|
Author
|
Topic: Dembski on functional subsystems in "Uncommon Dissent"
|
Pim van Meurs
Member
Member # 541
|
posted 27. July 2003 15:02
Pim: I presume that you have no additional data to support the metaphor used by Dembski either?
Nelson:
If you can provide support for your desert claim, then that would be data against the "sea" metaphor.
My dessert claim refered to the absence of supporing evidence for the sea of non-functionality. In other words you are supporting my claims.
Nelson: However, you have yet to substantiate that claim.
Why? You have substantiated it for me.
Nelson: AFAIK, the export machine is the only "stand alone" function within flagella.
Interesting but what relevance do you think this has to you supporting Dembski's claim?
I also traced down Dembski's claim Richard Wein provides us with an excellent overview:
quote:
Dembski: Third, and perhaps most telling, Wein needs fitness to vary continuously with the topology of configuration space. Small changes in configuration space need to correlate with small changes in biological function, at least some of the time. If functions are extremely isolated in the sense that small departures from a functional island in configuration space lead to complete nonfunctionality, then there is no way to evolve into or out of those islands of functionality by Darwinian means.
The notion of "functional islands" is misleading, as I will show below. But the essential point that Dembski seems to be making here is that there might not be a viable evolutionary pathway to the bacterial flagellum. This is just another appeal to ignorance. In his previous section, he made this appeal more explicit: "But what guarantee is there that a sequence of baby-steps connects any two points in configuration space?" Science is not in the business of giving guarantees, but of making inferences to the best explanation. (Note: "configuration space" is equivalent to the term "phase space" which Dembski used in No Free Lunch.)
Btw it seems that the research that was supposed to support Dembski's islands of nonfunctionality, 'to be published anytime in the next two years' may not have lived up to its promise. Anyone knows what happened to this research?
Nelson in the past seems to have suggested that the NFL 'random assembly' calculations for the flagellum were evidence of the sea of nonfunctionality. But if that were the case why the need to refer to yet to be published research?
Is Dembski refering to the Discovery Org sponsored research by Douglas Axe (THE WEDGE PROJECTS: Phase I. Scientific Research, Writing & Publication, Individual Research Fellowship Program · Molecular Biology Research Program (Dr. Douglas Axe et al.))?
I have seen Axe's work used occasionally to suggest that protein evolution may face some problems.
For instance
Is intelligent design testable
quote:
n fact it is -- notably in the bacterial flagellum. Internet mavens have been pestering me for actual calculations of complexity involved in such systems. I address this in my forthcoming book (_No Free Lunch_), but such calculations are out there in the literature (cf. the work of Hubert Yockey, Robert Sauer, Peter Rüst, Paul Erbrich, Siegfried Scherer, and most recently Douglas Axe -- I'm not enlisting these individuals as design advocates but merely pointing out that methods for determining specified complexity are already part of biology).
An investigation of above claimsby Yehouda Harpaz
[ 27. July 2003, 19:26: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 28. July 2003 22:15
Nic,
That is simply not true. A pen is similar to a pencil in that it is a writing instrument. A tree is similar to a pencil in that they both contain sections that are made of wood. A tree is not similar to a pen at all.
A particular section of the sequence may be similar in both Mots and Tols, but ExbD may not have anything in common with MotB.
Reference a paper which shows homology between MotB and ExbD, show me the % sequence similarity.
[ 28. July 2003, 22:19: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 28. July 2003 22:17
Pim writes:
quote:
Nelson: However, you have yet to substantiate that claim.
Why? You have substantiated it for me.
how did I substantiate it for you? I showed that there is no stand alone function besides the T3SS. That is not a substantiation.
quote:
Nelson: AFAIK, the export machine is the only "stand alone" function within flagella.
Interesting but what relevance do you think this has to you supporting Dembski's claim?
The flagellum only has 1 alternative function. Where are all the other functions that natural selection can reward that would lead up to flagella?
[ 28. July 2003, 22:20: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 28. July 2003 22:23
RBH,
I'll respond more fully tomorrow as I actually did download Avida and am setting up in my own environment, for the most part, I'm mostly interested in checking your claims, since it has been already established that the experiment has absolutely no bearing on IC systems ID theorists are talking about.
RBH wrote:
quote:
As for emailing data, that's impossible: the files are gone. As I said, those were exploratory runs looking at sampling frequencies, and once I settled the sampling frequency question to my satisfaction the data files were over-written in subsequent exploratory runs.
I'm completely disappointed. Particularly because you made such a strong claim concerning the the 3 intermediate run (the number of intermediates got lower and lower each time you posted on the subject, which raised a red flag for me), which required 2 functions that were not rewarded and yet persisted. That raised a red flag for me, in that I suspect EQU was frontloaded into the starting conditions, and if true this can deal a serious blow to the experiment.
RBH writes:
quote:
Neither Dembski nor Nelson Alonso seem to understand what cooption means nor what kinds of things can be coopted by evolution.
There's not a shred of evidence to support this claim.
[ 28. July 2003, 22:25: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 29. July 2003 00:10
Nelson Alonso wrote quote:
... it has been already established that the experiment has absolutely no bearing on IC systems ID theorists are talking about.
"Established"? To your satisfaction, perhaps, but not to mine or others' satisfaction. "Absolutely no bearing" is a substantial exaggeration. It has been asserted, but not by any means "established."
And Nelson wrote quote:
RBH writes: quote:
Neither Dembski nor Nelson Alonso seem to understand what cooption means nor what kinds of things can be coopted by evolution.
There's not a shred of evidence to support this claim.
The clearest evidence is this remark that Nelson made earlier in the thread: quote:
I find it completely outlandish that NOT and OR_N and AND_N is sufficient for EQU, in which case, the program is even more plastic then even I had thought.
That remark makes sense only if Nelson believes that evolutionary processes must coopt functions as whole units, in this case NOT, OR_N, and AND_N, pasting them together as discrete units, and cannot make use of parts of functions. Since it can make use of parts in both the Avida simulation and in biology (see yersinia's remarks above), I infer that Nelson's conception of cooption is flawed. Since Nelson attributes the same position to Dembski, the same must be the case for him.
Finally, that I reported the results of several series of runs over several postings raised a red flag for Nelson. In fact, I made just three postings in which I mentioned the number of intermediates in the pilot and exploratory runs. There was this one on July 23: quote:
My own pilot work over the past several weeks with the installation of Avida 1.6.0 on a Beowulf cluster and Ver 1.3.0 on a PC network suggests that even fewer intermediates may be necessary, and that the 'slope' of the fitness function need not be as steep as that of Lenski, et al. Thus the evolution of irreducibly complex assembly language programs that perform the input-output mapping corresponding to EQU are not dependent on any specific set of intermediates nor on rewarding "many" intermediates on some steeply stepped fitness function.
and this one, also on July 23, where I identified the actual number of intermediates (4) that were mentioned in the quotation above quote:
I can't show you the two-intermediate case (yet) - I'm still dealing with installation bugs. But as and if I have time I'll look at it. I know it happens with four intermediates, none of them NAND.
That was after several dozen pilot runs over two weeks with four rewarded intermediates in both 1.3.0 and 1.6.0, simultaneously dealing with the installation bug in ver 1.6.0. Finally there was this one on July 25, after another series of pilot runs made over a week or so quote:
While I haven't done a two-intermediate run in the course of program testing and familiarization, I have done several very long (500,000 cycles) runs rewarding only three 'intermediate' functions. One of them evolved to perform the input-output mapping corresponding to EQU in less than 500,000 cycles. (Recall that Lenski, et al. , used runs of 100,000 cycles maximum.) The test run that evolved a lineage that performed the EQU input-output mapping had a fitness function that rewarded only NOT, OR_N, and AND_N (and EQU, of course) with the same weightings as in Lenski, et al. OR_N is a 2-nand operation and AND_N is a 3-nand operation. The other two-input mappings (corresponding to the 1-nand NAND itself, 2-nand AND, 3-nand OR, 4-nand NOR, and 5-nand XOR) were not rewarded; fitness bonuses for them were commented out of the appropriate program control file.
If serially reporting a series of outcomes from a series of different runs raises a red flag for Nelson, then most research programs that extend over time whose series of results are reported in a series of papers should also raise red flags for him.
Nelson's complete disappointment over the unavailability of the output files is regrettable, but it can't be helped. I'm not doing these pilot runs to study the conditions under which complex structures evolve nor to satisfy Nelson's curiousity. The outcomes I've described here are by-products of exploratory runs performed for other purposes; they were peripheral to the pilot work and I decline to spend dozens of hours doing analyses peripheral to my concerns in these runs. As and when we do systematic studies, actual experiments, the data will be properly analyzed and reported as they're published. (Assuming, of course, that they're publishable.)
Incidentally, I don't understand Nelson's remark that quote:
Particularly because you made such a strong claim concerning the the 3 intermediate run ... which required 2 functions that were not rewarded and yet persisted.
What 2 functions are those? And what is meant by "required"? If it means what I think it means, it's another bit of evidence for my inference that there's a problem with the understanding of what cooption involves.
RBH
P.S. If by "checking your claims" Nelson means to do the runs necessary to see if the EQU mapping evolves with the three intermediates I've identified, he had better be prepared to do weeks of runs, not a day or two. I've been doing 500,000-cycle exploratory runs virtually continuously in Ver 1.3.0 on two fast (2.4 GHz) PCs and in Ver 1.6.0 on a Beowulf cluster with 10 dual-processor machines. If Nelson is running on a single PC he's got a whole lot of runs to do.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 29. July 2003 00:46
RBH wrote:
quote:
If by "checking your claims" Nelson means to do the runs necessary to see if the EQU mapping evolves with the three intermediates I've identified, he had better be prepared to do weeks of runs, not a day or two.
Methinks it is like a weasel ![[Wink]](wink.gif) . Anyway, I'll be back tomorrow or thursday to address the co-option issue.
[ 29. July 2003, 00:46: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 30. July 2003 00:14
RBH writes:
quote:
That remark makes sense only if Nelson believes that evolutionary processes must coopt functions as whole units, in this case NOT, OR_N, and AND_N, pasting them together as discrete units, and cannot make use of parts of functions.
Actually RBH, if those "parts" of functions are also functional, then yes, natural selection can select them. However if they are non-functional, then it is an unselectable step. Is this the only evidence you have that Dembski and I don't understand co-option? I never said partial function could not be selected for, but once again, what does this have to do with flagella? Where are all the simpler functions and their even simpler partial functions that would be selectable?
RBH wrote:
quote:
Since it can make use of parts in both the Avida simulation and in biology (see yersinia's remarks above), I infer that Nelson's conception of cooption is flawed. Since Nelson attributes the same position to Dembski, the same must be the case for him.
But how do you see my conception of co-option as flawed? I don't agree that a function, even it's a partial function, can persist unless totally by chance (in that case it's so simple as to be irrelevant, or it's front-loaded), or if it's rewarded. You stated that two functions persisted and yet were not rewarded,or are you saying that because they were part of the other three functions you listed, they were rewarded? But thats begging the question of how those two functions pre-existed and persisted in the first place.
But if they were so simple that they arose by chance and persisted only because by further chance events were incorporated into the other three functions, then again, I suspect either this is so simple as to be completely irrelevant to Behe's thesis of IC, or that they were front-loaded.
However, all this seems to be moot now in this thread. You can't show me the data, no one but you can independantly confirm anything you have said. So I suggest we stick to the published data which shows many selectable steps that is completely irrelevant to Dembski's discussion of the some biological IC systems.
[ 30. July 2003, 00:16: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 30. July 2003 01:18
Addressing me, Nelson Alonso claimed that quote:
You stated that two functions persisted and yet were not rewarded,or are you saying that because they were part of the other three functions you listed, they were rewarded? But thats begging the question of how those two functions pre-existed and persisted in the first place.
I have just reread all my postings in this thread. I did not state "that two functions persisted and yet were not rewarded, ...". I have said nothing whatsoever in any posting on this thread about functions "persisting," rewarded or otherwise. I did describe the control conditions in the Lenski, et al., study in which evolutionary runs where each single simpler function and all pairs of simpler functions were removed from the fitness function, but I said nothing at any time about the persistence of unrewarded functions. Nelson Alonso has twice asserted the "persist" business; I have not. Either provide a specific reference to such a statement on my part or withdraw your claim.
Regarding Nelson Alonso's comments about my description of the pilot runs I've made, I'll ignore the implications of his remarks. When the systematic experiments are done they'll be published along with the appropriate data.
Nelson Alonso further wrote quote:
But if they were so simple that they arose by chance and persisted only because by further chance events were incorporated into the other three functions, then again, I suspect either this is so simple as to be completely irrelevant to Behe's thesis of IC, or that they were front-loaded.
Nelson Alonso has the simulation, or so he says. Show us specifically where it's frontloaded. Or calculate the probabilities of chance formations: the information is all there. Like John Bracht in the earlier discussion of the Lenski, et al., paper, Nelson talks about chance but doesn't provide any numbers.
RBH
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 30. July 2003 12:43
Nelson: However, you have yet to substantiate that claim.
Pim: Why? You have substantiated it for me.
Nelson: how did I substantiate it for you? I showed that there is no stand alone function besides the T3SS. That is not a substantiation.
You showed that you have no supporting evidence for the sea of non-functionality claim you quoted from Dembski. Hence the dessert claim to refer to the absence of much such data.
Nelson: The flagellum only has 1 alternative function. Where are all the other functions that natural selection can reward that would lead up to flagella.
Still no evidence of a sea of non-functionality. At most a sea of ignorance but that should not be evidence of ID I hope.
RBH notices this as well when he states "Nelson Alonso has the simulation, or so he says. Show us specifically where it's frontloaded. Or calculate the probabilities of chance formations: the information is all there. Like John Bracht in the earlier discussion of the Lenski, et al., paper, Nelson talks about chance but doesn't provide any numbers."
Lenski et al did all the hard work to provide us with the tools and data to allow us to study complex systems, evolution, IC systems etc. As I commented elsewhere it seems ironic that all this work on such systems seems to originate from outside the ID movement. In my opinion ID had the upper hand here to define and direct the research into CSI, IC etc.
Nelson ironically ends his posting with "You can't show me the data, no one but you can independantly confirm anything you have said. So I suggest we stick to the published data which shows many selectable steps that is completely irrelevant to Dembski's discussion of the some biological IC systems."
Is it? Why?
Btw Nelson seems to contradict himself
"I never said partial function could not be selected for,"
and
"I don't agree that a function, even it's a partial function, can persist unless totally by chance"
Which is it Nelson?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 30. July 2003 14:23
I have decided to respond a little further to Nelson Alonso's posting in which he wrote quote:
However, all this seems to be moot now in this thread. You can't show me the data, no one but you can independantly (sic) confirm anything you have said. So I suggest we stick to the published data which shows many selectable steps that is completely irrelevant to Dembski's discussion of the some biological IC systems. (Emphasis original)
In fact that's false: There is sufficient information in my description of those runs so that anyone of marginal computer competence (you just need to know how to use a text editor) with the ability to read for comprehension and an Avida installation (freely available) can do the appropriate tests to independently confirm my claim. That fulfills my obligation to the professional community (and also to Nelson Alonso) to provide enough information that any interested person can replicate the observations. If Nelson Alonso so chooses, he can expend the roughly 3000-4000 CPU hours I have over the last month or so and can test any of my claims about what I've observed.
I am still waiting for documentation (or withdrawal) of his claim that quote:
You [RBH] stated that two functions persisted and yet were not rewarded,or are you saying that because they were part of the other three functions you listed, they were rewarded? (Emphasis added)
RBH
[ 30. July 2003, 14:25: Message edited by: RBH ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 30. July 2003 15:43
RBH,
One interesting question that sounds like it would be simple to answer with your Avida setup would simply be a table with:
# of selected functions --- # timesteps required for EQU
0 --- 500,000
1 --- .
2 --- .
3 --- .
4 --- .
5 --- .
6 --- .
7 --- .
8 --- 50
(just putting in sample numbers IIRC)
Obviously getting enough successful/unsucessful runs to get good averages for e.g. 0 alternative would be tough, but something like this would be a very useful summary, and would give some sense of how much probability amplification occurs due to co-optable intermediates.
You probably already thought of that, but just in case.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 30. July 2003 16:01
yersinia,
Yup, I thought of it. It'd take on the order of 45 - 60 CPU/days (figuring 10 randomizations per condition, which is probably a low-ball estimate of what would be required for decent estimates) to do 10 randomizations for just one set of intermediates from each of the 9 #-of-intermediates levels, not counting data summary and analysis time. I'm not willing to tie up that amount of resources for that length of time at this stage of the game.
That's not even taking into consideration the necessity (if one wants to answer the # of intermediates question legitimately) of running all possible pairs in the 2-intermediate condition, all possible triplets in the 3-intermediate condition, and so on. Each combination would require at least 10 randomizations, so the number of CPU/hours required is enormous.
RBH
[ 30. July 2003, 16:05: Message edited by: RBH ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 30. July 2003 17:49
I missed this post in my replies:
Nic wrote:
quote:
But in fact, once a subsystem exists performing some nonflagellar function, it's perfectly possible for just a part of it (especially if there is duplicated portion of the chromosome containing and extra copy of those genes)
First, that part of it must also be functional. Secondly, organisms have mechanisms to silence gene duplicates. So just bring up these two facets in an ad hoc fashion is still problematic with regard to IC systems.
Nic writes:
quote:
this may be why the flagellum contains homologues to a part of the F1F0-ATPase (FliI)
The similarity could be due to common design rather then co-option events. So the similarity there doesn't really mean much in positive ID terms.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 30. July 2003 18:01
RBH wrote:
quote:
I have just reread all my postings in this thread. I did not state "that two functions persisted and yet were not rewarded, ...".
In a previous post RBH wrote:
quote:
The remainder of the two-input logic functions were not rewarded with reproductive resources; their reward functions were disabled.
This sentence was written directly after this one:
quote:
NOT, a single-input function, and OR_N and AND_N, both two-input functions, as rewarded intermediates
However, it seems as though you are now saying (or originally meant) that the two remaining functions were not rewarded, did not persist, and therefore, were not required.
RBH wrote:
quote:
Nelson Alonso has the simulation, or so he says. Show us specifically where it's frontloaded. Or calculate the probabilities of chance formations: the information is all there. Like John Bracht in the earlier discussion of the Lenski, et al., paper, Nelson talks about chance but doesn't provide any numbers.
How can I provide any numbers when you won't e-mail me any data?
RBH wrote:
quote:
In fact that's false: There is sufficient information in my description of those runs so that anyone of marginal computer competence (you just need to know how to use a text editor) with the ability to read for comprehension and an Avida installation (freely available) can do the appropriate tests to independently confirm my claim.
Well I have access to several networks that could possibly confirm your claim. But of course, you're claim cannot be independantly verified at the moment. In fact, due to the stochastic nature of the program, your claim may indeed never be verified. I need to know exactly how many intermedaites were rewarded, and what the starting conditions were. As well as exactly how many instructions EQU used. If it was a very simple form of EQU, then I wouldn't be surprised that only 3 intermediates were required. If it was one of the complex forms of EQU, which Lenski report required way more than 3 intermediates, then I would need to look at the data, because I find it completely outlandish, that EQU would either be so plastic that 3 intermediates, along with many chance events, would evolve EQU (although that wouldn't really surprise me). However, if the instructions themselves exhibited complexity, but they still required only 3 intermediates, then I would think something was frontloaded.
All I can do is experiment with the simulation and test the things you have already said (which I am currently doing as time permits).
[ 30. July 2003, 18:47: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 30. July 2003 18:06
PvM:
quote:
You showed that you have no supporting evidence for the sea of non-functionality claim you quoted from Dembski. Hence the dessert claim to refer to the absence of much such data.
But you didn't address the fact that the only "stand-alone" function within flagella is the T3SS, again, where are all the other stand alone functions?
PvM writes:
quote:
Btw Nelson seems to contradict himself
"I never said partial function could not be selected for,"
and
"I don't agree that a function, even it's a partial function, can persist unless totally by chance"
Which is it Nelson?
You removed this part of the sentence, which clears away any assertion that it is a contradiction:
quote:
(in that case it's so simple as to be irrelevant, or it's front-loaded)
IP: Logged
|
|
|
Topic Closed
Printer-friendly view of this topic
