|
Author
|
Topic: Evolving Inventions
|
John Bracht
Member
Member # 5
|
posted 09. March 2003 21:44
Yersinia, Francis,
You have it backwards. You two want to assume that bacteria and redwoods are related by an evolutionary history, hence they must be in the same hypervolume. I'm challenging that, and you claim it's tautological. I think the burden of proof is on the making the extravagant claim, which in this case is you. Since you already assume they're in the same hypervolume, you commit the error of simply defining them as being in the same hypervolume (and then accuse me of making a tautology!)
Any metazoan which has a multi-tissue, differentiated bodyplan that originates from a single cell is in a different hypervolume from any single-celled, autonomously replicating organism. It's like Dawkin's biomorphs program and going from a black one to a blue one; moving about in the single-cell hypervolume (and creating variants of single-celled systems) will never get you a re-writing of the genetic program that allows development and differentiation. There's nothing tautologous about it; indeed, you guys are the ones tautologically insisting that they are in the same hypervolume (because of a pre-assumed evolutionary history). I'm saying that when one looks at the hyperspace of possibilities for a single-celled organism, it does not extend to any multi-celled organism with a developmental program. The developmental program consists of a re-writing of the paramters that evolution can evolve in.
Look, I'm sorry if you guys don't like it, but this is the way I've been arguing the issue since I wrote the TRIZ paper. It's not my fault that it's taken this long for you to understand it.
Francis now accuses me of constructing a strawman. But the fact of the matter is that he imposed some other definition of hypervolume on MY argument, and tried to show how it was a weak argument. I'm sorry, but this is very poor debating strategy. If you're going to address my argument, use my terminology to do so--else you ARE constructing a strawman. Period. Furthermore I don't give a darn what the "common" use of the term "hypervolume" is since to my knowledge nobody else is using the term in quite the same way as I am. When I'm constructing an argument, I get to define the terms I'm going to use. And you can respond by saying my definitions are unreasonable or that my argument is weak as-is. But it is illegitimate to argue that I really meant something different and then try to show that the argument is wrong. If Francis continues this tactic I will just start ignoring his comments.
Furthermore, just to clarify for Francis (who keeps on insisting that I haven't clearly defined "hypervolume"): tell me what is wrong with this definition:
quote:
Of course, if you define hypervolume in a different way (and a rather ad-hoc way, it appears) you can show that programs increase it. But if you use it as I've been using it, consistently, all along (as the space of possibilities that a program can explore, given its encoding), then I think it's pretty clear that GA's and trial and error processes can't increase it.
Hypervolume = the space of possibilities that a program can explore, given its encoding
What's so hard to understand here? This definition is clear and concise. What's tautological?
John
[ 09. March 2003, 21:56: Message edited by: John Bracht ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 09. March 2003 22:11
John still insists on attacking a strawman. I stated clearly that I was addressing my interpretation of his usage of the term hypervolume which suggested the number of state variables. Now we realize that instead of using hypervolume to define a rigorous scientific concept, hypervolume basically refers to the solutions possible. Of course such a definition is meaningless since it basically states that a GA cannot find solutions for which it cannot find solutions. As such hypervolume indeed has no relevance anymore and concepts such as GA's cannot increase the dimension of their hypervolume are per definition true.
If that has become the definition of hypervolume then things are even in worse shape for John's arguments then I initially believed because as Yersinia also has argued, we now have come to the realization that hypervolume has become not only a meaningless concept due to its circular notion but it also show how this concept seems to have gone the way of CSI and IC.
So let me recap: A definition of hypervolume which states "the space of possibilities that a program can explore, given its encoding" loses its relevance to the argument proposed by John that GA's cannot increase their hypervolume? While in the original argument it was phrased in concepts of state space it now has been reformulated to be tautological.
Of course GA's cannot increase their hypervolume under such a definition but then the relevance of this to non-routine design has been lost. As shown by GA researchers who have looked into the abilities of GA's to find innovative and creative solutions to non-routine problems, the ability of the GA to not only explore state space but also expand/contract the size of the state space is what allows GA's at least in principle to be creative/innovative. While John's original usage of the term hypervolume seemed to mirror the usage by these researchers, it has also become clear that such a definition of hypervolume would not prohibit GA's from finding creative solutions.
The present definition of hypervolume has made the statement GA's cannot increase their hypervolume tautological but also meaningless as to whether or not GA's can find innovative solutions.
May I also point out that contrary to John's claims that
"Yersinia, Francis,
You have it backwards. You two want to assume that bacteria and redwoods are related by an evolutionary history, hence they must be in the same hypervolume."
I have not made any such claims or assumptions. Whether or not bacteria and redwoods are related by evolutionary history can be supported by quite a bit of independent data. But that's besides the point: I did not make such an assertion.
May I also point out that the new usage of the term hypervolume by John in this posting seems to be begging the question when he claims that:
"Any metazoan which has a multi-tissue, differentiated bodyplan that originates from a single cell is in a different hypervolume from any single-celled, autonomously replicating organism."
Since John has neither shown that GA's cannot explain the origins of multicellularity nor that there is a rigorous mathematical concept of hypervolume. Perhaps John can walk us through the calculations that led him to assert that there are differences in hypervolume since hypervolume does not seem to be a property of solution anymore but a property of the pathway proposed. Under John's original definition the claim that there is a different hypervolume would have made some sense but under the new definition the term hyperspace does not refer to the properties of the "design" anymore. So how did John determine the hyperspace for a multicellular organism versus a single cellular one under his new definition?
May I also inquire as to whether or not John still considers the following a valid definition of hypervolume as well
quote:
For each program, there is an n-dimensional hypervolume of possibilities in which that program operates, with n equal to the number of variable parameters.
And would or would John not agree that if the number of variable parameters is variable that there may be an n+1 dimensional hypervolume which may be reached by the program?
[ 09. March 2003, 22:33: Message edited by: Frances ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 09. March 2003 22:19
Maybe what John is after is a program that interprets itself, and wants to see a new instruction enter the vocabulary of the self-interpreter. That sounds rather fun, actually. Does anyone know if such a thing has been created in a sufficiently minimal form for it to be tractable to run a genetic algorithm on it/them?
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 09. March 2003 22:24
John,
Here's the problem. You're making an argument that a barrier exists. Let's call it X. We are arguing that natural evolutionary processes can cross this barrier.
But you've defined the barrier as:
"Hypervolume = the space of possibilities that a program can explore, given its encoding"
...and therefore whenever we demonstrate that evolution can explore something, then by definition it hasn't crossed X. X has the appearence of being a bold, revolutionary, scientific claim, but actually it just eternally recedes to just beyond whatever has been convincingly demonstrated. Goalposts permanently on wheels, as I said way back in the beginning of this thread when I was first concerned about definitions.
As for the supposed barrier between single-celled critters and multicelled creatures, what prevents mutation and selection from:
1) Causing single-celled critters to stick together in environments where this is advantageous (low-food conditions are one common case)
2) Taking critters resulting from #1 and causing specialization of function of various cells? E.g., I think sponges have only three different types of cells.
...and yet you are claiming that these are "different" hypervolumes that cannot be crossed based on the logical circle you constructed rather than any evidence that either of these two processes is difficult for any reason whatsoever.
Evolution of animals – choanoflagellates and sponges is a brief intro to how recent molecular data confirmed the 100+ year old idea that the basal metazoans -- sponges -- were derived from colonial choanoflagellates. If you like we can start a thread on the origin of multicellularity -- spring break perhaps, everyone is busy about now...
[ 09. March 2003, 22:26: Message edited by: yersinia ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 09. March 2003 23:00
Another possible cause of multicellularity includes predation
"Martin E. Boraas, Dianne B. Seale, and Joseph E. Boxhorn. Phagotrophy by a flagellate selects for colonial prey: a possible origin of multicellularity. Evolutionary Ecology 19: 153-164, 1998."
See also Review: Volvox carteri: Molecular Genetics of Cell differentiation
Slides on Emergence of Multicellularity Furusawa & Kaneko (1998)
Or On the Evolution of Multicellularity and Eusociality By Larry Bull
quote:
In this article versions of the abstract NKC model are used to examine the conditions under which two significant evolutionary phenomena - multicellularity and eusociality - are likely to occur and why. First, comparisons in evolutionary performance are made between simulations of unicellular organisms and very simple multicellular-like organisms, under varying conditions. The results show that such multicellularity without differentiation appears selectively neutral, but that differentiation to soma (nonreproductives) proves beneficial as the amount of epistasis in the fitness landscape increases. This is explained by considering mutations in the generation of daughter cells and their subsequent effect on the propagule's fitness. This is interpreted as a simple example of the Baldwin effect. Second, the correspondences between multicellularity and eusociality are highlighted, particularly that both contain individuals who do not reproduce. The same process is then used to explain the emergence of eusocial colonies.
Keywords: Baldwin effect, eusocial colonies, genetic algorithm, multicellular organisms, NKC model
Adventures in Multicellularity The social amoeba (a.k.a. slime molds)
While we surely do not have (all) the answers, science seems to have formulated some very interesting hypotheses on the origins of multi-cellularity.
Enough for today since I have gone over my three postings and despite the moderator's comments to Nelson and others that such limitations are to prevent non-relevant topics/discussions and not to perse prevent discussion, it seems a good time to spend with my loved ones.
[ 09. March 2003, 23:01: Message edited by: Frances ]
IP: Logged
|
|
Argon
Member
Member # 276
|
posted 10. March 2003 09:52
I'm curious about the ability of humans to invent in the context of reaching alternate hypervolumes.
If one hypervolume is inaccessible to things in another hypervolume, then how is it that humans display inventiveness (so defined)? How do we jump to different hypervolumes? Couldn't one claim instead that we are simply operating within the "human-accessible" hypervolume?
[ 10. March 2003, 11:41: Message edited by: Argon ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 10. March 2003 13:42
Rex wrote quote:
Maybe what John is after is a program that interprets itself, and wants to see a new instruction enter the vocabulary of the self-interpreter. That sounds rather fun, actually. Does anyone know if such a thing has been created in a sufficiently minimal form for it to be tractable to run a genetic algorithm on it/them?
I don't know offhand of it being done, but I can imagine doing it in something like a minimal LISP. But I'll be darned if I'm going to spend the time on it.
RBH
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 10. March 2003 17:09
I wouldn't mind constructing such a program--once I have time, which is unlikely to occur soon--as long as we could clearly decide upon the conclusions we would draw from various different observations. I have exactly zero interest in wasting my time coding a program whose results will be dismissed as "within the hypervolume" regardless of its output.
IP: Logged
|
|
ASCSCommanding
Member
Member # 682
|
posted 10. March 2003 17:25
Gedanken,
quote:
One point that has still not been addressed is my example of the software that had no limit set on the number of “genes” or other “dimensions”, and adds dimensions when it decides to do so. More “dimensions” or greater search range may depend on the search time, but is not inherently limited in the algorithm. In this case the theoretical dimensionality of the raw search space is effectively infinite. Thus no “inventiveness” could possibly happen no matter what was invented, because all points were reachable within the infinite search space -- it is simply sifting of an ever greater and semi-infinite range of possibilities.
I agree that this is where we are headed in this discussion.
quote:
An example: A state is reached by some sort of direct jump to a greater degree of complexity. But it could, alternately, be reached by a series of small steps. Is the first case “inventive”, while the second not? The solution is exactly the same, the only question was how it was reached.
This too is on the mark. I discuss this at length below.
John,
I do appreciate that you’ve finals and all, and have rather a lot of people to respond to. Sorry if we’re piling on.
quote:
While some cases are not very clear-cut, there are definitely cases where we can point to both GA's and biological changes that would qualify as inventive. For GA's, consider again Dawkin's biomorph example. The Blue biomorph is clearly in a different hypervolume and is inaccessable from the hypervolume of the initial program. There are clearly two hypervolumes of possibilities here. For biological examples, Dawkins again leads the way with his insight that humans cannot just sprout angel wings from their backs--to do so would require re-engineering the developmental process such that angel wings are possible (in other words, such that this new solution is included in the hyperspace of possibilities). Furthermore, I can see examples where precisely this sort of change has occurred. The origin of developmental programs is precisely this sort of change, where the constraints that map out a hypervolume of possibility come into existence. I map out other examples in my paper. Perhaps you can explain why you feel that my examples can be included in the same hypervolume. How, exactly, is a single-celled organism in the same hypervolume as a metazoan? Surely there are possibilities available to the metazoan body plan which are simply not possible for the single-celled organism? (Things like arms and legs strike me as a good example, or circulatory/respiratory systems).
As you say, the blue biomorph is inaccessable in the initial program. So, it would seem that two objects are in different hypervolumes depending upon the underlying program. As for the angel wings example, that’s not so clear to me. I agree, of course, that humans are not just going to go sprouting angel wings, but that only indicates that there is no way to make that change directly. It would seem that for the issue of hypervolumes to be relevant then the question is could there be a trial-and-error search from some starting point to each of the two ending points. The absence of a direct pathway does not seem to be sufficient to put two entities in different hypervolumes. I note, for an admittedly problematic example, (but I don’t think the problems detract from my point) that none of your examples of separate hypervolumes in living things postdate the rise of vertebrates. If humans and condors can lie in one hypervolume it is certainly not clear to me that something very much like humans with wings [BOLD]must[/BOLD] be in separate hypervolume from humans. However silly that example might seem, the basic question is this, if there does exist a trial-and-error pathway from some starting point to each of two ending states, but to go from either ending state to the other requires inventiveness, are the two ending states in one hypervolume or two? The answer to this question could help me put together a better example to illustrate my thinking.
Here, and later, you state that multicelularity lies in a separate hypervolume from single celled life, but I don’t really see a demonstration of this. I’m not sure that I can show that they are in one hypervolume, but unless you provide evidence that separate hypervolumes exist, my failure would only leave the question undecided. Is your basic claim that there is good reason to conclude separate hypervolumes, and therefore natural mechanisms cannot account for the diversity of life, or that the question of hypervolumes is undecided?
That said, I will readily admit that I don’t see much problem in putting multicelled and single celled creatures in a single hypervolume. It seems like trial-and-error could lead single celled organisms to operate in close proximity, from there to have them be linked in different geometries, from some of those geometries to tubes, etc. Perhaps it’s just my imagination, but I don’t see the sharp divide. The appearance of a divide seems to come up only when looking at far removed points within the hypervolume. So yes, single celled creatures are not going to immediately go to having arms and lungs, and yet there seem to be numerous intermediate stages linking those states. Yersinia has posted a fair bit on this as well.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 10. March 2003 20:00
Nelson:
The point about telling a story about how a bacteria got a flagella, has to take into account the complex interactions and the irreducible nature of the bacterial flagellum itself. Irreducible complexity is a pattern that is not a fabrication.
Francis:
And how do we determine whether or not the pattern is not a fabrication?
Nelson:
A fabrication is an ad hoc pattern. The difference between a fabrication and a specificiation can be distinguished with statistical rigor as Dembski has done in Chapter 2 of No Free Lunch , and The Design Inference. As Dembski says:
quote:
"Unlike specifications, fabrications are wholly unenlightening. We are no better off with a fabrication than without one."
You can tell a pattern is a fabrication when you have to read it off the event in order to get it. If you are not the mathematical type you can simply ask why a particular pattern exhibits specificity and, if you get no answer, then you can be sure it's a fabrication.
Francis:
I would say that since the probability of the system would depend on our knowledge we may have a problem. In fact as Sobel has argued, specification is trivial for almost any event, making many chance events appear to be designed.
Nelson:
I don't know who Sobel is. Can you quote his argument? I don't want to take this thread off topic, since it's dedicated to TRIZ. Suffice it to say that everything about science depends on our knowledge. This doesn't make it a problem.
Francis:
Nelson's example about the poker hand seems irrelevant since in the case of the flagellum the specification is NOT done after the fact. But using Sobel's approach one can generate specification which includes all possible hands in poker hence the issue of specification seems to be trivial.
Nelson:
Again, I have no idea who Sobel is but he's wrong. A specification can be attributed after the fact, such as getting prime numbers sent from outer space.
Francis:
The case of the flagellum is more like a poker hand which has been dealt which now needs to be specified. For many hands I am sure that one can find some card game in which such a hand was dealt.
Nelson:
The flagellum is not an example of just any outcome, since it's pattern is not ad hoc. In all bacterial genomes , I see the same 20 core parts, or I see none at all. The pattern itself is irreducible complexity.
Francis:
From Sobel
quote:
From this second illustration can be gathered that Dembski's theory enables a moderately imaginative person, with a list of possible delimitations of an event, easily eliminate relevant chance-hypotheses for the event; if they all make more probable that not its non-occurrence, and avoiding 'false negatives' concerning relevant chance- hypotheses for this event is somewhat (it need not be very) important to him. From the two illustrations, one may gather that by the lights of Dembski’s book, we are entitled, and will always be entitled to conclude, that not much happens by chance.
Thus when Nelson states "When probabilities descend and they are attached to specifications, we usually attribute such things to intelligent design." he has to take into consideration the likely false positives.
Nelson:
Sobel was not talking about the risk of false positives here. But that Dembski's notions would lead to the conclusion that "not much happens by chance". The possibility of false positives is a risk for any theory. Unless Francis has a false positive in mind, I don't think that this criticsm amounts to much. Secondly, Sobel's point here is not about the risk of false positives
Francis:
Dembski seems to be struggling with this concept, on the one hand he seems to accept false positives but on the other hand he has to reject any such occurrence for the design inference to have any practical value.
Nelson:
Where does Dembski accept a false positive?
Francis:
Thus when Nelson states that "I really don't see how showing the possibilty of an intermediate step results in a more likely pathway. As Dembski stated in his reply to Miller, it is possible that pure chance alone can build the flagellum. But even evolutionary biologists regard this as unlikely and therefore reject it. It is not what is logically posssible but what is empirically probable. We can now consider the possibility space as smaller."
And yet showing intermediates does reduce the complexity of the whole system.
Nelson:
Showing possible intermediates does not reduce the complexity of the whole system. Removing a component and having the thing still work reduces the complexity.
Francis:
If Dembski is arguing against a pure chance flagellum construction then I wonder why he spent so much effort on this strawman when calculating the chance probabilities for the flagellum?
Nelson:
Dembski responded to this:
quote:
In No Free Lunch, I offer a way to try to get a handle on such progressions through what I call perturbation identity and tolerance factors (see section 5.10). The idea is to take a functional system, perturb it, and determine how perturbation affects the probability of retaining function. If the probability of retaining function is high, then this would constitute evidence that a Darwinian pathway could readily lead to the system in question. Essentially the idea here is one used in AI search strategies. Miller's task, to vindicate Darwinism in regard to the flagellum, is to exhibit a forward chaining search through genomic space that issues in a genome coding for the flagellum. But neither he nor anyone else in the biological community can do this. So an alternative approach is to try a backward chaining search that preserves function. What I show through my perturbation probabilities is that such searches face huge probabilistic hurdles. What this means is that if a forward chaining search succeeds, it does so as a highly specific and isolated path through genomic space. In that case the step-by-step probabilities moving forward from A_i to A_(i+1) could still be large enough not to overturn my universal probability bound. But absent a successful forward chaining search, there is no reason to think that success is even possible. Successful forward chaining assumes that a sequence like A_1 through A_n and can be made explicit. There is no evidence of this.
In other words, what Dembski showed in his book was not really a random spontaneous formation of the flagellum, but he actually, like Dermott did with RNA, put in mathematics that the flagellum was irreducible (by showing how improbable backward-chaining really is).
To make this relevant to TRIZ (the subject of this thread), Dembski states:
quote:
In fact, if we look to human invention, we have all the more reason to think that the Darwinian mechanism cannot account for successful forward chaining searches and thus for systems like the bacterial flagellum. The field of technological evolution broadly distinguishes between routine and innovative problems (see Savransky 2000 as well as Dembski 2001 and Bracht 2001). Routine problems are amenable to trial-and-error problem-solving techniques (of which the Darwinian mechanism constitutes an instance). Innovative problems, by contrast, require conceptual insights that transcend trial-and-error tinkering. Moreover, in human experience, irreducibly complex designed systems are invariably solutions to innovative, not routine, problems. Since we don't expect trial and error to produce irreducible complexity in the human context, why should we expect it to produce it in the biological context? The usual counterargument here is to charge anthropomorphism and invoke deep time -- natural selection should not be compared to human activity and natural selection has unimaginably more time to work with than human trial-and-error tinkering. But neither of these criticisms holds water. Humans can mimic undirected selection and they can now do it very fast on the computer, thereby compressing deep time into ordinary time. And nevertheless, it remains the case that no genetic algorithm or evolutionary computation has designed a complex, multipart, functionally integrated, irreducibly complex system without stacking the deck by incorporating the very solution that was supposed to be attained from scratch (Dawkins 1986 and Schneider 2000 are among the worst offenders here).
By the way, you're jumping from point to point here and you are doing so via assertion, it's hard to see what exactly your point is.
Francis:
So far the mere possibility of indirect
Nelson:
The logical possibility of an indirect route does not speak to the improbability of such routes. Again, it is quite possible for random chance alone to get a flagellum. But even evolutionary biologists reject this. But indirect routes smuggles in too many random chance events , as shown by irreducible complexity and Dembski's calculation, then, although it too is possible, it should be likewise rejected.
Francis:
and even direct Darwinian routes
Nelson:
As even Darwinists David Ussery and Thornhill admit, direct Darwinian routes can be successfully elimanted by irreducible complexity.
Francis:
suggests that ICness is not a helpful concept in determining likelihood of design.
Nelson:
Since IC is able to eliminate direct routes and show that indirect routes are unlikely, it certainly is helpful in determining the likelihood of it's intelligent design.
Francis:
Dembski may want specific scenarios but specific scenarios do not make an ID inference more or less likely or do they?
Nelson:
They sure do. If a specific scenario can be shown in detail, as Dembski suggests, which can be tested, as Behe suggests, which breaks down the flagellum and shows a likely causal history for it's evolution, then yes it would make the design inference less likely.
Francis:
Nelson may object to the stories so far but what Nelson cannot do is conclude that there is evidence of design in the flagellum since that depends on showing that no indirect pathways to IC systems exist.
Nelson:
Showing that indirect pathways are unlikely is enough to warrant a design inference. It also helps that I can show that no indirect pathways exist, since the T3SS post-dates the flagellum.
Francis:
At the moment the best we may be able to suggest is: "We don't know". I doubt that this would be an acceptable design inference for most people :-)
Nelson:
And here I thought that ID was a science stopper. When it came to the bacterial flagellum, it was "we don't know", when it came to the lobster eye, it was "we don't know", when it came to gene duplication, it was "we don't know". When it comes to all the systems that Behe talks about, it's "we don't know". This string of "we don't knows" is not going to get shorter, it's going to get longer. This is why intelligent design's time, in my opinion is now. While Darwinists sit around and twiddle their thumbs saying "we don't know", scientists still need to earn their paychecks. It seems as though it is stacking the cards for Darwinism to say "we don't know" when another alternative explanation can be just as fruitful a scientific endeavor, where Darwinism has failed.
Francis:
Or as Nelson states it so clearly:
Nelson:
I didn't say this, I was quoting Dembski.
quote:
Science must form its conclusions on the basis of available evidence, not on the possibility or promise of future evidence. This means that eliminative inductions need to be local inductions, based on detailed testable models and hypotheses that are currently available.
Francis:
Science thus must form its conclusion s not on the possibility or promise of future evidence of intelligent design.
Nelson:
ID theorists don't base it's conclusions on promises of future evidence for intelligent design. I regard the bacterial flagellum as evidence for intelligent design right now. Which is why it has no evolutionary history. What I hear more often then not, and even you stated this in the front-loading thread, is that "just because we don't have an explanation for it's evolution now doesn't mean we won't in the future".
Francis:
This means that eliminative inductions need to be local inductions, based on detailed testable models and hypotheses. But Dembski also suggests that such models are not likely forthcoming from an intelligent design perspective making the concept of ID not very attractive as an example of what Dembski believes science should be doing.
Nelson:
Where does Dembski suggest that such models are not likley forthcomming? What specific models was he referring to? Quote Dembski.
Nelson:
"Now in my opinion, the possibility that the bacterial flagellum arose from a primitive export machine has been effectively eliminated due to the amount of pure chance events invoked to get the job done."
Francis:
We seem to be able to identify the following problems 1) Pure chance is not invoked by such mechanisms
Nelson:
They are invoked by such mechanisms. An export machine that is selectively advantageous has the claws of natural selection imbedded in them. Pure chance alone can only break them free in order for them to spontaneously change the function that selection was once pruning. This has to happen multiple times. These are pure chance events.
Francis:
2) Since we do not fully know the full pathways for likely events we cannot ascribe any probabilities.
Nelson:
How can you say that you cannot ascribe any probabilities for those events when you just described those events as "likely"? Whether an event is "likely" or "unlikely" depends on probability. Furthermore, that Darwinists have laid their cards on the table about how the flagellum evolved, we can , by looking at the systems in question, and whether they are irreducibly complex, ascertain their probability.
Francis:
I may agree with Dembski that detailed evolutionary pathways may be lacking but ID seems to suffer from similar and if I understand Dembski correctly unsurmountable problems.
Nelson:
Which ones?
[ 11. March 2003, 21:45: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
NeilUnreal
Member
Member # 676
|
posted 10. March 2003 21:30
re: Rex & RBH, self-modifying programs.
It's been a while since I had automata theory, but...
It seems to me any self-modifying Turing machine with access to a "stochastic register*" could be converted to any other Turing maching by the following process: 1) start with a Turing machine which which writes a random sequence of bits from the stochastic register to a blank area on the tape; 2) self-erase the first machine and begin executing the new sequence. The second sequence could be any possible Turing machine (though it's unlikely to be any given Turing machine).
This implies that either 1) self-modifying Turing machines occupy difference hypervolumes, but there exist Turing machines capable of escaping their hypervolumes, or 2) all self-modifying Turing machines occupy the same hypervolume, so all things which can be described using self-modifying Turing machines are in the same hypervolume.
-Neil
*i.e. Under the influence of QM it could be anything capable of storing a bit.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 11. March 2003 02:06
Neil: actually, something very much like a minimal Turing machine is what I had in mind. However, John isn't using any standard definition of hypervolume. By his definition, no Turing machines would be in the hypervolume because to get any given Turing machine is too unlikely.
Nelson and Frances: It's a good idea to carefully define what you mean by "irreducible complexity". The naive interpretation of Behe's original version published in DBB had serious flaws, and I normally assume that this is the one people are talking about when they say IC. Some of the flaws have been removed in later versions.
Nelson wrote:
quote:
indirect routes smuggles in too many random chance events , as shown by irreducible complexity and Dembski's calculation
I'm not sure what he bases this claim on. Dembski's calculation in NFL absolutely does not take indirect routes into account. NFL, p.286:
quote:
The fundamental intuition underlying irreducible complexity is that irreducibly complex systems cannot be substantially simplified and yet preserve function. The Darwinian mechanism requires such simplification if step by gradual step it is to succeed in generating an irreducibly complex system.
Dembski argues that the flagellum cannot be substantially simplified and yet preserve function.
But the T3SS shows that it can be substantially simplified and yet preserve function--a different function, but the organism hardly cares about that as long as it aids survival.
Dembski's argument wasn't that indirect routes smuggled in too many random chance events--it was that by inspection of a flagellum we could conclude that there were no indirect routes. Oops.
[ 11. March 2003, 02:10: Message edited by: Rex Kerr ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 11. March 2003 16:10
Rex:
It's a good idea to carefully define what you mean by "irreducible complexity". The naive interpretation of Behe's original version published in DBB had serious flaws, and I normally assume that this is the one people are talking about when they say IC. Some of the flaws have been removed in later versions.
Nelson:
What serious flaw are talking about? As for which definition I'm using, take your pick. They are all equivalent as far as I can see.
Nelson
indirect routes smuggles in too many random chance events , as shown by irreducible complexity and Dembski's calculation. I'm not sure what he bases this claim on. Dembski's calculation in NFL absolutely does not take indirect routes into account.
Rex:
Im not sure what he bases this claim on. Dembski's calculation in NFL absolutely does not take indirect routes into account. NFL, p.286:
Nelson:
This quote states my point quite clearly, and I'm basing this claim on the perturbation and tolerance factors of Dembski's calculation.
quote:
The fundamental intuition underlying irreducible complexity is that irreducibly complex systems cannot be substantially simplified and yet preserve function. The Darwinian mechanism requires such simplification if step by gradual step it is to succeed in generating an irreducibly complex system.Dembski argues that the flagellum cannot be substantially simplified and yet preserve function.
Rex:
But the T3SS shows that it can be substantially simplified and yet preserve function--a different function, but the organism hardly cares about that as long as it aids survival.
Nelson:
You cannot remove a substantial portion of the parts, in fact, you can't remove any of the core parts, and still preserve function --- motility, you have to change the function. Dembski's perturbation and tolerance calculations showed this. Preservation of function is impossible with an IC system. You have to change it, not preserve it. This is why you need to invoke co-option. Behe, back in 1996, as you yourself pointed out, admitted this much, that indirect routes were possible. But as I keep saying, indirect routes get more improbable the more complex the irreducible system is. It doesn't matter what the organism cares about, it matters what natural selection has to work with, and how much pure chance has it's hand in the process.
Rex:
Dembski's argument wasn't that indirect routes smuggled in too many random chance events--it was that by inspection of a flagellum we could conclude that there were no indirect routes. Oops.
Nelson:
That is completely false. Both Behe and Dembski state qiute clearly that there can be indirect routes. The fact that you have to go from motility to protein secretion (which has nothing to do with motility) shows just how irreducible the function is. You cannot preserve the function at all by removing any part from a bacterial flagellum. You have to change the function in order for your story to work.
[ 11. March 2003, 16:17: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 11. March 2003 19:45
quote:
Nelson:
You cannot remove a substantial portion of the parts, in fact, you can't remove any of the core parts, and still preserve function --- motility, you have to change the function. Dembski's perturbation and tolerance calculations showed this. Preservation of function is impossible with an IC system. You have to change it, not preserve it. This is why you need to invoke co-option. Behe, back in 1996, as you yourself pointed out, admitted this much, that indirect routes were possible. But as I keep saying, indirect routes get more improbable the more complex the irreducible system is. It doesn't matter what the organism cares about, it matters what natural selection has to work with, and how much pure chance has it's hand in the process.
I want to thank Alonso for focusing our attention on the primary misunderstandings present in the ID position.
As Rex has pointed out, what “natural selection” has to work with is systems that are using most of the same parts (parts of the so-called “IC” system in question) functioning as part of a different system that performs a different function.
This helps point out how the ID position typically tries to present the evolutionary scenario in stark contrasts, as though the organism had no process of motility for example if the flagellum was not functional in the form labeled “IC”.
And apparently this “different function” gives a degree of advantage over the ancestor, in some specific environmental niche. This means that the particular organism can reproduce in that niche.
Only later is the chance event of the further change from the previously useful components to the new function.
What is very important is the mistaken view of a “search” for the particular function that is identified. But there is no reason to expect if the “tape” were rewound that evolution would produce the flagellum. There are many different patterns that could have developed, and which would be equally effective. In fact we see precisely that in the tremendous variety of lifeforms we can readily observe.
So the greater the complexity is correctly interpreted by Alonso as lowering the probability of taking a particular exact pathway to a particular system from which no part can be removed without that particular function failing to be present. But of course if the “tape” were rerun, one would expect other functions to develop.
And since each and every function whatsoever has a minimum number of parts below which that particular described function no longer occurs according to the precise description (and thus every function that can be identified whatsoever has an irreducible core) each new function that might have developed could also have been claimed to be IC and also of low probability. And the claim would be correct, that different function did develop with low probability. But the probability is high that some set of functions would develop -- functions that will be beneficial to the ancestors. And given the just prior states, the next state that was developed would be likely to have a high probability measure if that could be calculated by knowing sufficient detail of the evolutionary history.
The “contradiction” of TRIZ is found when a particular difficulty arises, like the motility problem. But one could say that the “flagellum” is produced with overcoming a “contradiction” if it were arrived at directly. But arriving at it indirectly is exactly correctly labeled as “improbable”. The problem is that one has not identified that arriving at some solution to the so-called “problem” was improbable -- only arriving at the particular so-called “solution”. But once again nature was not “searching” for that particular solution -- rather variations were produced by natural processes, and those that had advantage became more prominent. That the changes that gave greater fitness must have produced some sort of function seems obvious.
The entire issue of claiming “low probability” is a combined failure to recognize that the further a tree branches the inherently lower the probability of any branch becomes -- and this is inherent in the tree structure and is not an aspect of the particular solution arrived at. (For one think one must consider what starting state is the reference for the probability calculation -- further back in the branching tree gives lower probability for the very same specified state, and thus the probability of a given state is not a single number but a function of the current and selection of a past state.) And then the lack of recognition of the tremendous density in the so-called “search space” of the different “solutions” or functional alternatives is not recognized. This is why there tends to be a focus on reaching or searching for the “particular” solution -- and finding that the TRIZ process might have been required to resolve the apparent “contradiction” of the particular advancement. As the complexity of systems increases, the number of possible functions that can be generated involving varying subsets of the components also increases -- thus providing greater density in the possible branches to be taken and greater density in potential “IC” so-called “solutions”. (But they are solutions to a perceived “problem”, and are only better or more fit alternatives in the actual organism form “space”. Solving a different “problem” may be equally likely and equally beneficial to the succeeding generations.)
No problem has been presented for evolution, but a good example to clarify the arguments has been given. And premises that were correct but with missing consideration of their consequences gives an opportunity to explain how evolution actually works.
[ 11. March 2003, 20:28: Message edited by: gedanken ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 11. March 2003 21:26
Ged:
I want to thank Alonso for focusing our attention on the primary misunderstandings present in the ID position.
Nelson:
Your welcome, but you got it a little wrong here. I am focusing on the misunderstandings present in the minds of ID critics about the ID position.
Ged:
As Rex has pointed out, what “natural selection” has to work with is systems that are using most of the same parts (parts of the so-called “IC” system in question) functioning as part of a different system that performs a different function.
Nelson:
Rex actually pointed out to a small subset of the flagellum within the Type III secretory system. Not "most of the parts". Just some.
Ged:
This helps point out how the ID position typically tries to present the evolutionary scenario in stark contrasts, as though the organism had no process of motility for example if the flagellum was not functional in the form labeled “IC”.
Nelson:
The the ID position here is scientifically accurate. You cannot remove any component from the flagellum without losing motility.
Ged:
And apparently this “different function” gives a degree of advantage over the ancestor, in some specific environmental niche. This means that the particular organism can reproduce in that niche.
Only later is the chance event of the further change from the previously useful components to the new function.
Nelson:
Exactly. So with this ancestral function that has absolutely nothing to do with motility,(protein secretion), natural selection would conserve it. Unfortunately, what you want is a flagellum that functions for motility, not protein secretion. To get that, you have to invoke multiple chance events.
Ged:
What is very important is the mistaken view of a “search” for the particular function that is identified.
Nelson:
This is irrelevant. What John Bracht is doing with TRIZ is showing that a particular solution is needed to solve a particular problem. The problem to be solve? How can we get bacteria to move despite being in a "brownian storm"? The blind watchmaker can never hope to solve this particular dilemma precisely because it is not a search for that particular solution (or any solution for that matter). Intelligent design can.
Ged:
But there is no reason to expect if the “tape” were rewound that evolution would produce the flagellum. There are many different patterns that could have developed, and which would be equally effective. In fact we see precisely that in the tremendous variety of lifeforms we can readily observe.
Nelson:
Yes what we see among these organisms are very effective solutions to technical contradictions, such as the eukaryotic flagellum which is also IC. The main point here is that evolution unfortunately, is a blind search. So those who espouse Darwinian evolution and hold it as a blind search would have to see things as different when the "tape is rewound" (although there are other views held by biologists such as Michael Denton, that no matter how many times you rewind the tape, you will get flagella.)
However, as I said, no IDer has ever claimed that a blind search must always get our bacterial flagellum. So lets test it. Start with an export machine. Mutate and artificialy select for motility. See if we get an irreducibly complex structure that is used by the organism to move. No ID critic has ever accepted this challenge . If you say that this experiment wouldn't prove anything either way because we cannot test something that occured millions of years ago by random events, well then, I must question the utility, the falsifiability, and testability, of Darwinian theory. It boils down to that bad weatherman prediction, it will rain or it will not rain tomorrow.
Now, TRIZ details exactly why IC systems are such a problem for Darwinian evolution. And you accentuate this problem for Darwinian evolution when you correctly state that Darwinian evolution is not looking for that particular function, it is blindingly grasping around in the dark taking anything it can get, cobbling things together like a clumsly baby playing with legos, or locking on to a particular function "created" by a random mutation.
With IC systems the problem becomes greater for Darwinian evolution. Not only is natural selection completely blind to the solution it needs for the particular problem (a propeller strong enough to sail a brownian storm), it can't get there by walking a straight path. It has to go around in circles holding on to a different function, perhaps never attaining motility at all. Like a ping pong ball that has no help in getting to that slot at the end of the ping pong machine, it bounces around in a circuitous route fruitlessly conserving the wrong solution for the bacteria's dilemma of sailing the brownian storm. Pure chance alone can only hit natural selection in the ass in order to get it to another function, and this has to happen multiple times.
As you so correctly point out, we can define neo-Darwinian evolution so that it excludes intelligent design and instead involves only random genetic changes and natural selection/drift, an intelligent designer would have the particular solution in sight.
So as you also correctly point out, either IC molecular machines evolved from the combination of more simple machines that had different functions, or they were built by an intelligent designer who indeed could search the space of possibilities with the particular solution in mind.
What do we find in nature instead? Do we find the blind cobbling together of different functions using different parts, from wiggly things, to a somewhat functional flagellum, to the efficient, precise, fine-tuned, bacterial flagellum? Or do we find a core 20 parts that is extremely resistant to the clumsly cobbling of natural selection or the reductionism of Darwinian evolutionists? As Stanely Salthe so eloquently put it,
quote:
Moving now into consideration of details of the formal properties of the idea of natural selection, we can start very broadly by noting that it is basically a theory of, as Einstein might have remarked, higgledy-piggledy. That is, it is a theory of constraints on randomness -- or, indeed a theory of accidental changes.
http://www.nbi.dk/~natphil/salthe/anacri.99.08.html
Now as far as the claim that the further a tree branches the more likely it is that a solution will be found, I disagree. I disagree because IC eliminates so many of these possibilities by it's very own nature. As we see it eliminates all direct pathways. Direct pathways lies in a zone where the potential for probable pathways mostly exist.
Instead, what do we have? We only have indirect pathways. All indirect pathways to date have been shown to both not exist and/or be completely unlikely. So there is no large branching of abstract possibilities. And vaguely pointing to such things is also untestable and unfalsifiable, therefore it is not science. There are so many ways that the flagellum can come about without natural selection, through pure chance. There are perhaps millions of ways but all of these ways are rejected even by Darwinian evolutionists. This is exactly why we must be constrained by local inductions and Dembski's probability calculation is paving the way.
Let me not be so negative here. There is a positive side here for intelligent design. I think that a lot of IC systems do indeed, or will in the future, exhibit similarity that might look like it was the result of co-option. In fact, I would go ahead and say that the origin of IC systems were the product of co-option. The problem is I think co-option is so unlikely to have been accomplished by natural selection,drift, and random mutation that if it did happen this way, it was done through intelligent design. In fact, one can even make a testable ID hypothesis from this.
[ 11. March 2003, 22:03: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
|
Printer-friendly view of this topic
