PZ Myers Does It Again

_{Clive Hayden

August 20, 2009

Darwinism, Education, Evolution, Intelligent Design, Science}

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

PZ Myers has, once again, railed against something that he doesn’t understand at his blog Pharyngula. Hi PZ! Notice that he doesn’t actually address the content of Dr. Dembski and Dr. Marks’ paper, which you can read here: Conservation of Information in Search: Measuring the Cost of Success, published at the IEEE. Given his argument, he doesn’t know how to measure the cost of success, yet claims that Dr. Dembski doesn’t understand selection. A bit of advice PZ, the argument presented by Dr. Dembski and Dr. Marks is very sophisticated PZ, your mud slinging isn’t PZ, you need to step it up PZ. I know this new stuff isn’t ez, but you may want to consider a response that has actual content PZ. Your argument against this peer-reviewed paper is still in its infancy, or, more accurately, still in the pharyngula stage, embryonic in its development.

Since evolution of the kind PZ subscribes to cannot be witnessed, the argument has moved into genetic algorithms with the advent of computational abilities to determine the affair, and the IEEE is an entirely appropriate place to publish on that subject. We’re not going anywhere, we’ll give him time to catch up and educate himself to the tenets of the paper’s actual content. And if/when he does, maybe he’ll write another blog, and possibly write one with active information, that is, actual information, or else his argument will never reach it’s target.

Comments

Any targeted search is evidence for ID as non-telic processes to not have a target in mind. IOW if Dawkins wrote a program to scrammble letters, without selection towards a target, and it hit on "Methinks it is like a weasel", he would have something that supports his position. Selection in that scenario would be whatever survived.Joseph_{August 22, 2009
August
08
Aug
22
22
2009
08:32 AM
8
08
32
AM
PDT}

1- Dawkins uses the weasel program to illustrate cumulative selection 2- Cumulative means to increase by succesive additions 3- Dawkins used cumulative selection to show that once something is found you don't have to keep searching for it- you have it. You don't keep searching for something you already have. 4- Dembski/ Marks used the words "partitioned search" and "ratcheting". 5- In a partioned search once you have something needed you don't need to search for it any more/ 6- Ratchet means to move in degrees in one direction only. So the bottom line is anyone familiar with the English language can see that Dembski and Marks were not wrong and their reference to TBW supports their claim.Joseph_{August 22, 2009
August
08
Aug
22
22
2009
08:27 AM
8
08
27
AM
PDT}

Lerned Hand:
Stripping out the subjective weasel words, this is an admission that a reversion is possible in any nonzero population with any nonzero mutation rate.
Only to those who are twisted and demented. As I said given a large enough population and a small enough mutation rate there will NEVER be a reversal. IOW using realistic numbers latching/ ratcheting will ALWAYS occur.Joseph_{August 22, 2009
August
08
Aug
22
22
2009
08:18 AM
8
08
18
AM
PDT}

kairosfocus#92
It has already been demonstrated above at 79 that Weasel c 1986 can legitimately be interpreted as a partitioned search
That is not the case. In fact, it has been conclusively demonstrated here: http://www.softwarematters.org/more-weasel.html that it is not possible to interpret the Weasel algorithm as defined in The Blind Watchmaker as a partitioned search based on a line by line reading of the actual text. The only people who seem to have this misconception are Royal Truman and Dembski and Marks. I suggest that you try the exercise of going through Dawkins' very clear prose to implement the Weasel algorithm yourself to see why it is obviously not a partitioned search.DeLurker_{August 22, 2009
August
08
Aug
22
22
2009
07:29 AM
7
07
29
AM
PDT}

Alex73#87
Indeed, you got a point, Dawkins does not say that. However, his code may still implicitly work as I described if it does the following: (0. Each generation is a well organized list of child strings.) 1. Once a generation has been created, it starts from the first one in the list. 2. It compares it to the Weasel string and gives it a score and makes it the candidate for the new parent. Then proceeds to the next string. 3. If the 2nd string has higher score than the first one, then the 2nd one becomes the new parent. In any other case, the first string will remain the next parent. 4. Continue to do #3 with all other strings. Now due to the low mutation rate, it is very likely that the first highly scoring string will be identical to the parent string or only differ in non-matching letters. In this case my description is still valid in the far majority of the cases.
The Weasel algorithm is clearly defined in The Blind Watchmaker. While your description may describe the majority of cases, it is definitely not valid. Your original statement to which I responded was:
1. The program chooses a winner only if it has more matching letters than the parent string. (i.e. in case of the best child string(s) having the same number of matches as the parent one the entire generation is discarded) It seems to be perfectly in line with Dawkins’ own words.
This is simply not the case. Further, I'm not sure that your suggestion
What Dawklins could have done is to gather all child strings with the highest score and choose one from them randomly.
would change the output when sampled every ten generations. It is just as likely that the first child with a given fitness would have a reversion as that any other child would. Finally, your last sentence
Then why is it such a holy cow?
The only issue I have is that Dembski and Marks continue to misrepresent Dawkins simple example, even after having been repeatedly corrected, with references to videos and Dawkins' own statement that no latching was used. The Weasel algorithm would be of little interest if Dembski and Marks didn't bring it up so often.DeLurker_{August 22, 2009
August
08
Aug
22
22
2009
07:22 AM
7
07
22
AM
PDT}

I don't see how any kind of demi-, quasi-, etc latchet ratchet tisket tasket can be seen as an argument for ID. Your saying that there is a strong emergent property of history based, population based methods.Nakashima_{August 22, 2009
August
08
Aug
22
22
2009
06:56 AM
6
06
56
AM
PDT}

Clive @ 77:
I made no concession.
It was lame of me to say you made a concession. I'm unimpressed when others crow when I drop a subject, so it was stupid of me to do the same. I apologize.
“Weasel” has nothing in common with life
Does cumulative selection (selection acting on the product of previous selection) occur in Weasel? Does it occur in life? If the answer is yes to both, then they have something in common. This is pretty simple.
And having no goal does mean that there is no goal, not even fitness as a goal, for you would have no criterion for fitness, for there is nothing to compare any accumulation to.
Dawkins says that life has no "distant ideal target" or "long-term goal" or "long-distance target" or "final perfection". Here is how he describes life: "In real life, the criterion for selection is always short-term, either simple survival or, more generally, reproductive success." "No long-term goal" does not mean "not even fitness as a goal." You seem to be of the opinion that fitness must be evaluated against a long-term goal. If so, you're wrong. In life, as well as in many virtual environments, fitness is not forward-looking at all. Perhaps organisms need only be better than their cousins at something in order to reproduce, and that "something" may even change over time.R0b_{August 22, 2009
August
08
Aug
22
22
2009
06:45 AM
6
06
45
AM
PDT}

PPPPSSSSS Please try and actually read my post @88 instead of coming up with excuses to avoid dealing with the issue on its merits.BillB_{August 22, 2009
August
08
Aug
22
22
2009
04:54 AM
4
04
54
AM
PDT}

18 –> In either case, these searches would for sure fail to arrive at and preserve say METHINKS*IT*IS*LIKE*A*BEAGLE, as the embedded target in teh program now pushes the string away from the target. That is, the Weasel type algorithm with the wrong target is sure to fail, even while just plain random chance has 1 in 10^40 or so odds of getting home on any one toss.
Think about what you are saying - if you want the search algorithm to find one target but you specify a different target, then it won't find the target that you haven't specified.
You know what you need to do to make amends.
Again, I'll turn my cheek to your insults, accusations of dishonesty, and abuse.BillB_{August 22, 2009
August
08
Aug
22
22
2009
04:52 AM
4
04
52
AM
PDT}

PPPS: Onlookers, recall as well, ratcheting advance to target, which effectively latches in successful letters, can be done BOTH explicitly and implicitly, especially in the case of selecting "good" runs to showcase. So, Drs Marks and Dembski are strictly correct to analyse on what is in front of them in BW, circa 1986: ratcheting behaviour without apparent reversion of ANY letter that once goes correct, across a sample of 200+.kairosfocus_{August 22, 2009
August
08
Aug
22
22
2009
04:42 AM
4
04
42
AM
PDT}

PS: Mr BillB, you have forfeited the right of civil discussion through your uncivil conduct, and that to someone who in your presence only a few days back in this blog had to deal with the threat or fact of malicious false report to the US Homeland security Dept as a terrorist threat, for warning on the dangers of incivility. You know what you need to do to make amends. PPS: Onlookers: It has already been demonstrated above at 79 that Weasel c 1986 can legitimately be interpreted as a partitioned search, given the meaning of the term cumulative [= "Increasing or enlarging by successive addition." ], the CRD stated context of rewarding the "smallest" increment of advance to the target [i.e one letter], and the published "good" runs -- 40+ and 60+ gens to target -- of 1986 that do not show reversions in samples amounting to 200+ letters that could have reverted; with the sharp contrast of the BBC Horizon 1987 videotaped, much longer and plainly unlatched run underscoring the difference in what was going on 1986 vs 1987. And, all of this is on a red herring led off to strawmen soaked in ad hominems and ignited, on a matter where the issue in the main -- one that shows the SCIENTIFIC progress of Design theory resented and resisted to the point of repeated, now more or less routine abuses and uncivil behaviour by the Darwinist establishment and their supporters -- is the significance of active information and the rise of a new ID metric associated with it.kairosfocus_{August 22, 2009
August
08
Aug
22
22
2009
04:32 AM
4
04
32
AM
PDT}

YD: Please note, the introduction to 79 shows why I first focus on the main point Drs Marks and Dembski were making, and then in part I I addressed the non-technical level of that main point. In Part 2 I addressed the red herring talking points that have been used since publication was announced, to try to shift attention from what was achieved, and to go down a sadly familiar rhetorical garden-path lit by the light of burning ad-hominem-soaked strawmen. In particular, the talking points on weasel (constituting a second level of red herrings and strawman distortions) are a red herring. I think part of the challenge many face is need for a bit of background on what design theory, information search and cost of search are about, and how these relate to bio-systems (and to Weasel). Pardon, therefore, some step by step notes on points: 1 --> ID is about the empirical study of signs of intelligence, among which complex, specified information [CSI] is a key sign, including of course the development of methods of measurements and calculation for quantities associated with such signs. (In the context of both known engineered systems, and biosystems, functionally specific, complex information [FSCI] is a particularly interesting sub-set of CSI.) 2 --> Information exists in the context of contingency: e.g. strings of capital letters and spaces are such that they can in any one position take up 27 values: A/ B/ . . . / Z/ *, the * standing in for the space. 3 --> As a result, if we have one position, say X, it can take up 27 possible states. For two member strings: XX, there can be 27 x 27 states, and for 28 length strings we have 27^28 possible values [~ 1.197 * 10^40], including of course METHINKS*IT*IS*LIKE*A*WEASEL 4 --> Particular configurations can come about in various ways: e.g. Shakespeare could compose -- intelligently -- a sentence; a random letter generator could get very lucky in one shot, or there may be ways to cumulatively -- by step by step progressive additions -- start from a random initial string and create it. 5 --> In any case, the point is that the Weasel string is not just any ordinary string: it functions as a specific, meaningful sentence in English. 6 --> It is also COMPLEX, not just because it has many interacting parts, but because these parts are such that they may take up manifold configurations, so there is a space of possible configurations that is large [over 10^40 possible configs]. 7 --> In that space, only a tiny fraction will fulfill the function of being the Weasel sentence, and this can be recognised in the particular case by looking at the string. [And, we may compose measures for how close to this we are.] 8 --> Now, we come to the problem of how do such things originate? By enormous good luck is in this case strictly possible but not sufficiently probable to be credible. 9 --> By intelligent action is a routinely observed cause. 10 --> Mr Dawkins claimed c 1986, that instead of "single step" good luck, if an initially at random string were allowed to multiply into a population generation by generation with some small chance of mutation of the letters, where the "closest" to Weasel were selected and the process would iterate, in a rather short time we will arrive at Weasel. He wrote two programs [one in BASIC, the other in PASCAL] and published some "runs." 11 --> However, on closer inspection, the programs already embed the Weasel sentence as a target, and the process of moving to the sentence is one of targetted search; and as he admitted, this is fundamentally dis-analogous to the claimed mechanism of evolution by chance variation and natural selection. indeed, it is evolution by artificial selection. 12 --> But, we are here dealing with a discussion of searches in configuration spaces, and since DNA is also a digital text string which is functional and complex, searches and search mechanisms and challenges facing them are relevant to the origination of FSCI and to the question of inference to design as the best explanation for cell based life. 13 --> in that context, Weasel is a case of a search, and teh ststement by CRD that it was by cumulative progress to taarget is a search method. The print-offs of 1986 and the remarks on cumulartive progress to target make it a reasonable implementation of QWeasel to do what is called partitioned letter-by letter search with explicit latching of the successful leters. (that otehr versions are alos possible is much of the focus for a side-discussion, already addressed.) 14 --> What Weasel demonstrates in fact -- though not in CRD's intent -- is the advantage conferred by active information, that makes the "cumulative selection" targetted search more likely to succeed than the average or reference search algorithm. 15 --> That is, since we already know the target, the simplest search would be to find the statement of the target in the WEASEL program and reproduce it. 16 --> Another solution is to split up the string into 28 sub-searches, with each letter now facing odds of 1 in 27 of being correct. Test for being on target letter by letter. Lock up the successful letters. Vary the others at random, and repeat the test. Do again until the whole target is achieved. 9noteteh role played by the known target in the search, and how saving successful letters form further change makes the search far more effective.) 17 --> Another solution is not so direct, but by getting into a band where population per generation, mutation rate per petter and selection filter are working together right, "good" runs will IMPLICITLY do in effect the same thing, as already explained. but again, the location of the target is a key aspect of the relatively advantageous performance of the search. 18 --> In either case, these searches would for sure fail to arrive at and preserve say METHINKS*IT*IS*LIKE*A*BEAGLE, as the embedded target in teh program now pushes the string away from the target. That is, the Weasel type algorithm with the wrong target is sure to fail, even while just plain random chance has 1 in 10^40 or so odds of getting home on any one toss. 19 --> In this case, the active information in the algorithm has a NEGATIVE effect, of sending the strings to bark up the wrong tree. (Pardon puns.) 20 --> What Dembski and Marks discuss in the paper is that active information in some cases will greatly assist searches, but in others will do just the opposite, so on average they will be no better than random search. [If you were to pick a pool of active searches out of a hat and run them against random walks, on average the one would do no better than the other.) 21 --> So, once we inject active information as a way to structure a search, we see that the search for a search that will run well on this track at this distance will become a problem. [By way of illustration, my countryman Mr Bolt was almost accidentally discovered at 15 on the cricket pitch, and put on a 200 m track. With only a little serious coaching, he ran world beating times. The rest is history, including what he is doing on the straight run race after a couple of years of trying. But he is not known as a 5,000 m champion. If someone with an educated eye were not looking at that cricket pitch, where would he be today?] 22 --> We can then build technical measures of the impact of a particular active information approach, by comparing its performance against random search as a yardstick of the average of all searches. 23 --> Where does such AcI come from? In all known cases: intelligence, e.g by knowing he target or knowing the way different configs perform , i.e the lay of the performance landscape so to speak. [We have long known how to take advantage of peakiness and the slopes of hills leading up to peaks through hill-climbing techniques.] 23 --> in short, active information is a sign of intelligence, and it is a measure of the impact of such intelligence on search relative to the yardstick random search. 24 --> As such it is inherently about design theory, and it is about the developments of metrics for design inferences, allowing quantitative measurement of key features of designed systems and the distinguishing of such systems from those that credibly cme about by chance and blind mechanical necessity without intelligent intervention to create directed contingency rather than stochastic, undirected contingency. GEM of TKIkairosfocus_{August 22, 2009
August
08
Aug
22
22
2009
04:13 AM
4
04
13
AM
PDT}

And there are some formatting errors, but you get the drift I hope.BillB_{August 22, 2009
August
08
Aug
22
22
2009
04:04 AM
4
04
04
AM
PDT}

Oops, it appears that whilst quoting from the paper the letters in the search phrases got pasted in reverse order. I wonder how that happened ;)BillB_{August 22, 2009
August
08
Aug
22
22
2009
04:03 AM
4
04
03
AM
PDT}

KF: The issue being discussed here is the fact that Dembski and Marks describe WEASEL as a partitioned search thus:
Partitioned search [Reference to Blind Watchmaker] is a “divide and conquer” procedure best introduced by example. Consider the L = 28 character phrase LESAEW * A * EKIL * SI * TI * SKNIHTEM Suppose that the result of our first query of L = 28 characters is MAS?EVOLUTIONARYI?NFROMATICS 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 finished. For the incorrect letters, we select 26 new letters and obtain LISTEN?ARE?THESE?DESIGNED?TOO
Dawkins does not describe his algorithm as such, and if you implement his algorithm as he describes it you produce a programme that does NOT perform a partitioned search and DOES produce multiple 'mutant progeny'. This is the issue being discussed, it is what PZM is complaining about and his complaint is what this whole post is about. As you yourself have admitted, WEASEL does not require an explicitly coded latching mechanism to produce the published results. Including a latching mechanism and not generating a population of candidate progeny converts WEASEL from a non-partitioned population based search into a partitioned iterative search. Lets go over this again: The description of the WEASEL algorithm provided by Dawkins is of a non-partitioned (non-latching) search that produces multiple candidates for each generation where every letter has a probability of mutating, and only the 'fittest' is selected for the next generation. Dembski and Marks describe an algorithm (as detailed above) where letters in a phrase are constantly randomised, and for each letter the randomisation is halted when that letter matches the target. These are two different algorithms. I'll repeat the central point in case you missed it: This debate concerns Dembski and Marks incorrect portrayal of another work in their own peer reviewed paper. They never justify describing WEASEL as a partitioned, non-population based search in their paper. If they believe that this is a valid interpretation then they should make this clear in their publication and give their reasons. This is the issue, everything else you keep bringing up is obfuscation and distraction. Now can you please try and stay on-topic.BillB_{August 22, 2009
August
08
Aug
22
22
2009
03:59 AM
3
03
59
AM
PDT}

DeLurker, Indeed, you got a point, Dawkins does not say that. However, his code may still implicitly work as I described if it does the following: (0. Each generation is a well organized list of child strings.) 1. Once a generation has been created, it starts from the first one in the list. 2. It compares it to the Weasel string and gives it a score and makes it the candidate for the new parent. Then proceeds to the next string. 3. If the 2nd string has higher score than the first one, then the 2nd one becomes the new parent. In any other case, the first string will remain the next parent. 4. Continue to do #3 with all other strings. Now due to the low mutation rate, it is very likely that the first highly scoring string will be identical to the parent string or only differ in non-matching letters. In this case my description is still valid in the far majority of the cases. What Dawklins could have done is to gather all child strings with the highest score and choose one from them randomly. This is, of course, not mentioned in the book. However, in this case, we should have seen plenty of changes among the incorrect letters and also some correct letters reversing. In his examples when he goes from 2 non-matching letters to 1 non-matching letter, there could have been many of other intermediate generations with different 2 non-matching letters until a lucky strike fixes one of them while not ruining any other one. Dawkins did not publish the code. We do not know why, but he certainly can avoid much criticism by doing that. The results he published look like explicit laching could be used. However, even in case of no explicit latching his code could likely behave just like one with low mutation rate and certain programmatic solutions. So those analyisng his 'search engine' as a partitioned search have a strong case. He also admits that it is not a good simulation of real life, so proving that evolution cannot work like this will not bring down the Darwinist establishment. Then why is it such a holy cow?Alex73_{August 22, 2009
August
08
Aug
22
22
2009
03:56 AM
3
03
56
AM
PDT}

kairosfocus, I don't see anything in part I of your post that discusses weasel specifically. Recall that in his post, Dr. Dembski referred to the weasel section specifically, saying that the analysis supported ID. I'm still at a loss as to the point of that section, if as you say, weasel is a very poor analogy for Darwinian evolution and hence is essentially irrelevant to the ID/Darwinism debate.yakky d_{August 22, 2009
August
08
Aug
22
22
2009
12:22 AM
12
12
22
AM
PDT}

FB: Remember, we are accounting for "good" runs that appear to latch and are claimed to not have used explicit latching. So, IMPLICIT latching and associated ratcheting through cumulative advance towards the target, become relevant concerns. As you will see in my linked discussion, in App 7 the always linked, I speak of QUASI-latching -- close enough to your term -- to speak of the case of relatively rare reversions (which I have usually seen with substitutions). When the parameters are detuned enough, we see far from latched behaviour similar to that BBC Horizon 1987 video, and often that happens with exploding numbers of generations; also similar to the BBC Horizon tape. This last, often as the last few letters keep getting substituted out so the deal is very hard to close. GEM of TKI PS: I guess I should note too that one aspect of sci methods is the replication of results. thanks to Atom's adjustable Weasel, we can replicate all sorts of patterns, which strongly legitimates the spectrum of weasels: explicitly latched, implicitly latched (esp for good runs), quasi-latched, unlatched. (Note too that Apollos showed us a case where an explicitly latched weasel can be programmed to do reversions etc.)kairosfocus_{August 22, 2009
August
08
Aug
22
22
2009
12:09 AM
12
12
09
AM
PDT}

YD: Kindly look at part I of my comment at 79 above, which I specifically gave as first priority. You will see there how the further analysis extends the frame of thought for inference to design. Also the onward, linked Atom comment and exchange with Rob make for interesting insights from one active with the EIL. Fresh meat. GEM of TKI PS: When a red herring-strawman-ad hominem distractor has been heavily used [and, rememeber I have been accused now of "gutter politics" and just over a week ago or so was threatened -- if it was just a threat (why I emailed and have written) -- with being reported to the US HS Dept as a potential terrorist threat to be put on a watch list for pointing out that this is an increasingly commopn and dangerous rhetorical pattern], it is necessary to address it correctively, and that justified what is part II.kairosfocus_{August 21, 2009
August
08
Aug
21
21
2009
11:58 PM
11
11
58
PM
PDT}

kairosfocus,
a –> Targetted search that uses mere proximity and which eliminates and even dismisses complex functionality as a threshold [why CRD derides "single-step selection"], is NOT a good analogy to what chance variation and natural selection are said to do; for want of basic similarity.
Let's stipulate that this is true: Weasel is not a good analogy to Darwinian evolutionary theory. How then does Dr. Dembski's critique of weasel become a critique of Darwinism, and somehow lend support to ID?yakky d_{August 21, 2009
August
08
Aug
21
21
2009
11:40 PM
11
11
40
PM
PDT}

kairosfocus in 79: "with big enough generations, and big enough mutation rates, multiple mutation cases may become common enough that we begin to see other effects in the runs of champions: substitutions where one letter advances to correct even as another reverts, and cases of double etc letter advances." This is why I think "demi-ratcheting" is the proper term to describe this Weasel business, as I mentioned in March in this comment: https://uncommondescent.com/intelligent-design/the-simulation-wars/#comment-312801 Since correct letters can sometimes change, "latched" doesn't quite work, while "ratcheting" implies change in only one direction (the good direction). By using the term "demi-ratcheting," we capture the rare occasions of correct letter reversions, while at the same time noting that the number of correct letters, the as it were "fitness" of the offspring, ever ratchets upwards. At any rate, in his comment #7 to his post about his new paper, Dr Dembski makes it clear that the issue of demi-ratcheting ("locked or non-locked") is beside the point. So there isn't much use arguing about it here.feebish_{August 21, 2009
August
08
Aug
21
21
2009
11:32 PM
11
11
32
PM
PDT}

NB: Atom has a good discussion here, in the graffiti thread, as usual. [A, how is the Luminous One?]kairosfocus_{August 21, 2009
August
08
Aug
21
21
2009
11:12 PM
11
11
12
PM
PDT}

OOPS: a –> Targetted search that uses mere proximity and which eliminates and even ~~discusses~~ dismisses complex functionality as a threshold [why CRD derides "single-step selection"], is NOT a good analogy to what chance variation and natural selection are said to do; for want of basic similarity.kairosfocus_{August 21, 2009
August
08
Aug
21
21
2009
11:08 PM
11
11
08
PM
PDT}

PPPS: BTW, just to show that we are not singing off the same hymn sheet, there is a point of minor disagreement between Joseph and I. Since the filter rewards the closest to target in any one generatio0n, with big enough generations, and big enough mutation rates, multiple mutation cases may become common enough that we begin to see other effects in the runs of champions: substitutions where one letter advances to correct even as another reverts, and cases of double etc letter advances. So, I think there is a band on pop size to see the latching and quasi-latching IMPLICIT effects. As I vaguely recall, once I started to push Atom's adjustable Weasel up to the 999 pop limit and also push mutation rates, some of these effects popped out of the woodwork. (Atom, that sounds like an old fashioned product name . . . sorry.)kairosfocus_{August 21, 2009
August
08
Aug
21
21
2009
11:01 PM
11
11
01
PM
PDT}

Onlookers: Let's go straight to the main point. The latest wave of Darwinist rhetorical wave attacks on "Weasel" are clearly intended to distract attention from, distort our understanding of and get us to dismiss the latest key ID peer reviewed paper, and its author. (As well as of course those who would support it.) tot hat end a by now all too familiar ruthless and destructive pattern of rhetoric has been deployed by the Darwinists: distractive red herrings, led out to strawman misrepresentations soaked in ad hominems and ignited to cloud, confuse, choke, poison and polarise the atmosphere. And, if you dare point this out, you will be attacked through turnabout false accusations, in my case now amounting to accusing me of "gutter politics." [That's why I have publicly stated to BillB that he has gone beyond the pale of civil discourse. Sadly, he is now an example of what not to do. Prayers, of course are always indicated; for all of us finite, fallible, fallen, struggling sinners.] The most urgent thing therefore is to first refocus our attention on the main point, then expose the sleazy tactics for what they are -- and I think you all will know the fairly recent historical exemplars of where that sort of incivility points and what it can do if it snowballs across our civilisation unchecked. (And TME, the written complaint should now be at the HS Dept.) I] First things first: What Dembski and Marks have done is: 1 --> they have extended the reach of the design thought focussed on CSI, by now addressing the issue of how we get to specified targets in configuration spaces. 2 --> The baseline approach is a random walk algorithm, which as we know from the many discussions of FSCI, can be swamped by exhaustion of search resources. 3 --> For instance if the entity uses 1,000 bits of info capacity to define its functionality (a functionality which is of course vulnerable to modest perturbation and so comes in islands), it will be such that the observed universe across its working life will only be able to run through less than 1 in 10^150 of the number of configs specified by that many bits. 4 --> Now, the cost of search issue is that there is no generic search algorithm that for all relevant problems will consistently outperform random search. that is while it may do better in some cases, it will actually do worse in others, and on average will be at most as good or as bad as random search. 5 --> this imposes the point hat there are horses for courses, and "good" search algors for particular challenges. 6 --> therefore, before we get to the direct search, we have a second order search for a good search to deal with. And so the problem explodes in a vicious regress. 7 --> but, when we introduce the possibility of independent knowledge of the situation, we see that we might be able to pick a good search. 8 --> The impact of such a good search can be measured by the advantage it confers over random search. 9 --> that is, we have now defined ACTIVE INFORMATION at least up to the level of a concept. 10 --> the empirically known source of such active information: intelligence. OOPS! II] The Distraction In a few sentences of the new paper, Drs Dembski and Marks make mention of a partitioned search approach to the Dawkins Weasel sentence target of 1986. This actually makes a lot of sense, as it is a well-known, self-confessed instance of targetted search using cumulative -- step by step addition -- search based on knowledge of target and proximity to it. (All of this is directly stated by Dr Dawkins himself, in so many words,a s I have long since laid out here, for those concerned to be truthful and fair minded.) That has some consequences: a --> Targetted search that uses mere proximity and which eliminates and even discusses complex functionality as a threshold [why CRD derides "single-step selection"], is NOT a good analogy to what chance variation and natural selection are said to do; for want of basic similarity. b --> In the context of CRD's enthusiastic description, he stresses the cumulative selection that rewards the slightest increment to target, and eagerly published a 40+ and 60+ generation run to target. c --> In neither case do we ever see a letter that has gone correct revert [with over 200 possible places for hat reversion to happen, taking up the bulk of the 300+ published letters from runs] -- by sharpest contrast with the 1987 BBC horizon videotaped run, which also took up many, many more generations as a consequence of frequent reversions. d --> So, when we multiply the two aspects of the available evidence, we see immediately that it is a legitimate form of a weasel algorithm to consider a case of letterwise-partitioned, explicitly latched search. e --> This, Dembski and Marks did and this they (and those who point out the legitimacy of so doing) are being castigated for. f --> Now, we know also that subsequently CRD is reported c 2000 to have claimed that Weasel c 1986 was not explicitly latched. This raises the question, whether such is compatible with the evidence published c 1986. g --> the answer is yes, as without EXPLICIT latching of successful letters and further step by step ratcheting forward with those in the bag specifically protected from further change, we can get IMPLICIT latching, at least for "good" runs. h --> This works by having a due balance of generation size, mutation rate per letter, and an appropriate selection filter (which includes the specifics of the Hamming distance to target metric). i --> What will happen in such a case is that art least one no-change case will be present in an overwhelming majority of generations, and that single step changes will dominate the rest. j --> That way, filters can be set up so that the closest to target will be either a same-state case, or a one-step increment in such "good" runs. [Observe that in 40 gens to target for 28 letters, nearly 1/2 the time, no-changer must win, and one step advances win most of the rest of cases.] So, both explicit and implicit latching are legitimate approaches to Weasel in looking at the published behaviour c 1986. And in neither case do the red herring and strawman talking points over latching -- not to mention the ad hominems that have been tossed at pro ID UD regulars with such abandon -- make a dime's difference tothe force of the Dembski-marks case in teh main. GEM of TKI PS: Last time around, there were a lot of talking points on how the apparent latching was not real. It seems that for now the law of large numbers implications for the 200-letter sample have put that set of talking points to rest. [Cf the already linked for a discussion of that red herring trail.] PPS: Let us note: many of these involved in the latest wave of attacks here and elsewhere are clearly connected to academia and/or education and presumably also the media and/or wider chattering classes. What we have seen over the past few days therefore, sadly, speaks volumes about judgement, truth and fairness issues that seem to be all too commonly characteristic of this generation of the intelligentsia in our civilisation; reflecting the corrosive impact of evolutionary materialism driven secular humanism -- which is inherently amoral [spell that: is-ought gap] and thus enabling of immorality, especially when pet issues are at stake. I think it is time to create viable alternatives to a corrupted intellectual culture and institutions.kairosfocus_{August 21, 2009
August
08
Aug
21
21
2009
10:49 PM
10
10
49
PM
PDT}

R0b, I made no concession. "Weasel" has nothing in common with life, remember, life's not like that, it doesn't search out a target, so there is no Weasel to be got. And having no goal does mean that there is no goal, not even fitness as a goal, for you would have no criterion for fitness, for there is nothing to compare any accumulation to.Clive Hayden_{August 21, 2009
August
08
Aug
21
21
2009
09:37 PM
9
09
37
PM
PDT}

Yes in an unrealistically small population with an unrealistically high mutation rate there could be a reversal. Stripping out the subjective weasel words, this is an admission that a reversion is possible in any nonzero population with any nonzero mutation rate. Did Dembski restrain his allusion to WEASEL to a narrow band of "reasonable" numbers? I don't recall any such limitation.Learned Hand_{August 21, 2009
August
08
Aug
21
21
2009
05:28 PM
5
05
28
PM
PDT}

Fitness is usually defined as those who leave more offspring. And there are many reasons why that could be. Also there may be many different "beneficial" variations competing within the same population.Joseph_{August 21, 2009
August
08
Aug
21
21
2009
05:05 PM
5
05
05
PM
PDT}

DeLurker, Yes in an unrealistically small population with an unrealistically high mutation rate there could be a reversal. That is why I said in a large enough population and a small enough mutation rate latching/ ratcheting is a given. And realistic mutation rates are less than 1%. Much less.Joseph_{August 21, 2009
August
08
Aug
21
21
2009
05:01 PM
5
05
01
PM
PDT}

Clive:
and if there is no target that is being approximated to in the accumulation, then the accumulation is anything, randomness, and always will be, for there would be nothing to compare the accumulation to, no standard of comparison.
No. Just because fitness is not defined in terms of a long-term target, that does not imply pure randomness. Since you are no longer arguing that WEASEL has nothing in common with life, I'll take that as a concession. Thank you.R0b_{August 21, 2009
August
08
Aug
21
21
2009
03:53 PM
3
03
53
PM
PDT}

Prev 1 … 4 5 6 7 8 9 Next

You must be logged in to post a comment.

Leave a Reply