co-advocates of evo mat. About pa\r for their course I’d say.

So, let us correct, for those interested in the truth of the matter:

1 –> The above is in response to my: “Weasel is utterly unconnected to the real world, where the issue is that targeted search that rewards non-functional configs based on mere increments in proximity to a target, have nothing to do with the alleged designing capacities of chance variation plus natural selection across competing FUNCTIONAL sub populations.”

2 –> My summary — since Dec last — is an accurate summary of Dawkins’ own description as cited many times on Weasel from BW, ch 3. In short, there is no strawman. At least, on my part. For, targetted search that rewards mere proximity of non-functional configs ducks Hoyle’s challenge of getting TO shores of islands of functionality, and is utterly irrelevant to the vaunted BLIND watchmaker of natural selection. this last, is said to select by differential success of functioning life forms. No function, no differential reproduction; no natural serelction. WEASEL is not about the blind watchmaker but instantiates the power of intelligent design, here pursuing a rhetorical ultimate purpose. So, we must not let turnabout rhetoric distract us from the plain truth of the matter on the merits, regardless of the outcome on explicitness of observed o/p latching circa 1986.

3 –> Next, the claimed “small” sample size constitutes 300+ letters that could change [count 'em above], with 200+ being latched on output; never does a correct letter revert in he sample; which is avery low probability event indeed if the Weasel of 1987 is more representative of output circa 1986. On inference to best explanation, the output as published is a likely “good summary” on the Weasel’s performance, and it with high probability implies the outputs are latched. So, GLF is here distorting the record on the o/p of the 1986 Weasel pgm and the extent of the sample used form Mr Dawkins’ published “good” results. 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.

4 –> From the previous thread, 346 – 7, it will be evident that I proposed two alternative models for that observed o/p latching I noted on since Dec last: T2 — explicit latching, and T3 — implicit latching. On the 1986 o/p alone explicit latching is the best — most parsimonious — explanation.

5 –> On the principle of charity on testimony, I have accepted that per recent reports first published here in this thread above, Mr Dawkins did not EXPLICITLY latch. But that simply means that on further best explanation, he implicitly latched. The observed latching is the dominant feature of the o/p circa 1986, and in fact the implicit latching thesis has the problem that the letters that flick back circa 1987 then rapidly revert, hence the “winking” effect. On mere probability, that should not be likely.

6 –> Further to this, Apollos has shown that explicitly latching code can also revert temporarily. That is, only credible code is decisive, though on the evidence we can draw some fairly suggestive inferences.

7 –> E.g. as discussed above in this thread, the co-tuning of mutation rates [about 5% will give 1 - 2 letters per member of a generation] with an optimised population size will give effective latching, as 24% of all members will be no-change at that mutation rate with a string length of 28 characters. then,w e see that on observation the reported runs take 40 – 60+ gens to reach target for “good” runs. This means that > 50% of time, no-change wins in a generation.

8 –> Choose a population scaled so that we will get some typicality, perhaps 20 – 30 at base [just off law of large numbers as a pop size to start from], and small enough so that out in skirt cases of 1, 2 or 3 correct letters do not dominate outcomes, and you will [a] latch correct letters, and [b] progress to the target in the sort of range as published instead of within 30 gens.

9 –> So, the evidence is that the primary issue is settled independent of the debate [and I almost never use this in a positive sense; this is no exception] we have had over whether or no Weasel explicitly or implicitly latched the o/p circa 1986. Weasel is simply not blind watchmaker, period. And, with a sample size of 200 latched letters in 300 chances for letters to change, we are seeing a dominant pattern of o/p latching. On the current balance of evidence [giving heavy weight tot he reported testimony from Mr Dawkins] I am willing to go along with implicit latching, though I would like to see an explanation of the winking effect on flickbacks of correct letters circa 1987.

____________

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.

So, let us note the distruptive pattern of his threadjacking behaviour, and not reward him by followig the schools of wriggling red herrings lead out to ad hominem soaked strawmen ignited to cloud and poison the atmosphere required for progressive discussion.

GEM of TKI

PS: As for antievo and pandas thumb, I refuse to have anything to do with these notoriously disruptive and destructive fever-swamp of evo mat sites; strong cases in point of precisely the kind of destructive selective hyperskepticism and rhetorical manipulation and championing of the morally indefensible that is ripping our civlisation apart. If there is any merit in their arguments, GLF must present them here — without the nasty ad hominems, strawmen and red herrings, not to mention cherry-picked one sided analyses that major on minor distractive points to duck the main issue. (Indeed, the very Weasel program itself is a case in point: targetted search is precisely a begging of the question Hoyle et al have raised — getting TO shores of functionality, before hill-climbing to optimality can proceed. rewarding target- proximity of confessedly non-functional strings is exactly what should not have been seen, but was.)

Mr Kellogg, re 102:

Mr Musgrave is misleading you. (Utterly unsurprising for PT.)

Look at Apollos’ code form the previous thread and you will see just how short, relatively speaking an explicitly latched Weasel is. I don’t know if the EIL code is there at the site, as Atom can confirm.

In fact, Dawkins states that within each generation the closest to target is preserved. That means that he code has to have in it an assessment of proximity, which is going to be based on relative locations of letters to the target in any case. You already have the information you need to latch, just to get to the champion per generation.

Also, on the math at EIL, latched searches — whether explicit or implicit makes no difference — will as median run to 98 generations. This is consistent with “good” runs of 40 – 60.

GEM of TKI

PS: JT at 105: Do words like “of” have meaning on their own? Clearly they do: Without a much wider context,t eh glyphs or sounds implicated are meaningless. That wider communicative and indeed social context is massively based on intelligence and intentionality.

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/c.....=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.

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/c.....ntry140135

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.

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 TKI

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.

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?

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.

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:

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.

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.

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.

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?

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?

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?

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:

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,

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?

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.

JT

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.

[ . . . ]

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.

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!

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.

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:

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!)

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!

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.

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.

hi,
there seems to be a bug in the board, I post but nothing appears and no message about moderation either.

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.

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?

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!

Moderators, I posted this in the wrong thread: It should go here. Could you kindly delete that version and post this one? For once I am happy to have a post moderated.

I am reproducing a section of Wes’s refutation of kf’s response, with the names changed:

[kairosfocus] is plainly confused about a great many things.

His own choice of analogy provides a great example of an IDC advocate shooting himself in the foot. The dart and chart situation that would have some slight analogy with “weasel” isn’t one where all the dart-holes count; it would instead be like dropping the dart N times, and then only recording where, say, the right-most dart-hole of those N holes occurred, and shifting the chart to center the drop-point on that hole for the next set of N dart drops. Obviously, the larger N is, the less likely the chart will be moved to the right and not to the left for re-positioning. [kairosfocus]‘ assertion is like saying that we should expect one or more right-ward shifts of the chart during a process where N=50 dart drops, and the chart is re-centered on the right-most dart-hole after each N drops some modest number of times. [kairosfocus] is obviously clueless; the “law of large numbers” is a perfect counter to his argument, not a vindication of it.

300 characters in a printout of the best candidates per selected generations are not subject to change, not unless one is applying a 100% mutation rate. For a reasonable mutation rate of 4%, one would be greatly surprised if just 24 characters in so many ordinary, not best, candidates had actually undergone change of any sort in the unknown derivation from their parent strings.

But the whopper in [kairosfocus]‘ maunderings is the bland assertion that “we have reason to believe” [the sampled best candidate strings from various generations] “are credibly uncorrelated to the system”. No, [kairosfocus], we have a tremendous expectation that those results are “correlated to the system”: they are the result of a selective process from a population of candidate strings, taken from a pool of N such candidates at each generation. And the odds for very reasonable population sizes and mutation rates are that best candidates from N alternatives are strongly in favor of keeping “correct” bases from the parent string unaltered. I specifically analyzed the situation, provided the equation that delivers the probability at issue, and [kairosfocus] cannot be bothered to address those facts.

Joseph, the issue has been “latching”, not “racheting.”

And kairosfocus’s point that the program is intelligently designed is pointless – so is every computer program ever written. Does this mean that computer simulations can never be used to illustrate or explore some aspect of the real world?

kairosfocus @ 138
Taking your points in turn…

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.

The Blind Watchmaker, as you say, is the book and discusses many aspects of Dawkins’s major thesis. The WEASEL program is not the book. It is an illustration of one aspect of what is discussed in the book, namely, the power of cumulative selection. Whether or not you agree with the main thrust of the argument in TBW should make no difference to whether or not the WEASEL program is an effective illustration of just one point in the book.

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.

Yes, there is a difference between between your initials and your name. They are three letters which, in the search for your full name, are latched in place.

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.

An open website is public domain regardless of the volume of traffic.

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.

It would only be identity theft if someone were to use your name and pretend to be you for personal gain. No one here has done that as far as I am aware. Simply using your own name to refer to you is not an offense nor is it even a breach of your privacy. Your name is out there as I found in a simple Google search which retrieved this article by one Ian Boyne in the Jamaica Gleaner which begins comments about you as follows:

But G***** M*******, well-meaning but with a surfeit of zeal over knowledge, implies that there is a necessary conflation between theology and action.

Note that Ian Boyne did not asterisk out your name, I did.

JT (132), I commend you on not stopping with your first impression of what you thought I was saying. You continued to think about it and eventually arrived at a more accurate understanding. I certainly accept your apology.

Yes, if any of these programs were falling under the combinatorial explosion, they simply wouldn’t work. See my comments about random typing and Shakespeare’s sonnet, a point that Dembski has made.

In principle, there can be legitimate ways to evolve incrementally with function based selection giving a huge boost over random search, though unsurprisingly this is subject to conditions, restrictions, requirements and constraints. I know about software and I know there are also illegitimate ways to rig the system so that one gets an unwarranted advantage, i.e. one that steps outside the limitations one supposedly is modeling.

So the question is, does a program like Zachriel play within justifiable bounds, or does it step across a line and presumptuously take an unjustified advantage?

Just to be clear, I’ve never supposed that Zachriel is driven by proximity to a single fixed target. I know that Zachriel employs (among other things) a dictionary of possible selectable words, not just a single fixed string. Nor is that distinction (one vs. many) a necessary part of the issue I am raising. I am primarily talking about something else, namely the War Games fallacy, as I explained back in post 75.

Please take a look again at post 75 and let me know what part of that post seemed unclear. Thanks.

In contrast with the illusions of evolutionary word programs, here are two positive examples, one actual and one hypothetical, that do not commit the same errors.

Some have used genetic algorithms to explore possible new designs for improved antennas. Starting with one or more initial functional designs, the ability to modify these in various ways (perhaps change the lengths or angles of the arms, add forks, etc.) , and the ability to analyze the expected effectiveness of the results, such an algorithm can explore the space of possible antennas defined by the fitness function it has been given. It can look for optimal points as defined by that function.

Notice that it is doing so by evaluating function at each stage (albeit as defined by the provided function and within the limits of the modifications it can make).

One can also design software to compete in games. Though I cannot immediately provide a link, I believe I’ve seen games based on battles of software tanks. One could make an evolutionary simulation in which one starts with one or more programs for such a software tank, makes truly blind and undirected random modifications to the software, and then sets that generation of tanks into competition. The top surviving tanks are reproduced, potentially with additional modifications.

Notice that this also would have a genuine evaluation of competitive function / fitness as the criteria for preservation and propagation.

Typically, word games such as Zachriel or WEASEL are illusions in that they do not have a legitimate way of defining preferential preservation in terms of current actual effective function.

Since their basis for selection is not based on current function, even as examples of (non-functional) “cumulative selection” they are irrelevant to discussions of biological evolution.

Biological evolution simply cannot operate as they do.

hazel,

Have you ever seen a ratchet in operation?

It is a LATCHING mechanism that allows for the ratcheting.

George:

As you have the book at hand, could you provide a quote that supports this?

Well George, first get yourself a dictionary and look up the word cumulative:

1 a: made up of accumulated parts b: increasing by successive additions

There is a big difference, then, between cumulative selection (in which each improvement, however slight, is used as a basis for future building), and single-step selection (in which each new ‘try’ is a fresh one).- Dawkins TBW page 49

That’s ratcheting.

hazel:

Does this mean that computer simulations can never be used to illustrate or explore some aspect of the real world?

In order for a computer to simulate anything the programmer(s) need to have COMPLETE knowledge of that they are trying to simulate.

For example we can simulate flying in an airplane because we understand flight well enough to do so.

With biology we don’t even know what is responsible for the development of our eyes/ vision system so we have no idea how to simulate that.

We have no idea what makes an organism what it is.

We have no idea how many mutations can accumulate and what that accumulation will do.

So again we can only simulate that which we understand.

Seversky,

The problem with cumulative (natural) selection is that it only exists in the head of Dawkins.

The power of cumulative (natural) selection has NEVER been demonstrated in nature.

Take away the target and cumulative (natural) selection is nothing more than a blind, random, meandering walk, right off a cliff.

We have COMPLETE knowledge of flying.

That’s news to me.

In general, what you say is not true. Every day people program simulations based on mathematical models of real-world phenomena of which we have very incomplete knowledge.

The computer simulations give us results that we might never expected with the computational power of the simulation.

However, and this is critical, we then have to take those results and test them back in the real-world to see if our model is sound enough. If the results don’t match the real world, we go back and refine our
model.

To state the obvious, if we had COMPLETE knowledge of something we wouldn’t need to create a simulation.

Duh.

Joseph,

I’m sure you’ve seen the example of antenna design by means of evolutionary algorithms.

http://ti.arc.nasa.gov/project.....ntenna.htm

Do you agree that:

1) It works

2) The process does not involve “ratcheting” toward a target specified beforehand.

Note: I’m not asking whether you believe that this application accurately reflects what happens in nature.

hazel:

My bad.

I should have said all simulations are only as good as our knowledge.

So if we take our knowledge of biology computer simulations of any evolutionary aspect wouldn’t be very impressive.

madsen,

1) Yes it works

2) Target 1- optimize the Yagi-Uda. No idea how that was handled. But they must have tested the mutants against something. And yes I do understand in antenna design sometimes one-step back- a tweak one way- is required to get a bigger step forward- another tweak or two someplace else.

target 2 – more optimizing of a design

But anyway, what is your point?

This thread is about Dawkins and weasel.

I know people can write programs that can do things.

What no one can demonstrate is nature, operating freely doing something like that.

Thanks, Joseph, and I agree with you that real biology at the genetic level is too difficult for comprehensive, meaningful simulations both due to the huge amount of factors involved and to our lack of understanding.

That doesn’t mean that someone hasn’t, or can’t, write simulations of some aspects of genetic biology. As I explained before, all simulations work by modeling some simplified version of reality, after which one goes back and tests the results of the simulation against reality. In the results are verified empirically that gives one the confidence that the model has some merit, so that one can refine it a bit. If the simulation results don’t match reality, then you change the model until they do.

Dawkins “weasel” and similar programs are meant to test and illustrate a point. They are not meant to model real biology. Dawkins knows that, I’m sure.

Joseph,

But anyway, what is your point?

This thread is about Dawkins and weasel.

I know people can write programs that can do things.

What no one can demonstrate is nature, operating freely doing something like that.

Did you look at the ST5 antenna? There was no ratcheting toward a pre-specified target. Rather, just mutation and selection were involved.

Furthermore, the two (quite different) best resulting designs were superior to human-designed antennas in certain respects.

The reason I am bringing this point up in the weasel thread is that I want to understand more about this notion of ratcheting toward a specified target you have brought up. In particular, I would like to see if we can at least agree that evolutionary algorithms work even in the absence of such targets, the ST5 antenna being one example

madsen (and Joseph), first thanks for bringing up the antenna design example I was just alluding to.

As a quick comment, these evolutionary algorithms work within the limits of

a) the modification/mutations allowed by the programmer, and
b) the fitness function defined by the programmer to evaluate each of the resulting candidate designs.

This combination defines a landscape with peaks and valleys. The programmer doesn’t know where the peaks are in advance. (The working of the algorithm is essentially a search to find highest points in that defined landscape.) However, everything they have specified does indeed specify where those peaks are.

Whether you want to call that a “target specified beforehand” or not is a question of defining terms. But they have specified a landscape and at best the algorithm only finds the peaks (maxima) or else finds the low points (minima) that are determined by the model implemented.

Hi ericB,

Whether you want to call that a “target specified beforehand” or not is a question of defining terms. But they have specified a landscape and at best the algorithm only finds the peaks (maxima) or else finds the low points (minima) that are determined by the model implemented.

Of course I agree that the fitness function determines any maxima and minima in the fitness landscape. But the programmer doesn’t actually have to know the locations of these points ahead of time for the algorithm to work. I think we both would agree to that.

However, in view of this statement of Joseph’s:

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.

I wonder if he accepts that EA’s can solve problems without the programmer explicitly telling the computer what the answer is beforehand.

Also, to clarify and be fair to Joseph, I do realize he was referring to natural selection and the weasel program, and not to EA’s in that quote.

Onlookers:

It is now sadly evident that the much over-used evolutionary materialist advocate selective hypersketicism threadjacking tactic of red herrings led out to ad hominem soaked strawmen and onward to ignition that clouds and poisons the rtmopsphere for sertious discussion has reached the stage of turnabout accusations in this thread.

Such distractive tactics also inadvertently expose just how threadbare the evo mat case is.

In the case of Weasel, it is now abundantly clear that it uses a targetted, proximity based search strategy that rewards non-functional configs on proximity without reference to functionality. So, it is yet another misleading icon of evolution.

That is so whether or no it explicitly or implicitly latches or near-latches, and it is so in more modern Genetic Algorithm search simulations that are just as much characterised by active information fed in by designers.

Further, as it is plain that the evo mat advocates here and at the Anto Evo site are intent on personalities, rather than the merits, no further reasonable discussion with them is possible; sadly.

(And if these do not know the implications of trumpeting personal information to one and all online and/or linking one side of a dispute in which the Newspaper in question had to publish a corrective, they are a lot more naive than we have any reason to believe. So, let us learn from their behaviour, the dangers posed by their agenda and its underlying undermining of intellectual and ethical responsibility, to our civilisaiton. then, let us act resolutely in the defense of that which we hold dear.)

Let us now focus on correcting the misleading arguments they seem to be ever so fond of, as that will help us in defending ourselves from the epidemic of selective hyeprsketicism that has so many influenced by evolutionary materialism firmly in its grim jaws.

1] On the law of large numbers and practical identification of trends

You will observe that, above, I have repeatedly adverted to a key sampling theory concept, the law of large numbers.

In effect, as a rule, once we have sufficient samples of a population in hand and no reason to suspect undue bias, the trends of he population will come out in the sample, with pretty fair reliability. A typical threshold sample size for a linear [and that can be curvilinear too . . . ] trend is about 5 – 9 points [recall from school lab exercises] and for more stochastic situations 20 – 30 or so. Once you are much under that sort of range, you have to begin to resort to more and more exotic testing procedures. Of course as samples cost time, effort and money, there is also a practical upper limit.

In the case in view, we are dealing with a sample of scale 300 or so, with a trend that comes out in about 200 cases, with NO counter-instances.

That’s a strong trend in anybody’s book.

AND, as will be discussed a bit more below, we have good models for why that is so.

2] Mathematical models and simulations vs thought experiments

Mathematical and/or computer simulation exercises are interesting, but have this critical flaw: they are no better than their underlying assumptions and dynamical/ logical structure. So, GIGO — garbage in, garbage out.

That is why physicists tend to put a more serious weight on experiments and thought experiments where actually carried out experiments are not directly feasible (or where the thought exercise is sufficient to make the point).

So, in the thread above, I gave the cases:

(i) of a loaded die that 2/3 of the time comes up 6’s on a run of 300 tosses (this, to show that probabilistic trends can strongly come out in reasonable sized samples), and

(ii) dropping darts to scatter holes at random across a bell-chart divided into stripes on a floor (to show how after a reasonable number of samples, trends will be more and more evident and skirts will onward begin to show up as enough sample points make low probability outcomes more likely to become manifest).

The relevance of such to the case in view in the original post, should be plain to all who have ever had much to do with real world data collection and analysis for serious decision making or for scientific investigation across time. A LOT of real science has been based on data sets of scale comparable to — or a lot smaller than — the one above in the original post, and many crucial decisions have had to be made on that scope of data or less. And, the decisions or inferences were confidently and often correctly made, too.

The thought exercises bring out what is going on pretty well. Namely, selective hyperskepticism in the face of strong evidence that the Weasel program circa 1986 explicitly or implicitly latched [or if you want quasi-latched], as it ratcheted its way tot he target by rewarding mere proximity in the teeth of non-function.

That ratcheting off proximity to target in the teeth of non-function, is telling, especially in light of Mr Dawkins’ description of what he did, why above.

Weasel is utterly irrelevant to, strawmanises and begs the question of the need to generate functionally specific complex information to get TO the shores of islands of functionality. Instead of addressing the blatant fact that even toy-scope modest function [1 in 10^40 not a small searchable fraction of 10^180,000], it starts on the rhetorically convenient assumption that you are at shores of some minimal functionality and can find easy steps all the way to the peaks of optimal function.

But event that nicely stepped path to the peak — post Behe’s Edge of Evo — is a serious question mark!

What is clear is that Weasel is plainly NOT the work of the BLIND watchmaker of the title of Mr Dawkins’ 1986 book.

3] But Weasel (circa 1986) does not EXPLICITLY latch!!!!

On Mr Dawkins’ say-so as reported by Mr Elsberry and I believe Mr Kellogg above, we have accepted that; never mind that it is the most natural explanation otherwise of the observed 1986 behaviour of Weasel as reported and published by Mr Dawkins. [Which we can very safely assume was representative of the result of what he then thought of as "good" runs to target.]

But, we have also shown that:

1 –> At 5% mutation rate per letter per string per generation, typical strings will show 0 [~ 24% of time], 1 or maybe 2 mutations, with 3 or more being out in the low probability tail.

2 –> Weasel on “good” published runs circa 1986 runs to target in 40+ and 60+ generations which means that no-change is winning some 50% of the time.

3 –> that is consistent with generations of sufficiently large scope that 0-change strings show up reliably, but 2 and more are relatively rare. Otherwise, extreme tail end multiple mutation cases with correct letters would dominate the closest to target filter and runs to target would be fast indeed.

4 –> This is also consistent with the emergence of implicit [quasi-]latching. That is, the numbers of cases in a generation where a letter reverts and another letter advances so that the resulting flicked back population member becomes the advanced champion of the generation is very rare.

5 –> As a result, we see steady ratcheting of progress to target, with letters that make an advance preserved from one generation to the next, i.e latched or effectively latched.

6 –> And, most importantly, since there is no realistic functionality requisite, there is no material resemblance to a process that allegedly uses chance variation plus non foresighted natural selection to move towards function and by hill climbing thence optimal function.

++++++++++++++

The matter is plainly over on the merits, and we need not attend further to selective hyeprskepticiasm as it inadvertently publicly reduces itself to self-referential absurdity and incoherence; also in that exposing the underlying intellectual and ethical irresponsibility that lurk in evolutionary materialism. Indeed, that have lurked in materialism ever since Lucretius’ day, some 2,000 years ago.

GEM of TKI

____________

PS Re Mr Severski — if he had followed up just a little more at the Gleaner’s site he would have seen that the Gleaner — no friend of mine [that formerly great newspaper has long since lost its credibility (especially on its commentary pages), sadly; so much so that it is one of the motivators of my advice on spin tactics in the media here] — was led to publish a corrective article by me shortly thereafter. [And, yes I know the corrective appears above my name. I am forced to do that, given the abuse of my name by Mr Elsberry. Hopefully the spam surge will be short enough to be tolerable. Why spamming seems connected to my name appearing in fairly high traffic locations, I know not, save that there are web crawling bots out there that will search for information on names. I hope the other Caribbean person with that name will not suffer unduly because of this. In any case, it is a matter of basic courtesy to treat people online in light of the handles they use.]

(In fact the “corrective” was not a “response” but a PRE-sponse that anticipated what Mr Boyne said, but was [badly] edited by the Gleaner to come across as a response to Mr Boyne’s article, without notification to the reader. See what I am speaking about on loss of credibility, Mr Severski? Did you check out the quality of the SOURCE and the implications of the context before you — and probably others art Anti Evo — pounced on the convenient quip? Whichever leg of that dilemma, the point on intellectual irresponsibility is underscored. In short: thank you for notifying us of your ill-informed anti-Christian bigotry, so we know what to make of your further comments here or elsewhere.]

PPS: Onlookers, FYI: Mr Boyne, in the series of articles in question, was accusing evangelical Christians in Jamaica, utterly unjustly, of being potential theocratic tyrants who were willing to set out on butchery of those who dared disagree with them, much like has become a common blood slander among radical secularists in Europe and North America. For instance here is a gem from Boyne: “[Evangelicals in Jamaica etc are] prone to bigotry, intolerance and the desire to impose their will on others just as the Islamic militants.”

Does Mr Severski also wish to agree with Mr Boyne’s assessment on that claim? On what evidence, please?

In fact — and pointing to this now increasingly inconvenient fact is what provoked the attack by Mr Boyne in the first place — there has in recent decades been a studious ignoring and/or suppression of the contributions to the rise of modern liberty and democracy by those coming from the Biblical framework, not to mention the religious background of several crucial heroes of Jamaica, or even that of the writer of our National Anthem. [Onlookers, Cf here for a documentation of this point.]

In short, Severski (sadly, predictably) has – yet again — resorted to adverse personal commentary and selective citation or linking, without doing due diligence to first find then present a true, well-warranted and fair view of the truth. Typical of selective hyperskepticsm at work.

PPPS: I am further forced to link two blog remarks on the Boyne affair, here and here [the latter being the rebuttal that I actually submitted to the Gleaner]. I trust that fair minded people will see for themselves the sort of anti-Christian bigotry we are dealing with, and will draw prudent conclusions. Again, I ask that my HANDLE be used in referring to me, or even my initials.

FOOTNOTE: While I am at it, onlookers need to be pointed to this blog comment on the importance of public standards of decency, which also gives highly relevant context on why Evangelicals stand up for modesty in dress [one of Mr Boyne's points of issue in the series that provoked my response and led to his hit piece that Severski so unwisely decided to cite as though it were the unquestionable truth].

madsen:

Did you look at the ST5 antenna? There was no ratcheting toward a pre-specified target. Rather, just mutation and selection were involved.

There was a pre-specified target. And as I said I don’t think ratcheting was involved.

Furthermore, the two (quite different) best resulting designs were superior to human-designed antennas in certain respects.

You mean unaided human design. Those two antennas were still designed by humans.

The reason I am bringing this point up in the weasel thread is that I want to understand more about this notion of ratcheting toward a specified target you have brought up.

I was talking ONLY about ONE SPECIFIC example- the example that is the topic of this thread.

In particular, I would like to see if we can at least agree that evolutionary algorithms work even in the absence of such targets, the ST5 antenna being one example.

But there was a target- an antenna that could do X.

However that does not mean ratcheting was involved.

hazel:

That doesn’t mean that someone hasn’t, or can’t, write simulations of some aspects of genetic biology.

Like what, for example?

madsen:

I wonder if he accepts that EA’s can solve problems without the programmer explicitly telling the computer what the answer is beforehand.

But EAs are written to solve specific problems.

Do you think that someone can write an EA that is not specified to solve a problem and somehow it will evolve the capability to do so?

I would love to see that.

kairosfocus [165],

Your first-parody seems to be a delightful self-parody of your own excesses. Good for you! I had thought your sense of humor was utterly absent.

I’m going to provide translations of a couple other paragrahs here for onlookers who may not see them in terms of the long-term debate:

In the case of Weasel, it is now abundantly clear that it uses a targetted, proximity based search strategy that rewards non-functional configs on proximity without reference to functionality. So, it is yet another misleading icon of evolution.

Translation: I agree that Weasel works as Dawkins always said it did: as “a bit of a cheat.”

That is so whether or no it explicitly or implicitly latches or near-latches, and it is so in more modern Genetic Algorithm search simulations that are just as much characterised by active information fed in by designers

Translation: While I, kairosfocus, am not conceding error on the latching issue, I insist that, if I were wrong, it wouldn’t matter.

Why you spent so much time insisting on your case for an issue that you now say doesn’t matter in the least baffles. All you’ve done is gone back to the notion that it’s a targeted search — which has never been denied.

correction: for “first-parody” read “first paragraph.”

Joseph,

But EAs are written to solve specific problems.

Do you think that someone can write an EA that is not specified to solve a problem and somehow it will evolve the capability to do so?

I would love to see that.

I agree that GA’s are written to solve specific problems. I’m not claiming anything about their ability to evolve any capabilities.

It is now sadly evident that the much over-used evolutionary materialist advocate selective hypersketicism threadjacking tactic of red herrings led out to ad hominem soaked strawmen and onward to ignition that clouds and poisons the rtmopsphere for sertious discussion has reached the stage of turnabout accusations in this thread.

Say what-y now?

Joseph,

If by “target”, you mean the region of the search space where the fitness function is greater than some set value, then I don’t disagree.

The only thing I’m saying is that the two actual winning designs for the ST5 antenna didn’t have to be specified beforehand in order for the algorithm to work. They are novel designs, created by, erm, …, well, I’m not sure who created them. But the point is, you don’t know what you’re going to get until you run the algorithm.

madsen:

The only thing I’m saying is that the two actual winning designs for the ST5 antenna didn’t have to be specified beforehand in order for the algorithm to work.

And the only thing I am saying was the target was a pre-specified result.

They are novel designs, created by, erm, …, well, I’m not sure who created them.

The person/ people who wrote the code.

Computers and therefor computer codes are TOOLS, nothing more.

But the point is, you don’t know what you’re going to get until you run the algorithm.

I am pretty sure they knew they were going to get an antenna that matched the results they were looking for.

madsen,

Yes or No-

In “The Blind Watchmaker” the example of cumulative selection known as the “weasel” program, uses a ratcheting process.

David Kellogg,

It is a targeted search that uses a ratcheting process to reach that target.

And it would be fair to say the “ratcheting process” part has been denied.

kf,

In the absence of source code this whole argument seems to be about whether Dawkins was cheating more than previously admitted, but it’s difficult to come to any solid conclusion due to lacking all relevant information. Personally I don’t see the point since all know that the program is irrelevant to the main issues for reasons already discussed.

I’ve quickly scanned this conversation and I’m looking for some clarification in order to bring this thread to an end.

By “explicit latching” do you mean that the latching mechanism is a function in the program?

By “implicit latching” do you mean that the runtime parameters of the program were fine-tuned to produce such results but nothing was explicitly hard-coded except that non-functional intermediates are allowed? If so, is this the scenario that Darwinists are preferring on other sites?

Patrick,

At issue is the paper by Marks and Dembski which uses the “weasel” program as an example of a partitioned search which “locks” the letters into place once they match the target.

The anti-ID side was quick to jump on the paper because of that.

However upon further investigation the example they reference is an example of a targeted search which uses a ratcheting process.

And that is as described by Dawkins.

The issue has been ratcheting vs latching.

Ratcheting just means, as far as I can tell, that the program incrementally moves from the original random sequence to the “methinks” phrase, which it does. The word was first used by Joseph, and is not part of the original issue. No one denies that the process “ratchets” in this sense.

Latching, or locking, is a word used by Dembski in the original post, to refer specifically to the idea that once a letter occurs in the correct place, such as the first m, then it never changes in further generations. This is different than ratcheting, and is the issue under discussion.

Everyone agrees that that 1987 video clearly shows that the program does NOT latch (or lock – they mean the same thing.)

At 164, kairosfocus (GEM), quoting himself writes,

4 –> This is also consistent with the emergence of implicit [quasi-]latching. That is, the numbers of cases in a generation where a letter reverts and another letter advances so that the resulting flicked back population member becomes the advanced champion of the generation is very rare.

5 –> As a result, we see steady ratcheting of progress to target, with letters that make an advance preserved from one generation to the next, i.e latched or effectively latched.

It seems clear that GEM has accepted that the latching is NOT built into the program, but is rather quasi or implicit. Since the program does eventually reach the target, all letters eventually become correct and stay correct, even though along the way they may have been correct, then changed to incorrect, then became correct again.

So the program does not display real latching even though it ratchets its way to the target. Calling what happens quasi or implicit latching obscures rather than clarifies the distinction being argued, which is that the program does not latch (or lock) letters once they are correct.

That’s my understanding of this long, and long-winded, discussion

Hope that helps.

Joseph,

madsen,

Yes or No-

In “The Blind Watchmaker” the example of cumulative selection known as the “weasel” program, uses a ratcheting process.

Please tell me exactly what ratcheting means in the context of the weasel program, and I’ll be happy to answer.

madsen:

Please tell me exactly what ratcheting means in the context of the weasel program, and I’ll be happy to answer.

see comment 152

hazel,

Ratcheting involves “locking into place”.

Joseph [177],

Yes to the first sentence if you agree with what hazel [180] says:

Ratcheting just means, as far as I can tell, that the program incrementally moves from the original random sequence to the “methinks” phrase, which it does. The word was first used by Joseph, and is not part of the original issue. No one denies that the process “ratchets” in this sense.

No to your second sentence: the ratcheting process has never been denied by anybody associated with Weasel. Recall that in [140], once you had a copy of TBW, you confirmed that this was the way it worked. So this has never been denied.
David

Joseph,

The person/ people who wrote the code [created the novel designs]

Ok, let’s stipulate that for now. As long as you allow that design can arise through the application of GA’s.

I am pretty sure they knew they were going to get an antenna that matched the results they were looking for.

Again, I’m talking about whether or not the particular design of the winning antenna was specified beforehand. The answer is clearly no.

Joseph: Do you see the difference between the two idea I described above. If by ratcheting you mean latching in the sense that Dembski defines it in the opening post, then the program does NOT ratchet.

If by ratching you mean that EVENTUALLY all letters fall in place, even though a letter might have been in place and later out of place, then the programs ratchets without latching.

Ratcheting (your word) and latching do NOT mean the same thing.

Hi,

I’ve been lurking in this thread for a long time and I feel a little guilty for not contributing, so I thought I’d add some fuel to the fire.

Just to be clear, I’m pro-ID, although sometimes I do waver in that. I’ve been trying to account for my own bias while reading this.

For a while I was leaning towards agreeing with GLF that there is no letter latching going on. At this point though, I think there is implicit letter latching. I don’t think there is any way to prove or disprove it, but based on the output printed from the 1986 run, it seems unlikely that “correct” letters reverted to “incorrect” letters and then back to “correct” letters. I’m not claiming that I “know” that to be true or that my conclusion should sway anybody. That is just my conclusion based on reading the thread and looking at the charts. There could be different versions of the program, different configurations, or whatever is used to generate randomness could have led to the (apparent) implicit letter latching.

I don’t think Dawkins is trying to mislead people with this example. My impression is that he’s demonstrating a concept, not trying to say that it proves the concept. Obviously he hopes to persuade people, and its likely that some people, especially people who are predisposed to agreeing with him, will read more into the results than is warranted. Based on what I’ve read (in particular, Climbing Mount Improbable), I get the impression that he is willing to assume far too much. For example, when discussing the evolution of the eye, he concentrated on how “easy” it would be to duplicate photoreceptors, while ignoring the much bigger problem of the initial evolution of the photoreceptors. To his credit, he did acknowledge that he was ignoring that issue, but he spent significantly more time showing how the “easy” part could happen.

This will probably sound like a personal attack, but its not meant as one. GLF seems more intent on proving KF wrong than he is on proving his beliefs to be right. I know it could be my own biases that are making me interpret (or misinterpret) what is being said. I try to account for that, but this seems more like a personal grudge than an attempt to show the truth. Even if it could be proven that there was no letter latching, it would have a negligble impact on the overall arguments. When I see so much effort spent on a relatively unimportant point, (and I’ve had the same reaction to some pro-ID posters), it makes me question the judgement of the person making the arguments.

Thanks.

dl

Joseph,

In post #152 you define ratcheting as the process described by Dawkins in his book. No locking is described or implied there.

Yet in post #183, you claim that ratcheting does involve “locking into place”.

So I have to ask, which definition of “ratcheting” am I to use?

kairosfocus @ 164

In the case of Weasel, it is now abundantly clear that it uses a targetted, proximity based search strategy that rewards non-functional configs on proximity without reference to functionality. So, it is yet another misleading icon of evolution.

In other words, it does no more than Dawkins said it did: illustrates that cumulative selection converges on a target much more quickly than any random search.

If it was elevated to “an icon of evolution” it was only in the minds of critics who either misunderstood it or saw in it an opportunity to erect a strawman which could be demolished much more easily and dramatically than the real thing.

As for the reference to the Gleaner article, that was, like the WEASEL program, just an illustration. It demonstrated only that it was easy to find evidence of the use of your proper name in the public domain. I had no knowledge of the nature of the dispute being reported there nor did I comment on it, so your inference about my views on Christianity is presumptuous.

As to your request that we only use your “handle” or initials in posts to this board I am happy to comply. I will only note, as before, that it is absurd to pretend that your name is not now widely known in these circles. I would also point out that any contributor who wants to preserve anonymity will usually choose a “handle” that has no obvious connection to them, as I have. If someone appends their actual initials to most of their posts and chooses a “handle” that leads straight to their website, we are bound to wonder just how serious they really are about remaining anonymous.

madsen,

“Locking” taken in context, means there isn’t any backward movement.

Ratcheting means there isn’t any backward movement.

The way Dawkins uses and describes the “weasel” program in TBW it is a ratcheting process in every sense of the word.

The person/ people who wrote the code [created the novel designs]

Ok, let’s stipulate that for now. As long as you allow that design can arise through the application of GA’s.

Yes design can arise from a GA if the purpose of the GA is create a design.

Again, I’m talking about whether or not the particular design of the winning antenna was specified beforehand. The answer is clearly no.

It designed the thing it was programmed to design.

And even though no one knew what it would be beforehand, they knew what it needed to do.

IOW they didn’t write some general GA and then that GA evolved into an antenna- designing GA for just the application they were looking for.

Good – it clear that you use ratcheting to mean the same as latching. That clears that up.

Now, just to clarify, do you agree that the 1987 video shows non-latching?

hazel:

If by ratcheting you mean latching in the sense that Dembski defines it in the opening post, then the program does NOT ratchet.

Perhaps not in the same sense but with exactly the same result.

IOW there isn’t a coded statement that locks each letter as they appear and match.

The process itself takes care of that.

If by ratching you mean that EVENTUALLY all letters fall in place,

No going backward.

THAT is what I mean by ratcheting.

The way Dawkins uses and describes it in TBW the output will always be greater than or equal to the input.

See comment 152

hazel,

The argument ONLY pertains to the book “The Blind Watchmaker”.

That is because it was the BOOK that was referenced in the Marks/ Dembski paper.

The video is irrelevant.

Joseph,

“Locking” taken in context, means there isn’t any backward movement.

Ratcheting means there isn’t any backward movement.

The way Dawkins uses and describes the “weasel” program in TBW it is a ratcheting process in every sense of the word.

Looking back through the thread and reading your latest comment, I think we actually agree on what is happening with the program—there isn’t anything in it which prohibits backward movement, but the probability of that happening given the parameters involved is very small.

Apologies; I meant to post the above in the moderation policy thread. 195 refers not to madsen but to a comment in that thread.

madsen: Again, I’m talking about whether or not the particular design of the winning antenna was specified beforehand. The answer is clearly no.

Joseph: It designed the thing it was programmed to design.

And even though no one knew what it would be beforehand, they knew what it needed to do.

One can say, more specifically, that they predetermined the process in two ways.

1. By the way they coded the modification / “mutation” software they controlled and defined the set of possible candidates that could be considered. (This is not by specifying a list, but rather by specifying the rules of construction, composition, variation.) The algorithm could not consider any kind of candidate design that their own rules did not allow for. This is necessary, since every candidate design must be of a kind that their fitness software will both understand and know how to score.

2. By the way they coded the fitness function, they also predetermine the manner in which competing candidate designs will be scored and compared. They have predefined what they want for specifying which design is “better.”

I think we all agree that they don’t know in advance which design will win. They don’t even have a list of all the possible designs. (This could be prohibitively large. The search algorithm doesn’t even consider all possible candidates. It’s job is to try to efficiently seek out the optimal ones.)

On the other hand, it cannot be that the algorithm considers designs outside of what they have expressly allowed for. It could not decide, for example, to try changing something that the programmers did not anticipate changing, e.g. changing to a different material if the program wasn’t expressly designed to search among and evaluate different materials.

In short, genetic algorithms are simply tools to quickly search within chosen very large sets of possibilities as specified and bounded by programmers for cases that may maximize (or minimize) conditions set by the programmers.

ericB,

I think we all agree that they don’t know in advance which design will win. They don’t even have a list of all the possible designs. (This could be prohibitively large. The search algorithm doesn’t even consider all possible candidates. It’s job is to try to efficiently seek out the optimal ones.)

I agree with everything you have said, and the above paragraph is the point I was trying to make.

OK, I’ve gotten interested in the Weasel problem. Let me summarize a bit, and then show some actual math that will help clarify things, I think.

1. The key issue is whether the program uses LATCHING (or locking). I take “latching” to mean that once a correct letter appears in the correct slot, it can NOT mutate again, and will therefore stay correct for the duration of the program.

2. Everyone agrees that the 1987 video clearly shows that the program does NOT use latching: some letters which are correct in one generation are later not correct.

As Dembski says in the opening post, “There [in the film] you see that his WEASEL program does a proximity search without locking (letters in the target sequence appear, disappear, and then reappear).”

3. The question is whether he used the same program when he wrote the Blind Watchmaker (BWM). In the book he shows examples of every tenth generation of two runs, and in those examples there are no incidences of a correct letter mutating to an incorrect letter and then mutating back again. As Dembski says, “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.”

4. Dawkins does not give many details about how his algorithm works. Some (notably Joseph and kairosfocus (GEM)) claim, based on the examples, that the BWM algorithm DID use latching (and was thus different than the video version), while others argue that the BWM algorithm and the video algorithm are the same, and that it is only the limited data set in BWM that makes us think that latching was used.

5. Dembski’s response to this was:

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).

I disagree. I think the natural thing to do is assume that the two algorithms are the same, and that the apparent latching in the BWM examples in just an artifact of a limited data set.

More importantly, I’m going to show some math that will help support my claim. I will show why a correct letter mutating to incorrect is a fairly rare event, so that we would not expect to “capture” this event by taking snapshots of every ten generations for just two runs, as illustrated in the BWM.

Dembski is a mathematician, and his specialty is probability, so perhaps he will comment on what I present below.

6. Here is what Dawkins says in BWM

a. We have 28 slots to fill and each slot can take one of 27 letters (counting a space as a letter.)

b. We begin by choosing a random letter for each slot. There are 27^28 = power ways of doing this, so the odds of getting the right phrase by pure luck is about 1 x 10^-40

c. each generation the letters mutate, although Dawkins does not state what mutation rate he used.

d. Here’s the key sentence:

The computer examines the mutant nonsense phrases, the ‘progeny’ of the original phrase, and chooses the one which, however slightly, most resembles the target phrase. [p.47 in my book]

7. Note that Dawkins does not explain the rules the program uses to decide which phrase “most resembles the target phrase.”

I am going to make some assumptions.

a. The the number of correct letters determines which phrase is to be kept. If the parent phrase has 14 correct letters and the child has 15 correct letters, we pick the child.

b. However, what if the parent has one correct letter mutate to incorrect (remember, I am assuming no latching) and also an incorrect letter mutate to a correct one, so that both phrases have the same number of correct letters?

I am going to assume, because I think it is the most reasonable thing to do, that in this case we keep the parent phrase.

c. Also, obviously, if the child phrase is worse than the parent phrase because a correct letter mutated to incorrect and no incorrect letter mutated to correct, then obviously we keep the parent.

d. So if a correct letter mutates to incorrect, the only way the child phrase could be more fit would be if two or more incorrect letters mutated to correct.

OK now let’s do some math!

8. First we need a mutation rate. Dawkins gives no hints as to what he used, but both Wesley Elsberry and kairosfocus (GEM) have used 5% in their examples, so I’ll use that. (Note that the argument I make is not dependent on any particular rate, but it’s much easier to make my points when we have actual numbers to think about rather than writing everything as formulas with a variable rate r.)

So, let p = probability of a letter mutating in any one generation = 5%
and let q = probability of a letter not mutating in any one generation = 95%.

9. So in each generation how many letters can we expect to mutate in the whole 28 letter phrase. The Binomial Probability formula nCk•p^n•q^n-k applies to this situation:

Probability of no mutations = 95%^28 = 24%
Probability of one mutation = 28C1•5%^1•95%^27 = 35%
Probability of two mutations = 28C2•5%^2•95%^26 = 25%
Probability of three mutations = 11%
Probability of four mutations = 4%
Probability of five mutations = 1%
Probability of more than five mutations = negligible

So note that most of the time the phrase will mutate three or fewer letters in any one generation.

10. Now let’s look at the probabilities of what happens to each letter each generation. A letter can either be correct or not correct in the parent generation.

a. Chance of correct staying correct is 95%, and chance of correct changing to incorrect is 5%. (Remember, no latching)

b. The chance of incorrect changing to correct is different, though. The incorrect letter has to not only mutate, it has to mutate to the correct letter, which will only happen 1/26 of the time it mutates. (There are 27 letters, so there are 26 other letters it could mutate into.)

Therefore the probability of incorrect changing to correct is 5% • 1/26 = 0.2%, or about 1/500 of the time.

Therefore the probability of incorrect staying incorrect is 99.8%

11. Looking at these two probabilities (correct to incorrect and incorrect to correct) might lead one to think that the program would take a very long time to reach the target. However let’s now look at things generation by generation as opposed to letter by letter.

Suppose we have a parent phrase with n correct letters so far, and suppose one of those n correct letters mutates. There are three possibilities

a. No incorrect letter becomes correct, and so the child phrase is worse, not better, so the parent phrase is kept and we never see that the correct letter mutated.

b. One incorrect letter becomes correct. There are now still n correct letters, and I decided at the beginning that in the case of a tie, we keep the parent. So again we don’t see that the correct letter mutated because the parent lived, not the child.

c. So the only way that a correct mutation can survive is if two or more incorrect letters become correct at the same time, because only then would the child phrase have more correct letters than the parent.

For this to happen, we would have to have three mutations: one correct to incorrect and two from incorrect to correct. As we say in 9 above, three or more mutations only happen 15% of the time anyway, and since incorrect staying incorrect even after a mutation is much more common than incorrect becoming correct, the probability of a correct letter mutating to incorrect AND that child phrase being better than the parent are very small because it would take at least two or more incorrect to correct mutations to compensate and improve the fitness of the child

The conclusion here is, then, that correct letters can mutate to incorrect but they very seldom survive because they weaken the fitness of the phrase too much.

Even if Dawkins had printed out all of the successful generations instead of every tenth one we might not see an example of correct mutating into incorrect. Only if he showed not only the successful phrases that lived but also all the rejected phrases that died would we see all the correct to incorrect mutations.

So latching is not needed. The process itself ensures that most of the time a correct letter will stay correct. It’s not that correct stays correct because there is some latching rule in the algorithm, but rather because correct mutating to incorrect is too detrimental to the fitness of the phrase.

My reasoned conclusion, as I stated in the beginning, is that the algorithm did not use latching.

[A disclaimer. I know that the situation is much more complicated than what I have explained, because you have 28 slots subject to the conditions in 10 and 11 above in each generation, and you have different numbers of correct letters at different times, and so on. Analysis of that would be extremely complicated, being full of various probability trees. I’m not going there – this has been time-consuming enough as it is.)

Comments on this analysis?

Very nice!

Hazel, personally I enjoyed your analysis., and reasd every word of it. I am simply not convinced of the exercise made by Dawkins to begin with.

I think of the analysis made by another mathmatician, David Berlinski:

- – - – - – - -

Advent of the Head Monkey

IT IS Richard Dawkins’s grand intention in The Blind Watchmaker to demonstrate, as one reviewer enthusiastically remarked, “how natural selection allows biologists to dispense with such notions as purpose and design.” This he does by exhibiting a process in which the random exploration of certain possibilities, a blind stab here, another there, is followed by the filtering effects of natural selection, some of those stabs saved, others discarded. But could a process so conceived — a Darwinian process — discover a simple English sentence: a target, say, chosen from Shakespeare? The question is by no means academic. If natural selection cannot discern a simple English sentence, what chance is there that it might have discovered the mammalian eye or the system by which glucose is regulated by the liver? A thought experiment in The Blind Watchmaker now follows. Randomness in the experiment is conveyed by the metaphor of the monkeys, perennial favorites in the theory of probability. There they sit, simian hands curved over the keyboards of a thousand typewriters, their long agile fingers striking keys at random. It is an image of some poignancy, those otherwise intelligent apes banging away at a machine they cannot fathom; and what makes the poignancy pointed is the fact that the system of rewards by which the apes have been induced to strike the typewriter’s keys is from the first rigged against them.

The probability that a monkey will strike a given letter is one in 26. The typewriter has 26 keys: the monkey, one working finger. But a letter is not a word. Should Dawkins demand that the monkey get two English letters right, the odds against success rise with terrible inexorability from one in 26 to one in 676. The Shakespearean target chosen by Dawkins — “Methinks it is like a weasel”-is a six-word sentence containing 28 English letters (including the spaces). It occupies an isolated point in a space of 10,000 million, million, million, million, million, million possibilities. This is a very large number; combinatorial inflation is at work. And these are very long odds. And a six-word sentence consisting of 28 English letters is a very short, very simple English sentence.

Such are the fatal facts. The problem confronting the monkeys is, of course, a double one: they must, to be sure, find the right letters, but they cannot lose the right letters once they have found them. A random search in a space of this size is an exercise in irrelevance. This is something the monkeys appear to know. What more, then, is expected; what more required? Cumulative selection, Dawkins argues- the answer offered as well by Stephen Jay Gould, Manfred Eigen, and Daniel Dennett. The experiment now proceeds in stages. The monkeys type randomly. After a time, they are allowed to survey what they have typed in order to choose the result “which however slightly most resembles the target phrase.” It is a computer that in Dawkins’s experiment performs the crucial assessments, but I prefer to imagine its role assigned to a scrutinizing monkey-the Head Monkey of the experiment. The process under way is one in which stray successes are spotted and then saved. This process is iterated and iterated again. Variations close to the target are conserved because they are close to the target, the Head Monkey equably surveying the scene until, with the appearance of a miracle in progress, randomly derived sentences do begin to converge on the target sentence itself.

The contrast between schemes and scenarios is striking. Acting on their own, the monkeys are adrift in fathomless possibilities, any accidental success-a pair of English-like letters-lost at once, those successes seeming like faint untraceable lights flickering over a wine-dark sea. The advent of the Head Monkey changes things entirely. Successes are conserved and then conserved again. The light that formerly flickered uncertainly now stays lit, a beacon burning steadily, a point of illumination. By the light of that light, other lights are lit, until the isolated successes converge, bringing order out of nothingness.

The entire exercise is, however, an achievement in self-deception. A target phrase? Iterations that most resemble the target? A Head Monkey that measures the distance between failure and success? If things are sightless, how is the target represented, and how is the distance between randomly generated phrases and the targets assessed? And by whom? And the Head Monkey? What of him? The mechanism of deliberate design, purged by Darwinian theory on the level of the organism, has reappeared in the description of natural selection itself, a vivid example of what Freud meant by the return of the repressed.

This is a point that Dawkins accepts without quite acknowledging, rather like a man adroitly separating his doctor’s diagnosis from his own disease.(6)

[(6) The same pattern of intellectual displacement is especially vivid in Daniel Dennett’s description of natural selection as a force subordinate to what he calls “the principle of the accumulation of design.” Sifting through the debris of chance, natural selection, he writes, occupies itself by “thriftily conserving the design work . . . accomplished at each stage.” But there is no such principle. Dennett has simply assumed that a sequence of conserved advantages will converge to an improvement in design; the assumption expresses a non sequitur.

Nature presents life with no targets. Life shambles forward, surging here, shuffling there, the small advantages accumulating on their own until something novel appears on the broad evolutionary screen-an arch or an eye, an intricate pattern of behavior, the complexity characteristic of life. May we, then, see this process at work, by seeing it simulated?

“Unfortunately,” Dawkins writes, “I think it may be beyond my powers as a programmer to set up such a counterfeit world.”(7)

[(7) It is absurdly easy to set up a sentence-searching algorithm obeying purely Darwinian constraints. The result, however, is always the same — gibberish.

This is the authentic voice of contemporary Darwinian theory. What may be illustrated by the theory does not involve a Darwinian mechanism; what involves a Darwinian mechanism cannot be illustrated by the theory.

hazel, you have presented a good case. I want to suggest one possible improvement. Dawkins refers to “mutant nonsense phrases, the ‘progeny’ of the original phrase.” In other words, each chosen phrase mutates multiple times in each generation, and the choice is made not betweeen the “parent” and the single mutated offspring but among multiple mutations.

Upright BiPed, Berlinski may once have been a mathematician, but the quote you present is not a mathematical analysis.

Let’s stay on topic, please. I’m just discussing the latching issue in Weasel – we are not discussing any larger issues. This is an exercise in math, and math only.

…my apologies

Onlookers:

I must first conclude some unfinished, rather distasteful business. Pardon.

For, S. having dropped a clanger, I see what looks uncommonly like a discreet silence on the part of those who so enthusiastically endorsed a dismissive statement in the Jamaican press, but which turned out on closer examination to be dripping with blood slander and with endorsement of public lewdness.

if this is so, such a tip-toeing away in silence would itself speak volumes. Sadly.

Let us hope S et al will explain themselves here and that the denizens of Anti Evo will correct their ill-judged use of the same remarks int eh Jamaican media.

Anyway, it is time to speak to a fairly substantial pair of comments; one of which inadvertently reaches the same conclusion that J and I have long since done on Weasel, while not recognising that we have spoken to evident latching of o/p and two different possible causes: [T2] explicit latching in program modules, {T3] implicit latching and/or quasi-latching.

One hopes that this arrival at consensus on actual analysis will finally lay to rest the hot rhetoric both here and at Anti Evo.

I will turn to that in a moment.

GEM of TKI

Hazel and Upright (and onlookers):

First, Hazel, thanks for taking time to engage on substance, not rhetoric. (you have missed a key point in J’s and my case, that we embrace not only explicit but internal latching or quasi-latching) but the analysis allows us to move the issue forward I believe. I believe the misunderstanding of our position is obviously inadvertent, not at all a calculated strawman.

Upright, thanks for putting up a very telling wider analysis by Mr Berlinski.

On points:

1] Hazel, 200: Some (notably Joseph and kairosfocus (GEM)) claim, based on the examples, that the BWM algorithm DID use latching (and was thus different than the video version), while others argue that the BWM algorithm and the video algorithm are the same

Both Joseph and I are more nuanced than that. The observation is evident latching on the o/p circa 1986 contrasted with a run in 1987 that on video multiply and frequently flicks back (including a “winking” effect where reversion to correct seems to happen rather quickly).

The issue we have had — and J and I are not simply equivalent — is explicit vs implicit latching, with the possibility of a quasi latching also in play. [J has stated that implicit latching is a most likely explanation, as a byproduct of the targetted search algorithm used as described by Mr Dawkins. I have pointed out that the core issue on Weasel is that it is targetted search that rewards non-functional configs, and that this obvioates any claims to being an exemplar of a BLIND watchmaker at worjk, ever since December last. As a secondary point I have pointed out to the evident o/p latching (as have many others over these 23 years) and have suggested that the most likely explanation is latching in the program, easiest as explicit, though also possibly implicit.]

On Mr Dawkins’ testimony as reported by Mr Elsberry [and I note, in light of a recent comment by a certain Dr Simmons, that "Mr" is strictly correct even in reference to one holding a PhD, save where disrespect is contextually evident; I have meant no disrespect], I have accepted it ar reasonable to address implicit latching.

In that context, I have pointed out that once a per letter per member of a generation mutation rate and population size are set in a range that a significant fraction will be 0-change and thereafter decreasing fractions will have 1 or more changes, multiplied by a population size that makes the skirt unlikely to appear in the cases where doubly changed letters substitute so that one reverts while one advances, we will reasonably get the sort of pattern observed in the 1986 runs. In particular, I commented on the point that if double and triple etc mutations are common, i.e. there is a fairly large population, we will see quite fast runs to target — instead of runs that are consistent with “good” runs for latching; which has median run length at 98 generations.

It is credible that such co-tuning can be achieved with a modest number of test runs and intuitively appe3laing mutation rates. [The 5% rate I put up is from others, I believe tracing inter alia to Rob. It averages out at a bit more than 1 per member of the generation.]

Next, I have had to point out that there is such a thing as a reasonable sample size in empirical contexts, that will be credibly typically representative of the population as a whole. And 200 – 300 in a context that we may safely presume means that the samples were selected for being typical of “good” runs, is beyond reasonable doubt, well within that range.

2] Upright, 202 and 205: Advent of the Head Monkey

Thank you. This aptly captures the wider context of the issue, and sets up any serious discussion on the merits. I both agree with Mr Berlinski — and yes, I know he holds a PhD (in I believe mathematics) — and find him to have admirably presented the matter.

3] H, 200: I take “latching” to mean that once a correct letter appears in the correct slot, it can NOT mutate again, and will therefore stay correct for the duration of the program.

This only describes a simplistic version of EXPLICIT latching; more sophisticated explicit latching is possible and implicit latching and quasi-/ imperfect- latching are also possible. By way of example, in the previous thread, Apollos presented code which will explicitly latch AND will have letters latched revert to incorrect status. (This means straight off that only credible code is actually demonstrative on the status of Weasel circa 1986.)

I have spoken to both implicit [T3] and explicit [T2] latching in processing, and to evident latching of output. The T2 and T3 options are alternative explanations for the observed output. In light of the further statement of Mr Dawkins circa 2000 received at second hand, we have in the main discussed T3 in this thread.

In short, the issue is somewhat mis-framed in the opening points of comment no. 200.

4] The question is whether he used the same program when he wrote the Blind Watchmaker (BWM).

Actually, th sis the secondary question, the primary one being the rhetorical status of Weasel vs its true substantial import. I and others have objected that in context it serves to lend persuasive force tot he BLIND watchmaker thesis, while in fact it is an example of designed, targetted search that uses a toy example and simplifying assumptions that evade the force of the Hoylean challenge of getting TO shores of functionality before hill climbing based on differential functionality can be brought to bear.

On the second level point, the issue is not whether the algorithm structure circa 1986 is different from that of 1987, but — on the presumption of some version of a T3, implicit latching or quasi-latching algorithm from 1986 on — whether parameters and o/p behaviour are credibly the same. In fact, early in the thread, Mr Dembski stated that such a shift could have material impact.

I have held that there is good reason to infer that the printed excerpts circa 1986 are representative of what was thought to be “good” performance at that time, and that this behaviour shows strong latching on the o/p, up to runs to target consistent with that. By sharp contrast, circa 1987, just as strongly, the o/p does NOT latch, but shows frequent reversions, often with winking.

I have concluded that the program circa 1987 is materially different form that circa 1986; which on T3 can be managed by changing two parameters: generation size and per letter per member mutation rates. It is probably noteworthy that an o/p that winks and runs fast but takes fairly long to get to target, is visually impressive.

[ . . . ]

5] I will show why a correct letter mutating to incorrect is a fairly rare event, so that we would not expect to “capture” this event by taking snapshots of every ten generations for just two runs, as illustrated in the BWM.

In short, you agree with my co-tuning of mutation rate and generation size analysis, and that it will lead to evident o/p latching or quasi-latching under certain circumstances. under others, esp by making mutation rate per letter per member high enough,a nd making generation size large enough that the far-skirt members are more likely to be present, we will see that the rewarding of mere proximity without reference to function will lead to domination of the succeeding generations by multiple correct letter members, and thus very fast runs tot target.

In an intermediate range, we should see more and more of reversion of letters to incorrect status,a s a double mutation substitutes another correct letter. The 1987 run seems to come from that intermediate range.

6] Chance of correct staying correct is 95%, and chance of correct changing to incorrect is 5%. (Remember, no latching)

No EXPLICIT latching. Latching (and quasi-latching) as I have discussed for quite some time in this thread, and in the previous one — and as J has also pointed out — can be implicit, an effect of interacting factors.

7] the probability of incorrect changing to correct is 5% • 1/26 = 0.2%, or about 1/500 of the time. Therefore the probability of incorrect staying incorrect is 99.8%

Thus, implicit quasi-latching,and effective latching in the case of appropriately co-tuned population scope and mutation rates. (Which can probably be done by trial and error looking for “good” results.)

8] the probability of a correct letter mutating to incorrect AND that child phrase being better than the parent are very small because it would take at least two or more incorrect to correct mutations to compensate and improve the fitness of the child

Oops.

This unfortunately neglects the case of the child simply being equal in [Hamming] distance to target as the parent, the latter being the champion from the previous generation.

This raises the interesting issue of a tie between zero-change members and a double change member that substitutes a new correct letter for an old correct letter that has reverted. there may be a randomiser in the program [flip a coin and pick], or a line that selects the changed phrase by preference in such cases.

It is in the case of the double-mutation with such substitution that we are most likely to see reversion behaviour. (My point of concern on this is that in the 1987 run, we see a further winking effect — very rapid reversion to correct — which seems odd if the letter is now simply on the long odds of getting back correct by mere chance. But, I am willing to not revert to the easiest [T2, Apollos variant] explanation for such, on grounds of taking testimony as so absent specific and strong grounds for rejecting it.)

8] The conclusion here is, then, that correct letters can mutate to incorrect but they very seldom survive because they weaken the fitness of the phrase too much.

This is exactly why Joseph and I have spoken of two patterns of algorithms, those that explicitly latch, and those that implicitly latch.

In short, your explicit binomial theorem based mathematical, probability analysis agrees with our narrative analysis based on programming patterns and population behaviour patterns. Indeed, you have confirmed that the implicit latching alternative is a viable mechanism for the o/p behaviour circa 1986.

HOWEVER, YOU HAVE ALSO UNFORTUNATELY MANAGED AT THE OUTSET TO MISS THE POINT WE HAVE REPEATEDLY MADE HERE AND IN THE PREVIOUS THREAD, THAT THERE IS IN ADDITION TO EXPLICIT LATCHING, IMPLICIT LATCHING (AND QUASI-LATCHING).

______________

So, your analysis agrees with ours, but your conclusion as stated is unfortunately flawed because you missed the point that we have looked at BOTH explicit latching and implicit latching. Also, I again call attention to Apollos’ code example that shows that EXPLICIT latching can be set up to mimic implicit quasi latching — with reversions.

So, can we agree that the Weasel program circa 1986 was probably a version of T3, with implicit latching, but in 1987, parameter changes would have made the latching shift to quasi-latching, without a tearaway streak to the target?

GEM of TKI

I cannot believe this thread is still going. Dawkins described what is known in the evolutionary computation community as a (1,n)-ES, where the ES stands for evolution strategy. I informed one of Dembski’s friends of this in email dated July 31, 2008.

In a (1,n)-ES, 1 parent generates n offspring in each generation, and the parent of the next generation is the fittest of the n offspring in the present generation. In a (1+n)-ES, the parent of the next generation is the fittest of the n+1 individuals in the present generation. That is, in the “comma” strategy, the parent “dies” after generating offspring, and there is a loose analogy to annual plants. In the “plus” strategy, there is a loose analogy to perennials.

There is a large body of formal analysis of evolution strategies. I believe that most of it involves Markov chain analysis. Some very bright people have already answered many fundamental questions about the behavior of evolution strategies. There is no need to reinvent the wheel.

Why Dembski is going on about “locking,” I have no idea. Until recently, the Big Three of evolutionary computation were evolutionary programming, evolution strategies, and genetic algorithms (including genetic programming). I find it difficult to believe that a critic of EC such as Dembski does not recognize an evolution strategy when he sees one.

At any rate, Google and ye shall find.

By the way, a (1,n)-ES is not a search algorithm in the sense of Dembski and Marks. I see no way to say anything about the active information of a “comma” ES without augmenting the ES with an extrinsic entity that registers the fittest parent of all generations. Of course, then the active information is not measured on the ES, but on the augmented ES.

It seems to me that the analytic framework of Dembski and Marks does not apply to the Weasel program — unless the framework is actually a Procrustean bed.

SG:

Weasel’s downfall is that it implements acknowledged targetted search that rewards NON-FUNCTIONAL “nonsense” phrases on mere proximity to target. This is NOT a matter of increments in current fitness being rewarded, in any sense of “fitness” worth using.

For, Weasel begs the question of achieving significantly complex information based functionality, in its leap to increment proximity of latest generation to target by selection off nearness to target for explicitly non-functional phrases. Thus, we see highlighted the key issue raised by two forms of the design inference: complex funcuoinality based on information is hard to find by random search, and irreducible complexity by direct means or by adaptation of existing parts to form a new function, is also hard to do by the engine of variation, chance. (And, without first having function, on pain of question begging, we cannot see nature culling based on better or worse degrees of function leading to differential success and reproduction.)

Further to this, in the recent Marks-Dembski paper, the T2, explicit latch version of the Weasel family of algorithms and implementations is addressed.

Searching for target zones and/or target points is . . . search in configuration spaces. (That is, this thread is over an incidental issue, whether in Weasel circa 1986 the correct letters to date are explicitly or implicitly latched, i.e locked or nearly locked in place. A glance at the Dawkins published runs of 1986 shows that once a letter is correct, it seems to be latched in place as Weasel marches on towards its target.)

Markov processes and evolutionary fitness strategies may be interesting but they are on a tangent to the key issue here.

GEM of TKI

PS: Onlookers: the above by SG and my recent response to Dr S in the mod pol thread, underscore how Weasel often succeeds rhetorically by such misdirection while failing to address the issue of increment in complex, functional information to get TO shorelines of islands of function. (The specific issue raised by Hole at eh beginning of the issue, and to which Weasel was claimed to respond.)

to David Kellogg: thanks for pointing out that each parent has multiple children. I had never thought about this Weasel program until yesterday when I got interested in the math, so I didn’t know that. But I understand now some aspects that weren’t clear to me before.

Of course this just makes my case stronger.

And this should lay to rest, I think, the question of whether the BWM algorithm contained a specific rule about latching, or not.

My conclusion is that it did not, and that the BWM algorithm and the video algorithm are the same. This was the issue in Demski’s opening post:

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).

I disagreed with Dembski, and I think I’ve proved my point.

hazel:

On that the issue is that Weasel circa 1986 and circa 1987 have significantly diverse output characteristics. Shifting the parameters controlling mutation rate and generation size would account for that, and for the purposes of our discussion, would be materially different.

GEM of TKI

madsen:

Looking back through the thread and reading your latest comment, I think we actually agree on what is happening with the program—there isn’t anything in it which prohibits backward movement, but the probability of that happening given the parameters involved is very small.

Given a target, a small enough mutations rate AND a large enough sample size, there will NEVER be any regression. NEVER.

IOW ratcheting toward the target, ie locking the correct letters in place, is inevitable.

hazel:

And this should lay to rest, I think, the question of whether the BWM algorithm contained a specific rule about latching, or not.

As I have already stated the program does not need a coded statement for something it does on its own.

I take it you cannot understand that.

Not my problem…

Sal Gal:

Why Dembski is going on about “locking,” I have no idea.

It’s like this- Marks and Dembski wrote a paper.

In that paper they referenced “The Blind Watchmaker” pertaing to the “weasel” program Dawkins describes in it.

In their paper they used it as an example of a partitioned search- meaning once the letters matched the target they get “locked into place”.

The anti-IDists have made a big to-do about it.

However when looking at ALL the data is is obvious that the program used and described by Dawkins in TBW uses a ratcheting process.

And that ratcheting process locks the matched letters in place.

There is a big difference, then, between cumulative selection (in which each improvement, however slight, is used as a basis for future building), and single-step selection (in which each new ‘try’ is a fresh one).- Dawkins TBW page 49

That’s ratcheting.

Joseph,

Given a target, a small enough mutations rate AND a large enough sample size, there will NEVER be any regression. NEVER.

Please prove this mathematically.

Now that I’ve gotten involved, I’ve gone back and read the start of the thread. I see that ROb made the same point I am making in post 5. All I did was supply a little math that shows why no latching rule is necessary, responding mainly to Dembski’s original point that it was “natural” to assume that a latching rule was in place.

That’s all.

Joseph writes,

As I have already stated the program does not need a coded statement for something it does on its own.

I take it you cannot understand that.

Not my problem…

I was responding to Dembski’s opening post, in which he referred to a partitioned search, “i.e., it locks on characters in the target sequence and never lets go).” I was not responding to you. If you accept that there is no latching rule in the algorithm, then we agree.

hazel:

I was responding to Dembski’s opening post, in which he referred to a partitioned search, “i.e., it locks on characters in the target sequence and never lets go).”

And that is what happens as described by Dawkins in TBW.

That is what happens with ratcheting.

madsen,

If you are REALLY interested just do it-

That means if you try you could find the exact parameters in which regression never takes place.

Ya see if the mutation rate is small enough it would mean that only one character could change- mutate.

It would also mean that there would be exact copies of the parent.

So if there are exact copies and mutated copies then AT A MINIMUM, the exact copies would be chosen.

Pretty basic actually.

Joseph,

madsen,

If you are REALLY interested just do it-

That means if you try you could find the exact parameters in which regression never takes place.

Ya see if the mutation rate is small enough it would mean that only one character could change- mutate.

It would also mean that there would be exact copies of the parent.

So if there are exact copies and mutated copies then AT A MINIMUM, the exact copies would be chosen.

Pretty basic actually.

While I accept that by adjusting the mutation rate and perhaps some of the other parameters, you can arrange that the probability of regression will be arbitrarily small, I don’t think you will be able to prove it will “NEVER” happen.

If you can present a proof, though, I’d be interested in seeing it.

David Kellogg:

the ratcheting process has never been denied by anybody associated with Weasel.

Yes it has:

All seven
characters are ratcheted into place.- Dembski/ Marks

IOW “rachet”ing was stated in the paper that you guys are trying to derail.

Read it for yourself:

Conservation of Information in Search: Measuring the Cost of Success page 5

That said it looks like they were correct as ratcheting does take place as described by Dawkins.

madsen

While I accept that by adjusting the mutation rate and perhaps some of the other parameters, you can arrange that the probability of regression will be arbitrarily small, I don’t think you will be able to prove it will “NEVER” happen.

I know you can never prove that regression can occur given the parameters I stated.

As I said given a small enough mutation rate and a large enough sample size the “worse” that can happen is the output = input.

But you could prove me wrong by going to one of the sites tat has the program and running it over and over again, until you find a regression.

I wrote comment #178 because I noticed people seemed to be generally agreeing in principle but yet–not realizing this–were still arguing, apparently due to objecting to certain terminology and descriptors. Never mind personal grievances.

To finish this thread, I’ll quickly do away with comment #212:

Hazel, you know you’re taking what Dembski said out of context. Dembski just noted that based upon ONLY the information in the book that it’s reasonable to presume a hard-coded latching function. The BBC video makes it obvious that this is not so, so he raises the possibility there are multiple versions of the program and that the issue should be investigated. As in, Dembski is leaving with a question, not an assertion. But given that you and several others have already found an explicit latching function is unnecessary given fine-tuned conditions*, and that Dawkins specifically says that explicit latching was not used, the best explanation at this point is that the program in the book and in the video were likely the same.

*Although as kf notes it’s possible to get the same results with multiple other approaches.

I think that about finishes this thread.

Joseph,

I know you can never prove that regression can occur given the parameters I stated.

But it’s your claim that regression can never happen. It’s up to you to demonstrate this.

Joseph, the example about Weasel in the paper you cited shows clearly that Dembski’s assumed that there was an explicit rule that locks letters in place. Here’s what he wrote:

Partitioned search [12] is a “divide and conquer” procedure best introduced by example. Consider the L = 28 character phrase METHINKS*IT*IS*LIKE*A*WEASEL (19) Suppose the result of our ?rst query of L = 28 characters is SCITAMROFNI*YRANOITULOVE*SAM (20) Two of the letters, {E,S}, are in the correct position. They are shown in a bold font. In partitioned search, our search for these letters is ?nished. For the incorrect letters, we select 26 new letters and obtain OOT*DENGISEDESEHT*ERA*NETSIL (21) Five new letters are found bringing the cumulative tally of discovered characters to {T,S,E,*,E,S,L}. All seven characters are ratcheted into place. Nineteen new letters are chosen and the process is repeated until the entire target phrase is found.

Notice that when you get two correct letters, you only “mutate” the remaining 26, and once you get five more correct, you only mutate the remaining 19.

This is an explicit rule. It’s is NOT quasi-latching or implicit latching or latching because it is highly improbable that a correct letter mutated to incorrect will get passed on. It is a RULE that says once we have found a letter it is not even considered for further mutation.

Period. This is absolutely clear.

Thanks for the summary, Patrick.

Hazel:

On the evidence of the o/p circa 1986, the simplest explanation is explicit latching of letters once they go correct. Implicit latching and/or quasi-latching is also possible, and that was noted before this thread ever began. In the D-M paper, they looked at the former case, which is a legitimate case of Weasel, given the many versions floating out there.

The 1987 o/p does not latch, and indeed seems to have a puzzling wink and get back pattern. On preponderance of evidence, including Mr Dawkins reported claim that he did not explicitly latch, the program, implicit latching as has been outlined many times above, is a reasonable explanation.

In either case, latching or quasi latching are secondary to Weasel’s real downfall: it is targetted search that rewards proximity without functionality. So, it is not a reasonable part of any case pointing to a BLIND watchmaker; the presumable aim of the book, BW.

That point was made ever since December last, but it seems that raising a secondary debate over what is highly evident form the 1986 o/p is highly distractive from that core flaw.

Bottomline: It is high time Weasel and kin were retired, as serving only rhetorical not illuminative, purposes.

GEM of TKI
________________

PS: Seversky, Kellogg and co associated with Anti Evo, I think you need to address the citation with evident approval here and also at Anti Evo, of someone indulging in, frankly, anti-Christian blood slander and through such ad hominems, dismissal of legitimate concerns over public lewdness. (As I recall, some Dancehall entertainers have actually been censured by the Jamaican courts over their beyond belief on-stage behaviour.)

kf@229, I agree with your first two paragraphs, so I think we can consider the discussion finished. I’ve never been interested in any of the other issues you mention, though, so I have no comment there.

Patrick,

In essence, if ratcheting is implemented, isn’t that what is going on?

That is with respect to parent and offspring.

IOW once a “parent” gets a matching letter, its offspring will also have those matching letters.

It is just a product of the process.

IOW once the letters are matched the search for them is over- ratchet clicks and fade…

Kellog at 202

I’m not sure on what grounds one could insiuate that David’s mathmatical acumen is somehow deteriorated.

You are correct however that his analysis was not pure mathematics, but, I am quite certain that was his specific intent.

He makes the larger point clear, there is a target in Dawkin’s exercise. The simulation reaches a predefined goal. Life, or more appropriately evolution, does not work that way. There is no target.

Produce “methinks its a weasel” without that phrase existing anywhere in the system, then that will be impressive.

Unitl then, Berlinski is 100% correct.

Joseph [223], I meant ratcheting in the sense defined by hazel. I’ll just refer to hazel [227] for clarification.

kairosfocus [229], I neither know nor care what you are talking about in your PS.

UB [232], Berlinski has a Ph.D. in philosophy and has done some work in mathematics but hasn’t contributed to the mathematical literature as far as I know. That’s what I meant. I would bet his mental faculties are as fine, or not, as ever. My point was that his essay does not address the mathematics of Dawkins’s program (the immediate point of discussion) or even its claims; it just makes an interpretation as to its value.

kairosfocus:

On the evidence of the o/p circa 1986, the simplest explanation is explicit latching of letters once they go correct.

How is the hypothesis that Dawkins added latching simpler than the hypothesis that he didn’t add latching?

The 1987 o/p does not latch, and indeed seems to have a puzzling wink and get back pattern.

This is explained by noting that the display cycles through all of the candidate strings in the population, not just the winner.

Bottomline: It is high time Weasel and kin were retired, as serving only rhetorical not illuminative, purposes.

I disagree that it’s not illuminative. It shows how the results of the oracle (or environment) simply communicating a fitness level can fool some people into thinking that more information is being communicated. Dembski, kairosfocus, and others thought that the oracle must not just be communicating the number of correct letters, but also which letters are correct. If nothing else, it shows that intuition isn’t reliable when it comes to genetic behavior.

Note that the artificiality of the fitness measure — i.e. the fitness measure has nothing to do with any observable functionality — has never been in dispute.

And I agree that Weasel and kin should be retired. Tell that to Marks and Dembski.

Nice post, R0b. It’s clear that you understood all this from the beginning, but I learned some things thinking through them myself.

I would add that our intuition is also often wrong when it comes to probability. Probability is a complicated subject and leads to many correct but counter-intuitive results.

R0b, KF, hazel, gpuccio, joseph, Patrick, et al.,

I finished coding Weasel Ware 2.0 this weekend, with lots of new features and goodies. I still have some testing to do and such, but one of the features I added was the ability for users to create their own fitness functions, via interpreted javascript. So we can see exactly how much active information a Proximity Reward fitness matrix contributes to the search and users can replace this with a reward matrix generated by a fitness function of their choice.

I expect a lot of new discussion to open up once this is released.

Oh, and R0b, you now will have both latching and non-latching versions side-by-side, for whichever you want to assume that Dawkins used for The Blind Watchmaker.

Atom

kairosfocus [229], out of curiosity, I tried to determine what you were talking about with respect to “blood slander.” I did a little reading, using your blog and the AtBC discussion as a guide.

I think the term “blood slander” is awfully strained in your usage; I’d even call it hyperbolic. Of course, there might be hyperbole on the other side as well, but I don’t know enough to judge.

I did want to thank you, though, for (perhaps inadvertently) bringing my attention to the beautiful Redemption Song Monument; it appears to be a powerful and moving sculpture.

Rob wrote:

“It shows how the results of the oracle (or environment) simply communicating a fitness level can fool some people into thinking that more information is being communicated. Dembski, kairosfocus, and others thought that the oracle must not just be communicating the number of correct letters, but also which letters are correct.”

Actually latching behavior requires less information: smaller code, less memory allocated, and the code is much faster and more efficient. Coding latched behavior is the shortest route to reproduce the output, so the conclusion that explicit latching is used is most reasonable based purely on examination, absent source code.

The non-latching implementation wastes a lot of time and memory to avoid locking letters. A side effect of hurling large mutating populations at what would otherwise be a simple search-and-latch implementation is that due to an unsophisticated target comparison a reversion is seen from time to time.

“If nothing else, it shows that intuition isn’t reliable when it comes to genetic behavior.”

There has been little disagreement here that Weasel exhibits no such behavior.

Apollos:

Actually latching behavior requires less information: smaller code, less memory allocated, and the code is much faster and more efficient.

That’s all true. But I didn’t say anything about code size, memory footprint, or efficiency.

In Dawkins’ non-latching code, the only information that the oracle communicates is the fitness level of the candidate. This is significant because the same is true of the environment in evolutionary theory. That is, the environment determines how successfully a given organism will reproduce, but it doesn’t tell the organism which genes to exempt from mutation.

WRT Weasel, the number of correct letters falls in the range of 0 to N, where N is the number of letters in the sequence. That means that the oracle communicates log2(N) bits of info in the non-latching algorithm. In the latching algorithm, on the other hand, the oracle communicates N bits of info.

“If nothing else, it shows that intuition isn’t reliable when it comes to genetic behavior.”

There has been little disagreement here that Weasel exhibits no such behavior.

Weasel is a genetic algorithm, so I’m not sure what your comment means.

Atom:

Oh, and R0b, you now will have both latching and non-latching versions side-by-side, for whichever you want to assume that Dawkins used for The Blind Watchmaker.

Very cool, Atom.

You’ll probably also want to change the “The Math” page, that claims to analyze Dawkins’ algorithm, but in fact analyzes an algorithm that has a mutation rate of 100% for unlatched letters, 0% for latched letters, and a population of 1.

Joseph

IOW there doesn’t have to be coded statement that locks the matching letters.

The locking is a byproduct of the program.

Atom

Oh, and R0b, you now will have both latching and non-latching versions side-by-side, for whichever you want to assume that Dawkins used for The Blind Watchmaker.

Atom, how did you know what behaviour to implement in the original version – where did you get the idea of latching letters (or not) from in the first place? Why did you do it that way?

Joseph,
If there is no coded statement in the program that locks the matching letters, how is it possible you need two different programs (as Atom says) to represent latching and non-latching behaviour?
Kariosfocus

To get implicit latching without reversions, the population as well as mutation rate actually need to be mutually tuned, so that you get steady advances without letter substitutions [one flicks back while another moves ahead].

Same question as to Joseph.

Tune the population and setttings you say? What settings, what program?

Atom,
Would it be possible for you to add a “FSCI meter” in the new version you are going to make available?

It would be interesting to see the FSCI values for the various strings change over time.

I realise this is “feature creep” but I think it would be a great feature!

So we can see exactly how much active information a Proximity Reward fitness matrix contributes to the search and users can replace this with a reward matrix generated by a fitness function of their choice.

Can we measure the active information in FSCI terms?

R0b,

I’m not in charge of the “Math” pages, I just code the GUIs. I’m sure Dr. Marks will update the pages, however, to include descriptions of the new features.

GLF,

I got the idea of latching from Dr. Marks and Dr. Dembski, who got that idea from the apparent (and possibly real) locking in the book version. With a low enough mutation rate and large enough population size (1 mutation per generation, 100 offspring) you get apparent locking behavior, as Joseph pointed out, even though no locking mechanism is in place for Proximity Reward Search.

As for the FSCI meter and feature creep: Sounds like something of an interesting idea, but the strings you search for can be any string, meaningful or not. I’d have to think through what FSCI would mean in that context (if it even applied), before I could even begin to code something like that. (FSCI requires a definition of functional states…I guess I could make the target the single functional state?)

Anyway, I have real (paid) work to do in the meantime, so I have to put off any major improvements or new features until I get a chunk of free time.

But there’ll be plenty to play with and discuss in the new version.

Atom

R0b wrote:

In Dawkins’ non-latching code, the only information that the oracle communicates is the fitness level of the candidate.

…and this “fitness level” encodes quite a bit of information about how far away a string is from the target; it is exactly inversely proportional to it, if I am not mistaken.

Without the information about target location encoded in the fitness reward matrix, the same algorithm sputters and can’t find the target well, if at all, even with reproduction, mutation and selection. In version 2.0 you can see this for yourself by choosing (or creating) a fitness function that has limited information about the target location.

Atom

Upright BiPed:

He makes the larger point clear, there is a target in Dawkin’s exercise. The simulation reaches a predefined goal. Life, or more appropriately evolution, does not work that way. There is no target.

Don’t tell Marks and Dembski this. If evolution, according to MET, has no target, then Marks and Dembski can’t apply their active information measure to biology without begging the question.

Produce “methinks its a weasel” without that phrase existing anywhere in the system, then that will be impressive.

Unitl then, Berlinski is 100% correct.

There are, of course, many such algorithms. A few years back, there was a kerfuffle in the ID debate over an algorithm that found the Steiner tree for a given set of points. This algorithm, like all useful searches, successfully inverted its objective function, providing solutions that were unknown to the designer of the program.

If Berlinski’s point is that Weasel doesn’t do this, then he’s trivially correct. Weasel is useless except as a pedagogical tool.

Dawkins evidently did not know, almost 25 years ago, that he had implemented an evolution strategy (ES). This does not give us license to ignore today that the Weasel program is an instance of an ES.

The shortcoming in the Weasel program is not in the ES, but in the fitness function. No evolutionist believes that the (un-)fitness of an organism is its distance in some space from a target.

The ES itself knows nothing about targets. It knows only that the solution space is the set of all length-28 sequences of uppercase letters and blanks. The ES passes candidate solutions to the fitness function, in which the details of the problem are hidden from the ES proper. The Weasel fitness function may be written

w(s) = 28 – Hamming(s, T),

where s is the candidate solution (a sentence of 28 letters and blanks), T is the target sentence, and Hamming(s, T) is the Hamming distance — the number of mismatches — between s and T. The ES uses the fitness values w(s) of progeny s to select the parent of the next generation, but has no “idea” how w(s) is computed.

In short, you get Dawkins’ Weasel program by plugging a particular fitness function into a generic ES. I can see only propaganda purposes in attacking the Weasel program instead of the combination of generic ES and fitness function. Do Dembski and Marks really want to make scholarly contributions to evolutionary informatics, or do they seek instead to make a big show of setting up and knocking over an ancient straw man?

The Weasel fitness function is very similar to one of the more heavily studied functions in the theory of evolutionary computation. The only difference between the Weasel function and the ONEMAX function is that ONEMAX restricts the characters to 0’s and 1’s, and the target is all 1’s. It’s a fair guess that some, if not most, ONEMAX analyses generalize easily to non-binary alphabets. In other words, there is quite body of theory to draw upon in analysis of an ES operating with the Weasel fitness function.

I believe that Dembski and Marks have known for quite some time that the Weasel program is an ES. So what game is Dembski playing? If you are a legitimate scholar and you know that you are analyzing an ES, you go to the ES literature to find prior analyses. You certainly do not conceal the prior work by making up new terminology like “proximity search” and “locking.”

I call shenanigans.

R0b says,

Weasel is useless except as a pedagogical tool.

Thank you. I had written something very similar before seeing your comment: Dawkins made it very clear that he was using the Weasel program as a pedagogical tool.

Anyone who has worked with evolution strategies — this should include any evolutionary informaticist — asks first about an ES whether parents compete with progeny for survival into the next generation. Dawkins answers this clearly enough in The Blind Watchmaker:

The computer 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.

I have to ask if Dembski has ever implemented an evolution strategy for himself. I cannot imagine that he would have found Dawkins’ explanation ambiguous if he had.

Sal Gal wrote:

You certainly do not conceal the prior work by making up new terminology like “proximity search” and “locking.”

I made up the phrase “Proximity Reward” search, not Dembski. I use the phrase since it clearly explains to the layman what the fitness function in the Weasel example is doing. It rewards a string based on proximity to a target, hence “Proximity Reward Search”

If that’s “shenanigans”, then so be it. I think it is clear.

Atom

Sal Gal,

Is your representation of Evolution Strategies correct? It seems ES deals with real numbers. Hence, Weasel is not an ES in the terminology of some. If so, your assertion of shennanigans is premature at best and possibly incorrect altogether.

Are you prepared to argue Dawkins uses Gaussian Random Noise onto real numbers?

From Data Structures

Evolution Strategy or Evolutionsstrategie

A search technique first developed in Berlin. Each point in the search space is represented by a vector of real values. In the original Evolution Strategy, (1+1)-ES, the next point to search is given by adding gaussian random noise to the current search point. The new point is evaluated and if better the search continues from it. If not the search continues from the original point. The level of noise is automatically adjusted as the search proceeds.

Evolutionary Strategies can be thought of as like an analogue version of genetic algorithms. In (1+1)-ES, 1 parent is used to create 1 offspring. In (u+l)-ES and (u,l)-ES m parents are used to create l children (perhaps using crossover).

Also from Wiki:

Evolution strategy – Works with vectors of real numbers as representations of solutions, and typically uses self-adaptive mutation rates;

Damn, back at 228 I actually thought I agreed with kairosfocus about something, but now I see I was wrong.

He wrote,

On the evidence of the o/p circa 1986, the simplest explanation is explicit latching of letters once they go correct. Implicit latching and/or quasi-latching is also possible, and that was noted before this thread ever began. In the D-M paper, they looked at the former case, which is a legitimate case of Weasel, given the many versions floating out there.

No, no, no. The simplest explanation is NOT explicit latching. There is no reason to think