The Simulation Wars

_{William Dembski

March 26, 2009

Intelligent Design}

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

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

Mendel’s Accountant: mendelsaccount.sourceforge.net

MutationWorks: www.mutationworks.com

MESA: www.iscid.org/mesa

Comments

Hazel: I have already pointed out that the very term latching implies that once a latched letter hits the target, its probability of further change falls to effectively zero. (A bit more than that in the case of quasi-latching, and also that of explicit latching that triggers post target shifting. These are why we need credible code to make a definitive conclusion beyond the preponderance of evidence.) I have also pointed out that (i) this is well warranted by Mr Dawkins' statements c. 1986 as already cited and remarked on, and that (ii) locking up on a letter by letter basis is not i8n principle different from locking up on the basis of hitting ther phrase. Given the partly hostile audience context here and elsewhere (which I need to always keep in mind), that extra bit is very important. GEM of TKIkairosfocus_{April 1, 2009
April
04
Apr
1
01
2009
02:36 AM
2
02
36
AM
PDT}

Mr Hayden (and others): My apologies. My intent was to advert to the fallacy of the closed mind; not to make a personal attack or to slight a person. I strictly intended to speak to the issue on the merits. GEM of TKIkairosfocus_{April 1, 2009
April
04
Apr
1
01
2009
02:28 AM
2
02
28
AM
PDT}

David Kellogg, and Kairosfocus, Kairosfocus: Don't call folks closed-minded. David: "Finally, you accuse me of close-minded repetition, yet your posts in this thread and the previous one have repeated the same nonsense over and over" Don't call folk's work nonsense. Both of you need to tone down the slighting and jabbing at each other.Clive Hayden_{March 31, 2009
March
03
Mar
31
31
2009
11:03 AM
11
11
03
AM
PDT}

kairosfocus [138], you are wrong in so many ways I scarcely know where to start. You say that I am providing
Capital illustration of trying to make something seem so by closed-minded repetition of selectively hyperskeptical claims
Not at all. I am merely saying what it true: that you can make no inference about trends when you have a highly biased sample from a population of unknown size. You know the sample is unrepresentative, but you persist. You don't know the population size, but you persist. I once heard a five-year old claim that "Most people are pink." He'd seen a sizable sample of people in his life -- certainly a great deal more than 200 -- but the sample he had observed was highly biased, and the world population was greater than he knew. It is as though you are drawing conclusions about the economy based on a survey of people who won the lottery! Finally, you have the sample size wrong. Each letter is not a member of the population: each phrase is. You can't sample below the level of the 28-letter phrase. Those are the "individuals" in the population. So your sample size is not 300, not 200, but 7 in one case and 8 in the other. Finally, you accuse me of close-minded repetition, yet your posts in this thread and the previous one have repeated the same nonsense over and over, for over 5000 words in this thread alone. DavidDavid Kellogg_{March 31, 2009
March
03
Mar
31
31
2009
05:17 AM
5
05
17
AM
PDT}

kairosfocus, you again responded to me on many issues that I didn’t bring up. I don’t understand why you do that. To summarize: You wrote, in 132,
Yes, in an implicit case, P(mut) is the same whether or no a let[t]er has already hit target. Yes, the p(mut) for non latched letters in explicit latchi[n]g is different from that of latched ones ….
I wrote, in 136, (shortened a bit)
Let p = the mutation probability and let p(mut) = the probability that a particular letter in a particular phrase will mutate. In the implicit case, for each letter p(mut) = p In the explicit case, for each letter if the letter is incorrect, p(mut) = p and if the letter is correct, p(mut) = 0
I think what I wrote is just agreeing with you, but is a little more specific about what p(mut) is in each case. Does what I wrote agree with what you wrote? Would you be so kind as to answer this question in a single paragraph or two? I’m not arguing any of the other points you’ve made, so I don’t think it’s necessary to bring them up again.hazel_{March 31, 2009
March
03
Mar
31
31
2009
04:53 AM
4
04
53
AM
PDT}

2] Hazel, 136: I don’t believe this is an issue that needs to be brought up anymore, at least not with me. Unfortunately, this is exactly the key issue that needs the most emphasis in the teeth of a barrage of attempts to dismiss or ignore or bury it. Weasel is targetted search that rewards proximity without regard to threshold of function. It is invalid as a claimed or implied example of a BLIND watchmaker. That is what shows what was going on in 1986, and it remains the problem with ever so much of Darwinian advocacy in the classroom and the popular media today. Worse, the want of basic realism of all too many computer simulations of "evolution" in the research literature shows that the problem of misleading computer icons of evolution is still with us. [Do I need to point out that once you can set up an algorithm, you can make a PC do almost anything and often make it look "real" enough? Only reality is reality! Simulations are not experiements! Computer projections/models are not FACTS -- contrary to what the Al Gores of this world may think. GIGO!!! models and theories must be validated against reality, and logic, and alternative assumptions, and must be seen as at best only provisionally supported, not true beyond any reasonable doubt. Ask the ghosts of Ptolemy, Galileo and Newton.] 3] “the probability of mutations of particular letters for members of the population.” On this, I object both that he individual probability of mutation is unrealistically large and the proportion of "good" possible mutations among these are also far too large. I also add, that the information capacity in question is also far too small. Such are highly relevant because we are in a search space context, and the cumulative effect it to give a highly misleading impression of relative ease of finding islands of functionality. 4] In the explicit case, for each letter if the letter is incorrect, p(mut) = p if the letter is correct, p(mut) = 0 Save for the case where as Apollos showed, we can program in mutations in explicitly latched cases [this is why only code, credible code, is demonstrative] that would be correct and has never been in dispute. After all, "explicit latching" plainly means that correct letters -- save as specifically programmed to do so -- will not revert. The temptation, though, is to then ignore the material context: letters arrive at that state because they have been mutated art random and rewarded on hitting target. That is, the target is partitioned letter-wise instead of seen as a whole. You will note that in by far and away most versions of Weasel out there, once the target is hit, no further mutations occur. An explicit latching case would "simply" do the same basic thing, but on a letter by letter basis. 5] Re GLF (et al): I ask you once more. How many letters total were in the run? It was not 300. It was not 200. How many letters were in the population as a whole? And of the 300+ letters shown, what section of the total population did they represent? Could there have been a sampling bias there? What is happening here is that, since these objections were answered with reference to the law of large numbers and sampling theory several threads ago, now in the archives of UD, they are being recirculated with selectively hyperskeptical assertions as though they were not cogently answered long since. Onlookers; GLF et al have NEVER been able to cogently address the issue that there is such a thing as the law of large numbers [which is foundational to statistical sampling, a generally accepted practice]. That is, the objection is selectively hyperskeptical as the samples in evidence do not point where they wish to go -- and, obviously, had it fit their agenda, they would have accepted the same results without a blink. Moreover, they have no good answer to the observation that there is reason to see that when Mr Dawkins showcased the o/p samples circa 1986 as "good results" we have every reason to take his statements that the o/ps showed "cumulative" progress to target at face value. Such circling back to already answered objections, sadly, is a hallmark of the fallacy of the closed mind. [I sampled this case only to illustrate the root problem, and to show why I no longer take GLF and ilk seriously; save as saddening examples of the rising tide of irresponsible intellectual and even uncivil conduct that threatens our liberties.] GEM of TKIkairosfocus_{March 31, 2009
March
03
Mar
31
31
2009
12:52 AM
12
12
52
AM
PDT}

Follow-up remarks: 1] Re DK, 135: the 1986 “observations” are a highly biased sample (by the very nature of the experiment) from a population of unknown size, you can conclude precisely nothing about latching from it. Capital illustration of trying to make something seem so by closed-minded repetition of selectively hyperskeptical claims, backed up by refusal to look seriously at how Mr Dawkins himself described the typical trends, i.e as cumulative, strong trends. 9not to mention his overall confe3ssions as I summarised above at 88 from easily accessible excerpts of the discussion, by those trying to justify what he did.] I have long since pointed out that once we have a reasonably sufficient sample of a population or a trend/timeline pattern and no reason to infer to bias, MOST samples are representative of the population and/or trends, especially those outcomes that dominate it statistically. This is called the law of large numbers, and is the foundation of sampling theory. So far as the population of possible outcomes of Weasel type exercises is concerned, 28 27-state elements have 27^28 ~ 10^40 states, which is a known size and in theory one that could be enumerated. But long before that, we will as a rule get a good enough look by sampling the population. Meaningful phrases in English, much less a unique one, will be vanishingly rare compared to nonsense, non-functional ones. That is why there is a challenge to get to shores of function. Mr Dawkins sought to overcome this by using a targetted search mechanism, but undercut its validity decisively by choosing to use a proximity rewarding search technique that makes no reference to functionality. Now, on picking up the trends as shown in the published o/p circa 1986, as a good rule of thumb, for linear trends not masked by excessive noise [including curvilinear], 5 - 9 samples are often enough to get a pretty good picture, though I am more comfortable with 20 - 30 if you have the time to get that, with further concentrations at knees. [But, too, data logging and automated point collection of thousands of data points are not always feasible, nor necessary.] For populations in general, 20 - 30 is a good enough point where the law of large numbers comes into force. 9tha tis a pint where the layman's law of averages has a point,though the layman's view ios often riddled with fallacies on reversions to the mean etc: the idea that if there has been a run of heads, a tail is overdue is not correct. But, if I were to see a coin you cannot directly inspect tossed 10 times and heads every time [odds on a fair coin being about 1 in 1,000], I would not be prepared to bet against it being a double-header. TRENDS, esp strong ones that lead to runs in data, often are very revealing.] In the case of the Weasel 1986, we have 300+ points where letters could change, and of these 200+ show letters going correct then staying that way, across dozens of generations, sampled by and large every tenth generation. There are NO cases of observed reversions: a strong run indeed. By sharpest contrast, the 1987 and later renditions of neo-Weasel, consistently show fairly frequent reversions. The striking difference is best explained by explicit or implicit latching in 1986, with implicit being the best on preponderance of evidence. 1987, on this model, was de-tuned so it shows reversions and winking. Neo-Weasel programs, extending the model, are designed not to fall into the now known trap of unpersuasively showing a dominant feature that spotlights what is wrong with the whole exercise: latching points to the invalidating significance of the proximity-reward search pattern that does not reckon with the search resources implications of the threshold of complex functionality manifested by real life forms. (This is also the reason why the same ilk objects so strongly to the otherwise obvious and easily exemplified concept of functionally specific complex information and its known cause.) As to the claim that the 1986 sample is "biased" we see that Mr Dawkins chose to showcase it. He and his supporters have to live with the consequences of such showcasing, and indeed, his further descriptions strongly suggest that this was typical of what he then considered "good" results. Then, the objections rolled in, and there was a shift to emphasising non-latching variations or at most quasi-latching variations. [ . . . ]kairosfocus_{March 31, 2009
March
03
Mar
31
31
2009
12:51 AM
12
12
51
AM
PDT}

kairosfocus, first you write,
The basic problem — as have noted since December — is, the program is pursuing a target, using a basic strategy that rewards mare proximity, with no reference to function. that means that it in effect suggests that it is rewarding differential function, but it is not. It is not BLIND watchmaker.
I agree with this, and think it is reasonably well stated. I think most people here agree with this. I don’t believe this is an issue that needs to be brought up anymore, at least not with me. Then you write,
Once that is in, questions on the probability of mutations of particular letters for members of the population are effectively moot.
For me, this is not a moot issue - this is the issue I’m interested in, because I’m interested in the programming logic. So, given that I accept your point in the first quote above, I’d like to continue the discussion about “the probability of mutations of particular letters for members of the population.” You write,
Yes, in an implicit case, P(mut) is the same whether or no a let[t]er has already hit target. Yes, the p(mut) for non latched letters in explicit latchi[n]g is different from that of latched ones ....
Good. I agree with this also. In fact, I think we could write the following: Let the mutation probability be p. For instance, many people have been using p = 5% in their examples. Also let us use, as you did, p(mut) as the probability that a particular letter in a particular phrase will mutate. In the implicit case, for each letter p(mut) = p In the explicit case, for each letter if the letter is incorrect, p(mut) = p if the letter is correct, p(mut) = 0 Do you agree that this adequately describes the difference between p(mut) in the implicit and explicit cases?hazel_{March 30, 2009
March
03
Mar
30
30
2009
05:37 PM
5
05
37
PM
PDT}

kairosfocus, since the 1986 "observations" are a highly biased sample (by the very nature of the experiment) from a population of unknown size, you can conclude precisely nothing about latching from it.David Kellogg_{March 30, 2009
March
03
Mar
30
30
2009
03:02 PM
3
03
02
PM
PDT}

George, "Then why is 99.999999999999999999999+% and more of the universe empty of life? That is the question you were asked. If the universe was designed for life, why is it basically empty of life? Your avoidence of that question has been noted." How much life would you deem necessary, percentage wise, in order to consider the universe designed for life? And what universe would you be using by comparison to say that this universe is not designed for life? Usually the argument that aliens do exist is used as an argument that our existence is nothing special, and so is the argument that there are no aliens. Folks need to chose which argument they'll advance against ID, and not use both. I see both arguments trotted out as "evidence" against ID, but they're logically contradictory. And secondly, if we had an idea of a healthy universe, then we could call ours sick. But we have no such knowledge, for the sample of the universe is 1.Clive Hayden_{March 30, 2009
March
03
Mar
30
30
2009
01:32 PM
1
01
32
PM
PDT}

George L Farquhar, "Is there some reason I was put into moderation? Clive?" So I can keep an eye on you :)Clive Hayden_{March 30, 2009
March
03
Mar
30
30
2009
01:23 PM
1
01
23
PM
PDT}

Hazel The basic problem -- as have noted since December -- is, the program is pursuing a target, using a basic strategy that rewards mare proximity, with no reference to function. that means that it in effect suggests that it is rewarding differential function, but it is not. It is not BLIND watchmaker. Once that is in, questions on the probability of mutations of particular letters for members of the population are effectively moot. Yes, in an implicit case, P(mut) is the same whether or no a leter has already hit target. Yes, the p(mut) for non latched letters in explicit latchig is different from that of latched ones [which will only mutate again if that is written in as Apollos demonstrated]. But, in context, we are seeking to explain what was observed in 1986, and what was said about it. What was observed is 200+ latched letters in a sample of 300+. No exceptions. What was said was, cumulative progress that rewards the slightest increment in proximity to target. Taken together, that more than legitimates understanding Weasel C 1986 to be explicitly latched, letterwise partitioned search, as Monash U showed -- until Mr Elsberry "corrected" them.. It is on the further statement that Mr Dawkins did not explicitly latch Weasel in 1986, that I have accepted that the best explanation per preponderance of evidence, a few oddities still sticking out, is implicit latching. And, we agree that there is a way to do it that with detuning also explains the 1987 videotaped o/p. Mr Dawkins' major problems are not with latching, but with setting up a foresighted watchmaker as if that answered to the challenge of a BLIND watchmaker. That is weasel's fundamental failing and it is why that program should be withdrawn, with an explanation of why it was misleading. GEM of TKIkairosfocus_{March 30, 2009
March
03
Mar
30
30
2009
12:59 PM
12
12
59
PM
PDT}

It seems that long comments get held up in moderation but short ones go right through. What a pity! Is there some reason I was put into moderation? Clive?George L Farquhar_{March 30, 2009
March
03
Mar
30
30
2009
12:48 PM
12
12
48
PM
PDT}

These are some comments I posted some time ago which did not appear at the time. Kariosfocus
Weasel has no legitimate illustrative or didactic role and only serves to distract from the Hoylean challenge
Then why do you suppose Dembski and Marks have a section of their evoinfo website dedicated to it?
Plainly, for 150 years now, Darwinism has had no real answer to this question. And, it is the key question.
If the question is "how did life arise" then no, there is "no real answer". Nobody knows. Do you know Kariosfocus? What insight does your position bring to the table?
Darwinism, once we move beyond minor micro-changes to already functioning organisms, is an empirical failure, a massive one; now increasingly sustained by the powers of an orthodoxy, complete with its own magisterium.
Is ID research being suppressed in the Caribbean too then?
In certain cases, the latching of the letters is practically all but certain. This is what on preponderance of evidence happened in 1986 in ch 3 of TBW, and in the NS run.
Is "practically certain" different to explicit latching then? That would be different to your original statement, right? Lets remind outselves of what you said originally
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.
And so we return to sampling bias, you said.
Indeed, there we can see that of 300+ positions that could change, 200+ show that letters, once correct stay that way, and none are seen to revert.
I ask you once more. How many letters total were in the run? It was not 300. It was not 200. How many letters were in the population as a whole? And of the 300+ letters shown, what section of the total population did they represent? Could there have been a sampling bias there?
Such a large sample provided by the man who in the same context exults in how progress is “cumulative,” is clearly representative.
How many letters are in the population as a whole? David Kellog has noted that Wesley Elsberry has made argued that the law of large numbers works against you. You have not responded to mathematical argument on its merits. and also I have pointed out that the data are highly non-representative because they are the products of a large selection bias (the best sampple from each generation provided). This was obvious from the text of TBW. Will you now respond to the mathematical argument on its merits? Or continue to claim a hollow victory? Will you address the bias issue or continue to simply ignore it and proclaim victory?
For the record, I have already long since pointed out that such o/p letter latching as samples published in 1986 show beyond reasonable doubt, can be implicitly achieved, not just explicitly so.
The maths says you have not. I find it interesting that you won't defend your position mathmatically yet claim it's correct because of the maths. Very interesting.
So, to put up a de-tuned case as if it were the sort of co-tuned latching case as we see in the 1986 runs, is a strawman fallacy. One that has been used in recent weeks over and over again. I note: the 1987 o/p is materially different from the 1986 one,a nd we have two viable mechanisms to explain the difference.
As has been explained several times, the difference is down to the fact that in the video the entire population was shown. And yet you persist with your misrepresentation.
All of this has been pointed out, and the meaning of “implicit latching” has long since been explained across three threads and hundreds of posts now in recent weeks.
And this is the critical point. You say that "implicit latching" proves your point - yet the solution is a mathmatical one and you refuse to address it. The probability of a letter changing once it is correct has been detailed several times, linked to several times. Yet you pretend that it simply does not exist? Why?
Indeed, the best explanation for our cosmos, in light of the factors and patterns we see, is that it is the product of design, powerful, elegant design too. And, it is COSMOLOGICAL design that points to an extra-cosmic, powerful and intelligent, artistically creative designer.
Then why is 99.999999999999999999999+% and more of the universe empty of life? That is the question you were asked. If the universe was designed for life, why is it basically empty of life? Your avoidence of that question has been noted.
but of course, one is free to shut one’s eyes to the obvious, and to dismiss the powerful testimony of the cosmos in which we live,
The obvious fact is that the only life in the universe that we are aware of is on this plaet. If the cosmos was designed for life, where is it all?
Repeat: Weasel ducks the real challenge — getting TO shores of function, not hill climbing to more or less optimal function.
What's your answer to the origin of life Kariosfocus?
Weasel begs the question of first having to get to shores of function before differential section based on degree of performance, and properly be taken into account.
What's your answer to the origin of life Kariosfocus?
And, by his own admission, that was to get away from the inconvenient fact that even in a toy example, a realistic functionality threshold would have been beyond the probabilistic search resources in the computer.
What's your answer to the origin of life Kariosfocus?George L Farquhar_{March 30, 2009
March
03
Mar
30
30
2009
12:46 PM
12
12
46
PM
PDT}

Have the problems with comments not showing been resolved? A number of mine have not yet appeared!George L Farquhar_{March 30, 2009
March
03
Mar
30
30
2009
10:40 AM
10
10
40
AM
PDT}

kf writes, "The imposed rule, that whichever member is closest to target will be champion — without reference to functionality — has the effect of latching, once the pop statistics are such that at least preservation of current position is very likely." Yes, I know that, and have agreed to that multiple times. Then, when I wrote, "Re: in this case the mutation routine never changes - every letter has a p% chance of mutating every time: it is the interaction with the rest of the system that produces the latching," you wrote,
The problem is, that being a member of the population in each generation is of secondary import.
But I would like to discuss this question that is of secondary import, given that we are in agreement about that which is of primary import. Just looking at the mutation part of the situation, and not at all the rest (which we are in agreement about), do you see anything wrong in saying this:?
In explicit latching, the mutation function knows about the target phrase because it only considers incorrect letters for mutation, but in implicit latching the mutation function itself does not know about the target phrase, and thus considers all letters, correct or incorrect, as subject to possible mutation.
Do you see anything wrong with the above?hazel_{March 30, 2009
March
03
Mar
30
30
2009
06:27 AM
6
06
27
AM
PDT}

Hazel: Re: in this case the mutation routine never changes - every letter has a p% chance of mutating every time: it is the interaction with the rest of the system that produces the latching. The problem is, that being a member of the population in each generation is of secondary import. The imposed rule, that whichever member is closest to target will be champion -- without reference to functionality -- has the effect of latching, once the pop statistics are such that at least preservation of current position is very likely. Also, on the sort of interpretation I deprecated earlier this AM, on seeing each letter as a function, the case can easily be viewed as optimisation in place so latching on hitting for the letters is "reasonable" and "justified" . . . you got to the target by random variation and cumulative selection. [And, recall, the overwhelming import of the text in 1986, including the published runs, shows latching, for which the simplest explanation is ratcheting based on explicit latching. I have gone with implicit latching as the better explanation of the 1986 results on the report that Mr Dawkins says he did not explicitly latch.] And, on the "tearaway run to target" on run no 1 case, Dawkins was clearly in the zone where multiple correct letters will crop up reasonably frequently and will be selected for by the imposed rule. (BTW, Jerry, it seems that this is a further data point on o/p latching -- i.e. he comes across as implying that that observed progress to target was cumulative and inexorable.) GEM of TKIkairosfocus_{March 30, 2009
March
03
Mar
30
30
2009
06:13 AM
6
06
13
AM
PDT}

Many people, as has been discussed here and elsewhere, have gotten results similar to those in the BWM with a mutation rate of 5% and populations around 100 or 200. I don't believe there are any other necessary parameters.hazel_{March 30, 2009
March
03
Mar
30
30
2009
05:31 AM
5
05
31
AM
PDT}

Yesterday, I found my copy of the Blind Watchmaker so I decided to read the chapter about the Weasel simulation. In Dawkins first example in the Blind Wathcmaker, after 20 generations, 20 of the 28 characters were found. Any one have any idea of what population size, mutation rate or other parameters were programmed in to get so quick a result?jerry_{March 30, 2009
March
03
Mar
30
30
2009
05:20 AM
5
05
20
AM
PDT}

Ah, progress. Re kf @ 123 First, when I wrote, “I will agree with kairosfocus about how implicit latching works,” he wrote, “Thus, after many cycles, finis on that point. Spirals do progress.” Actually, I’ve always agreed about the cause of implicit latching - I don’t think that has been a point of contention for me. kf writes,
Up to the point of letters being correct, there is little difference between (a) explicit and (b) implicit cases. In the former, once a letter is correct, it is removed from the search list, as it has attained its target . . . i.e. partitioned, letter by letter targetted search. In the latter, an interaction — note that point (in systems, interaction is a major source of emergent phenomena) — between per letter mutation rates, per generation sample of population [total of which is 10^ 40 or so] and the choosing of champions on mere proximity without reference to functionality — note Dawkins’ “mutant nonsense phrases” — cause lock-up of already successful letters. This, without an explicit per letter rule.
Good summary - I agree with all this also. Now, let me say this slightly differently. In explicit latching, the mutation function - that which goes along and decides whether to mutate each letter in the phrase - is influenced by or has knowledge of the target string because once a letter is correct the mutation function is instructed to leave it as it is. That is, the mutation function is no longer totally random - the chances of a letter mutating are p% (whatever mutation rate is being used) if the letter is incorrect but 0% if it is correct. In implicit latching, the latching is an emergent property of the interaction among such things as the mutation rate, the population, and the fitness routine which just counts how many correct letters a phrase has in respect to the target and then decides which phrase becomes the new parent. Note that in this case the mutation routine never changes - every letter has a p% chance of mutating every time: it is the interaction with the rest of the system that produces the latching. Therefore, the following distinction can be made: In explicit latching, mutation is not entirely random in respect to fitness: As the phrase gets more fit, fewer letters can mutate. In the implicit case, mutation is random in respect to fitness: no matter how fit the phrase is, every letter has the same chance of mutating every single time. Of course, because of the implicit latching, correct letters which mutate very, very seldom survive because of what you describe in the quote above. But the mutation routine itself does not know that - it just mutates without regard to or knowledge of what the rest of the process is going to do. Given that I absolutely understand what you are saying about the causes of explicit and implicit latching, do you agree with the distinction I am making in the above two paragraphs?hazel_{March 30, 2009
March
03
Mar
30
30
2009
04:52 AM
4
04
52
AM
PDT}

Onlookers (and current participants): Looking at onward exchanges, one would overlook that at what is currently 88 above, Dawkins' own words have laid out the matter clearly enough. Weasel is plainly highly "misleading" [teh word is Mr Dawkins'] on the central issue, fails to address the origin of functional bio-information, uses a filter that simply addresses proximity to target without function [thus actually is a form of intelligent design, as opposed to the BLIND watchmaker of the title of the 1986 book] and should be acknowledged as such and moved on from. However, there are some points that do need to be picked up and commented on. (i note to Jerry at 95 that I am writing for the record, showing he reductio ad absurdum plainly going on on the other side of this exchange.) On points: 1] Hazel, 90: I will agree with kairosfocus about how implicit latching works. Thus, after many cycles, finis on that point. Spirals do progress. 2] I am also not at all interested in the history or larger purpose (or non-purpose) of Dawkin’s program. Alas, from December on, that has been the key point: Weasel, as 88 shows, exists for a sadly rhetorical purpose; and works by in effect a subtle bait and switch that uses context to make its persuasive point, all the while, even more sadly, using -- here it comes -- weasel words to cover its tracks. (That concatenation of target phrase and rhetorical strategy chills me to the bones.) 3] in explicit latching, the mechanism is a rule about letters that is invoked in the mutation function, and in the implicit case, the mechanism involves members of the generation being selected, on probabilistic grounds, by the fitness function. First, recall, latching is in the first instance an observed o/p behaviour. The remaining question is mechanism, for which two candidates are credible. Up to the point of letters being correct, there is little difference between (a) explicit and (b) implicit cases. In the former, once a letter is correct, it is removed from the search list, as it has attained its target . . . i.e. partitioned, letter by letter targetted search. In the latter, an interaction -- note that point (in systems, interaction is a major source of emergent phenomena) -- between per letter mutation rates, per generation sample of population [total of which is 10^ 40 or so] and the choosing of champions on mere proximity without reference to functionality -- note Dawkins' "mutant nonsense phrases" -- cause lock-up of already successful letters. This, without an explicit per letter rule. In short, up to the insertion of a "fitness function," you are close to correct. (And of course as the coupling just described is relaxed, first we see occasional reversions then we see regular flicking backs, and also emergence of multiple-letter jumps in numbers of letters that are correct.] 4] DK, 92: please look a few of the the Weasel programs created by Zachriel, Patrick May, and others that claim not to use latching The answer to that is simple. First, simply look at the o/p of these neo-Weasels out there, by contrast with that linked for the original circa 1986. Latching is in the first instance recognised from the program's behaviour, a la 1986 o/p vs 1987. (Explicitness vs implicitness is a question of mechanisms to account for the observed o/p behaviour of the program as published in 1986 and discussed esp. in my Dec remarks.) I will (given the way the issues have been raised since 1986) guarantee that in by far and away most cases presented by evolutionary materialists and Darwinists more generally since 1987 on, such programs -- absent cases where they allow you to set parameters in the co-tuned range -- will most emphatically NOT latch. For, that was the obvious red flag that highlighted what was wrong with Weasel from 1986. (The Monash case is the obvious exception, and it was duly "corrected" by Mr Elsberry.) But, such neo-Weasels collectivley are simply yet another a bait-switch. The point I have always made is, as Mr Dawkins confessed in his weasel words on Weasel, circa 1986: Weasel is targetted search that rewards mere proximity, not functionality. As such it cannot be a reasonable illustration of the power of natural selection as a BLIND watchmaker. (For, NS is about differential PRESENT function.) And, the unanswered problem has always been to get to the islands of function per chance + necessity alone, not whether hill climbing once on such an island is more or less plausible. In short, the real question posed by the late, great, Sir Fred Hoyle and many others has been repeatedly begged and distracted from for 23 years. 5] Hazel, 96: I am trying to have an honest discussion about limited, specifics aspects of the topic. I am just trying to clear about one simple point - one that has nothing to do with the history of Weasel or it’s application, or lack thereof, to evolution. Why should kairosfocus, you, or anyone else feel that there is something wrong or threatening about this? Sadly, H, you are very much the exception. And, unfortunately, in answering you, I have to bear in mind the Anti Evo folks, there and here. So, like it or lump it [and I find it distasteful], I have to make it clear from immediate context that there is something fishy when they quote-mine. [To illustrate: Look at how they pounced on and trumpeted an ad hominem-laced dismissal in the J'can media that the Gleaner had to publish a corrective over; a dismissal that indulged in blood slander and served as enabling rhetoric for public lewdness. Then, when I blew the whistle on it, there was no serious accountability over that. And, that in a context of privacy violation. [BTW, once a real name is used these days, our friendly swindlers out there are perfectly capable of opening accounts etc in your name and doing nasty things with that identity theft.]) 6] SG, 118: I have argued that intelligence is a hypothetical construct. SG, your very first datum is that you are a conscious, choosing, thinking and acting, embodied agent, one who may sometimes make errors but is often right. Such as when you eat because you are hungry, choosing what you eat. Such as when you make sure no car is coming when you cross the road. If that is mere invisible, unobservable construct, then all else falls apart, including the claimed observations of the external world and the associated much vaunted 3rd person, onlooker perceptive. [Hint: observation is not to be equated to the 3rd person view, as we are capable of partly self-transcending reflection.] Is this not a plain case of self-referential incoherence? 7] if you feed programs in a Turing-complete language into the explanatory filter, Rice’s theorem makes things messy . . . . E.T. can say absolutely nothing about the purpose or function of your program. The linked reference of course continues to beg the same point, as though we do not personally and collectively instantiate semiotic agents who have track record of showing behaviours and creating artifacts that manifest aspects with distinct, empirically reliable signs of intelligence. Take for instance: There is no way to infer that a text is a program, let alone a unique machine for which the text is a program. Sounds impressive, but for the fact that we routinely do just that all the time. (And BTW, this reflects a significant part of the issue over FSCI. FIRST, observe function -- not just strings or what have you of symbols or glyphs in the abstract but symbols in action; then note that it is based on high-capacity storage of information, i.e is specified by function, is complex and is informational. Per experience such FSCI is a reliable sign of intelligence, e.g even your post, much less observed algorithmic function.) In short, sadly, self-referential incoherence yet again. 8] 121: I’m not saying that the string of characters [i.e. the methinks sentence] should function as a sentence. The environment pays off a particular character (trait) in each of 28 dimensions (positions). In essence, the simulated organism may function in as many as 28 different ways. This of course boils down to partitioned search, and to an absurdly question-beggingly low and unjustifiable threshold of function and probability [1 of 27], one that Mr Dawkins explicitly rejects in his direct statement "nonsense phrases." In short, SG here attempts to justify precisely a letterwise partitioned search as was the occasion of so much hearted dismissal overt the course of three threads to date in recent weeks. Sigh. _______________ Plainly, Weasel fails to illustrate teh BLIND watchmaker of the title of Mr Dawkins' 1986 book, and the attempts to justify Weasel and the various neo-Weasels, end up inadvertently underscoring the fundamental problem of evolutionary materialism. Namely, it has no credible account for the origin of functionally specific complex biological information. And, as 88 above shows, that was plain from 1986 [providing one read the text with a suitably critical eye], 23 years ago. GEM of TKIkairosfocus_{March 30, 2009
March
03
Mar
30
30
2009
12:10 AM
12
12
10
AM
PDT}

JT, 117 refers to 112. Having taught the theory of computation several times, and AI many times, I had to slip in a response. Good night.Sal Gal_{March 29, 2009
March
03
Mar
29
29
2009
09:11 PM
9
09
11
PM
PDT}

JT,
In the Weasel program, if the goal were “Me thinks it is a _______” and the blank could be filled in with any english noun, that would seem to make reaching a target easier.
It is easier.
I realize that in the WW, you say you mutate the target at random, but I’m thinking that what would be functional in nature wouldn’t be random.
Recall from 63 that I'm not saying that the string of characters should function as a sentence. The environment pays off a particular character (trait) in each of 28 dimensions (positions). In essence, the simulated organism may function in as many as 28 different ways.
But presumably there would be multiple N-bit string that were functional, that is there would be some programmatic description of what is a functional target.
You've guessed another modification I had waiting in the wings. With multiple environmental targets and a larger population, it is possible for subpopulations to track multiple targets simultaneously.
And it occurs to me that an individual target would not have to be an “exceedingly remote island of functionality” as for example an individual target need not be compressible.
There could be linked positions in targets. In other words, two targets could be constrained always to match one another in certain positions. To relate this to biology, consider that there may be multiple environmental niches for a species, and while they may differ in some ways, they do not necessarily drift independently.Sal Gal_{March 29, 2009
March
03
Mar
29
29
2009
09:05 PM
9
09
05
PM
PDT}

I’ve said before that to me, intelligence has to do with behavioral complexity, and also the degree of perceptual discrimination and acuity. So that if you have a program that does almost the same thing regardless of input (for example possibly because the program can only “see” the first n-bits of any input, then it is relatavely unintelligent. Or if you have a program that sees a much larger percentage of its input and furthermore the output it generates is highly variable, not trivially deducible from the input via any simple program for example, than that program is more intelligent. [also off-topic]JT_{March 29, 2009
March
03
Mar
29
29
2009
09:02 PM
9
09
02
PM
PDT}

SG [118]: What remarks of mine are you referring to? (I think I would agree that "intelligence is a hypothetical construct".)JT_{March 29, 2009
March
03
Mar
29
29
2009
08:52 PM
8
08
52
PM
PDT}

JT [off-topic], I have pointed out that if you feed programs in a Turing-complete language into the explanatory filter, Rice's theorem makes things messy. And I have argued that intelligence is a hypothetical construct.Sal Gal_{March 29, 2009
March
03
Mar
29
29
2009
08:46 PM
8
08
46
PM
PDT}

"multiple N-bit string" = "multiple N-bit strings"JT_{March 29, 2009
March
03
Mar
29
29
2009
08:41 PM
8
08
41
PM
PDT}

SG: I realize that in the WW, you say you mutate the target at random, but I'm thinking that what would be functional in nature wouldn't be random. But presumably there would be multiple N-bit string that were functional, that is there would be some programmatic description of what is a functional target. And it occurs to me that an individual target would not have to be an "exceedingly remote island of functionality" as for example an individual target need not be compressible.JT_{March 29, 2009
March
03
Mar
29
29
2009
08:40 PM
8
08
40
PM
PDT}

OK, I can think of an answer: Just make the goal "Me thinks it is a", that is, presumably you can make the target as short or long as you want.JT_{March 29, 2009
March
03
Mar
29
29
2009
08:27 PM
8
08
27
PM
PDT}

[113]:
JT:The simulation has only one target, a string of all 0’s (more about that in a minute).
SalGal:In both programs, it is irrelevant how the target is initialized. Any is as hard to locate as any other.
In the Weasel program, if the goal were "Me thinks it is a _______" and the blank could be filled in with any english noun, that would seem to make reaching a target easier. If MESA makes the weasel program more accurate by making it more difficult (by variable coupling for example), shouldn't it be more accurate wrt attributes that make it easier to reach a target as well.JT_{March 29, 2009
March
03
Mar
29
29
2009
08:20 PM
8
08
20
PM
PDT}

Prev 1 … 7 8 9 10 11 … 13 Next

You must be logged in to post a comment.

Leave a Reply