|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
Jack Foster
Member
Member # 79
|
posted 09. June 2003 02:03
Hi RBH:
(Just a note: I still can't read the paper you've linked, though the abstract is available. It's in postscript form I believe, which to my knowledge is not supported by my computer.)
quote:
In this domain of discourse a fitness landscape is not "determined by the internal properties of the replicator itself." If you want to use some sort of definition that invokes that notion, feel free, but know that it will only engender confusion since it's a purely idiosyncratic usage. To be blunt, no one will understand what you mean.
Fair enough. Let me tell you what I mean, and then you can tell me how I should have phrased the sentiment.
It's ridiculous to consider fitness landscapes divorced from populations of organisms. Panthers have different genotype space than do prokaryotes after all; therefore the fitness landscapes drawn upon their respective spaces must be of different scale, scope, and shape. I think of this as "internal" factors that affect the shape of fitness landscapes, and conversely, all of the intricacies of environment I think of as "external" factors that help shape fitness terrain.
There are some landscapes that exist completely independent of external factors. That is, regardless of changing environmental conditions, the shape of the fitness terrain remains the same. An example might be an indispensible, machine-like part that has no tolerance for variation . . . a heart-valve for instance. Regardless of environment, a mal-formed heart valve will produce a reduction in fitness.
W&A give another example:
quote:
Among the earliest experiments in evolutionary computation, Friedberg (1959) attempted to evolve functioning computer programs by mutating and selecting the code, but found that mutations effectively randomized the behavior of the programs, and adaptive evolution was impossible; there is no way to improve the performance of a conventional computer program by randomly altering letters in the source code. It became understood that the mutation/selection process is not universally effective in producing adaptation if favorable mutations cannot be produced (see for instance Bossert (1967), Bremermann et al. (1966), Eden (1967), or Simon (1965)).
In these cases, the fitness landscape is, as I said, "determined by the internal properties of the replicator itself." By "internal properties", I was referring to the genotype to phenotype map.
On to the mousetrap:
quote:
With respect to the evolving mousetrap analogy, I don't think it's at all useful because it is not an accurate model of relevant features of the Avida simulation. It displays a lack of understanding of what the Avida virtual machine is capable of. While the instruction set in Avida obviously contains the necessary raw material (instructions) for a program that performs EQU, Avida's virtual machine is Turing-complete, meaning that it has the necessary 'raw material' to perform any computable function. One can't say that about the mousetrap analogy: the parts list of that analogy is tightly constrained in a way that the Avida instruction set and virtual machine architecture is not. About the only functional 'complex' structure that one can construct (or evolve) using all the mousetrap parts is a mousetrap. However, the only constraints on the Avida system that evolved programs that perform EQU are precisely the constraints imposed by Darwinian evolutionary theory. So the evolving mousetrap analogy fails for lack of alternative capability: all it can evolve is a mousetrap. The Avida simulation in principle could have generated anything but didn't; it evolved what the evolutionary mechanisms permitted. And what it evolved in the Lenski, et al., study are structures - assembly language programs - that meet the operational definition of "irreducibly complex."
Turing-completeness is a red-herring. Biological evolution isn't "Turing-complete". And why do you assume that all that my system can evolve is a mousetrap?
quote:
The Avida simulation in principle could have generated anything but didn't;. . .
It could not have generated the phrase "Methinks it is a weasel", nor could it have generated a mousetrap. It was oriented towards a specific solution space. I agree that it evolved IC, but it did so by lowering the probabilistic hurdle associated with this particular form of IC. My system will evolve IC as well, and in the same way.
But once again, Behe never said that it was impossible for IC to evolve indirectly; he said that the probability of IC evolving indirectly goes down as the complexity of the IC system goes up . . . or something like that. (I wish I had DBB here! I've misplaced it, though I think it's at my workplace. Damn!)
regards,
[ 09. June 2003, 03:36: Message edited by: Jack Foster ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 09. June 2003 02:33
Jack: It could not have generated the phrase "Methinks it is a weasel", nor could it have generated a mousetrap. It was oriented towards a specific solution space.
If it is Turing complete then it could have generated in principle the phrase. It was NOT geared towards any specific solution space, when realizing that the solution space includes all Turing complete problems.
Let's look at proteins for instance. In principle sequence space can generate all possible proteins, regulatory networks etc. It 'merely' has to find them. As I have argued protein space is characterized by few very common proteins and many very uncommon where the common proteins in sequence space can be found all over and are in close position to eachother. Most solutions are found in common structures. Combined with scale free network features this seems to be strong evidence for evolutionary origins.
Jack: But once again, Behe never said that it was impossible for IC to evolve indirectly;
Did the Avida example evolve EQU directly or indirectly? And if indirect routes may be the prefered evolutionary pathway then Behe's comments do not seem to have much relevance.
Behe clearly dismissed indirect routes
quote:
Behe made the point on page 40 of Darwin's Black Box:
Even if a system is irreducibly complex (and thus cannot have been produced directly), however, one can not definitively rule out the possibility of an indirect, circuitous route. As the complexity of an interacting system increases, though, the likelihood of such an indirect route drops precipitously. And as the number of unexplained, irreducibly complex biological systems increases, our confidence that Darwin's criterion of failure has been met skyrockets toward the maximum that science allows.
Seems that Behe's argument may need some clarifications.
SOme additionally uses of the term IC
quote:
The bacterial flagellum is an example of what Michael Behe describes as an irreducibly complex system. In his book, Darwin's Black Box, he explains that such irreducibly complex systems could not have arisen by a gradual step-by-step Darwinian process.
also see Here
IC seems to suffer from similar confusions as other so many other terms in ID.
quote:
An irreducibly complex system cannot be produced directly by numerous, successive, slight modifications of a precursor system, because any precursor to an irreducibly complex system that is missing a part is by definition nonfunctional.[/url]
Source
Also [quote]Demonstration that a system is irreducibly complex is not a proof that there is absolutely no gradual route to its production. Although an irreducibly complex system can't be produced directly, one can't definitively rule out the possibility of an indirect, circuitous route. However, as the complexity of an interacting system increases, the likelihood of such an indirect route drops precipitously. And as the number of unexplained, irreducibly complex biological systems increases, our confidence that Darwin's criterion of failure has been met skyrockets toward the maximum that science allows.
While Behe was quick to dismiss indirect routes it seems evolution may be using such routes in quite a remarkable manner.
But in the end I would like to remark once again that IC is thus NOT a reliable indicator of design. If Behe wants to show that a particular system is IC and could NOT have evolved then he must either eliminate all natural pathways, known and yet to be considered (not a very attractive approach) or provide for positive evidence of an intelligent design.
Perhaps we can thus see quotes like this one be more carefully phrased.
quote:
For Dembski, this "Irreducible Complexity 2.0" is Darwin's invariant. I know that people from the audience I spoke with appreciated this definition for it seemed that Dembski took to heart many of the criticisms leveled at ID concepts, but still managed to improve those concepts. Darwin's unpayable debt was thus the fact that natural selection is incapable of producing irreducibly complex structures, as defined above.
Similarly [quote]Evolution simply cannot produce complex structures in a single generation as would be required for the formation of irreducibly complex systems. To imagine that a chance set of mutations would produce all 200 proteins required for cilia function in a single generation stretches the imagination beyond the breaking point. And yet, producing one or a few of these proteins at a time, in standard Darwinian fashion, would convey no survival advantage because those few proteins would have no function-indeed, they would constitute a waste of energy for the cell to even produce. Darwin recognized this as a potent threat to his theory of evolution-the issue that could completely disprove his idea. So the question must be raised: Has Darwin's theory of evolution "absolutely broken down?" According to Michael Behe, the answer is a resounding "yes."
quote]
Here
Maybe ID is overselling its claims?
[ 09. June 2003, 02:38: Message edited by: Pim van Meurs ]
IP: Logged
|
|
GP
Member
Member # 570
|
posted 09. June 2003 10:25
Jack,
Earlier you wrote me:
quote:
"Evolvability" as fourth condition for evoltuion is circular, so it really must be viewed as a placeholder. It seems that there is a disagreement between RBH and W&A about what condition or set of conditions produce "evolvability".
I just need a clarification here. Before you suggested that evolvability is a set of sufficient conditions. Now you seem to imply it is a set of necessary conditions -- whatever produces "evolvability." Which is it?
Continuing to your post this morning,
Turing-completeness is a red-herring. Biological evolution isn't "Turing-complete". And why do you assume that all that my system can evolve is a mousetrap?
As I understand where the mousetrap analogy was going, if you truly consider Turing-completeness to be a red-herring then your analogy ought to be seen as a strawman. The point seemed to be that the Avida simulation was "front-loaded" with instructions conducive to building EQU. But if Avida is demonstrably a universal Turing machine, then the set of computable algorithms that can be implemented is large -- so large that any specificity claims about the instruction set is rather unbelievable. Consider, for instance, what Micah said: quote:
An evolving system could be designed with a set of building blocks and an environment conducive to building mousetraps.
One half of this statement seems not applicable to the Avida experiment -- the "building blocks" half. Similarly, consider your version of the Avida simulation:
quote:
Let's invent a series of operatives that might prove beneficial in building a mousetrap. One operative might be "build spring", and one might be "build base". Another might be "combine parts x and y".
I think RBH was right to bring up Turing completeness given these two quotes. There was definitely a suggestion that the building blocks were specfic for EQU. But, I'd like to see a demonstration how your instruction set can implement as large a set of machines as Avida can.
I agree that it evolved IC, but it did so by lowering the probabilistic hurdle associated with this particular form of IC. My system will evolve IC as well, and in the same way.
This argument is substanceless without experimental support. The Darwinians could just as well argue that any IC structure evolved because the natural environment lowered the probabilistic hurdle towards its evolution. At the heart of the argument is a simple substitution of probability for incredulity.
postscript: It occurred to me that Avida is also a good base for verifying Dembski's displacement thesis. Suppose we give a generous estimate to the amount of CSI smuggled into Avida and take into account Avida's being a universal turing machine. The implication seems to be that there is an imposed limit on the amount of CSI that the evolved programs would have (by Dembski's law of conservation). Does anyone want to speculate how this claim would hold under scrutiny?
[ 09. June 2003, 10:35: Message edited by: GP ]
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 09. June 2003 10:57
I think there are some confusions in the discussions here. Am I correct to describe the following?
Claim from evolutionary theory:
-> All ICs are evolved -- (1)
Claim from ID theory:
-> All ICs are intelligently designed -- (2)
-> Some ICs are evolved ---- (3)
(or equivalently Some ICs are not evolved)
Claim from EQU experiment:
-> Some IC (e.g. EQU) are evolved -- (4)
or
-> Something (e.g. EQU) are evolved -- (5)
(depending on your evaluation of EQU)
If the above are the claims, they do not appear to have significant conflict. The main issue is the "sufficiency" claim.
Affirming (4) or (5) cannot conclude (1).
Affirming (4) or (5) cannot refute (3).
A test case for (2) may be the flagella.
A test case for (1) using Avida may be the "TESTPRIME" function, without the use of a previous algorithm proof. The issue is an explanation of ALL "complexity of life".
I am actually quite open to the result from the test cases. Indeed, I think in theory, Avida can generate the "TESTPRIME" function, given infinite CPU time and infinite memory. The only requirement is that it can be done within the life time of the programmer.
[ 09. June 2003, 11:33: Message edited by: YZ2 ]
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 09. June 2003 12:10
After reading GP's post, and after some more thought, I agree that the flexibility of the Lenski system (Turing complete) is indeed a strong point, and is not a "red herring". I apologize for the quick characterization.
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 09. June 2003 13:39
Hi GP:
quote:
I just need a clarification here. Before you suggested that evolvability is a set of sufficient conditions. Now you seem to imply it is a set of necessary conditions -- whatever produces "evolvability." Which is it?
Necessary and sufficient. Evolution occurs if and only if the system has evolvability. W&A:
quote:
Hence, the Darwinian solution of optimization problems is possible if and only if the problem is "coded" in a way that makes the mutation-recombination-selection procedure an effective one. The "representation problem" is how to code a problem such that random variation and selection can lead to a solution. The representation problem underlies the issue of whether selection, mutation, and/or recombination can produce adaptation.
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 09. June 2003 14:27
Hi Pim:
quote:
Did the Avida example evolve EQU directly or indirectly? And if indirect routes may be the prefered evolutionary pathway then Behe's comments do not seem to have much relevance.
Behe clearly dismissed indirect routes
(Of course when you say "evolve EQU" you really mean "evolve programs that perform the logic operation EQU", right? ![[Wink]](wink.gif) )
The IC claim is a bit difficult to evaluate, because with the Avida example, we're really talking about an analogy for IC. No new physical property has arisen due to this Avida example. The computer registers existed prior to the run, for instance. What "set of well-matched, mutually interacting, nonarbitrarily individuated parts" came into being as a result of this system?
You'll perhaps notice, however, that I've never actually made the argument that what has evolved here is not IC, though I spent a good deal of time pondering the question. To me, if this is not an appropriate IC analog, then nothing in Alife can ever be considered IC.
As to whether EQU evolved directly or indirectly . . . it depends upon how you look at it. If you look at the IC function as "to perform the logic operation EQU", then the evolution is indirect. I think this is the best way to view the example. (If you look at the function as "to perform logic operations", then the evolution might be viewed as direct, but now the successful strains are no longer IC; the complexity is reducible while preserving some functionality.)
You say Behe "dismisses" indirect routes. To me, it seems that he at least considers them, and recognizes that there is no absolute impediment:
quote:
Even if a system is irreducibly complex (and thus cannot have been produced directly), however, one can not definitively rule out the possibility of an indirect, circuitous route. As the complexity of an interacting system increases, though, the likelihood of such an indirect route drops precipitously.
So probability of attainment of IC via indirect route was always something to be considered. As I've said before, the Lenski result did not surprise me. Intuitively, the evolution of EQU did not seem probabilistically out of reach.
I've got a recording session. regards,
[ 09. June 2003, 20:56: Message edited by: Jack Foster ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 09. June 2003 15:10
Jack wrote quote:
The IC claim is a bit difficult to evaluate, because with the Avida example, we're really talking about an analogy for IC. No new physical property has arisen due to this Avida example. The computer registers existed prior to the run, for instance. What "set of well-matched, mutually interacting, nonarbitrarily individuated parts" came into being as a result of this system?
I really think this posting raises questions that are irrelevant. "No new physical property has arisen due to this Avida example."? Are you adopting Micah's argument that IC is a property only of 3-D physical objects? If so, the pattern of magnetic fields in the computer memory that stores the program that evolved is a "new physical property." Those electromagnetic patches are as physically real as the spring of a mousetrap.
What "set of well-matched, mutually interacting, nonarbitrarily individuated parts" come into being in Behe's mousetrap or the bacterial flagellum? All those proteins weren't specially created just for the flagellum, were they? They are all over the place in biology, it seems. The "set" that came into being in the Avida runs is the ordered sequence of instructions. They are well-matched, mutually interact, are nonarbitrarily individuated - it's easy to tell them apart, and that assembly language instructions are discrete entities is real old news. And the ordered arrangement is critical to the function of the program: a program is not a random aggregation of instructions. What more does one need?
Are you abandoning all the linguistic analogies IDists are so fond of? Will we hear no more about abstract information floating around independent of physical instantiation? (Please, please say 'yes'! ![[Smile]](smile.gif) )
The programs that evolved in the Avida simulation are not mere analogies. Those programs are real entities that meet the operational criterion for irreducible complexity on both Behe's original definition and on Dembski's modified definition. They are no less real entities than a strand of DNA. The quibbling about "parts" is amusing, particularly given all the linguistic examples we've seen from IDists. Every "complexity" estimate of a linguistic object that I've seen from IDists defaults to letters as parts, just as Dembski's informal analysis of the flagellum defaults to proteins as parts.
Jack further wrote quote:
To me, if this is not an appropriate IC analog, then nothing in Alife can ever be considered IC.
I'd take off the "A" in that remark. If this is not IC, then nothing in life can be considered to be IC.
RBH
[ 09. June 2003, 15:20: Message edited by: RBH ]
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 09. June 2003 15:18
For those who are interested in the testprime function, a web site explaining it is in:
http://crypto.cs.mcgill.ca/~stiglic/PRIMES_P_FAQ.html
I think it is quite a contribution if an evolutionary program can be written to generate this function, even if the relevant information is "smuggled" in.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 09. June 2003 21:31
RBH writes:
quote:
Those programs are real entities that meet the operational criterion for irreducible complexity on both Behe's original definition and on Dembski's modified definition.
Well, no they don't meet Behe's original definition. As has been pointed out before.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 09. June 2003 22:10
Roger wrote quote:
Well, no they don't meet Behe's original definition. As has been pointed out before.
The original operational definition was the knockout procedure, period. The evolved programs meet that one. Depending on who is doing the evaluation, they either do or do not meet the fuzzy criterion having to do with 'well-matched.' But since 'well-matched' has no operational definition such that neutral observers can reliably sort instances into the two classes 'well-matched' and 'not well-matched,' it's of no utility in doing research.
RBH
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 09. June 2003 22:19
quote:
A single system composed of several well-matched, interacting parts that contribute to the basic function of the system, wherein the removal of any one of the parts causes the system to effectively cease functioning.
If you don't have A single system composed of several well-matched, interacting parts that contribute to the basic function of the system, then knockouts can't indicate IC. You've thrown away a good part of Behe's pretty short definition, and yet claim to satisfy it. Sorry, that won't play.
Whether you think it is of utility in doing research or not does not change the fact that it was a critical part of Behe's original definition that the EQU program fails to satisfy.
[ 09. June 2003, 22:23: Message edited by: Roger R ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 09. June 2003 23:08
Roger:
quote:
Whether you think it is of utility in doing research or not does not change the fact that it was a critical part of Behe's original definition that the EQU program fails to satisfy.
Does it? One may argue that the components/parts are those than when knocked out destroy the function.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 09. June 2003 23:27
Y2Z:
quote:
I think it is quite a contribution if an evolutionary program can be written to generate this function, even if the relevant information is "smuggled" in.
In fact the arisal of prime numbers in nature has been observed in for instance predator prey interactions.
For instance
"A Biological Generator of Prime Numbers" by Eric Goles et al in _2000 Nonlinear Phenomena in Complex Systems_
Here
quote:
In the present work, we shall allow for the merging of two seemingly unrelated subjects, one
being periodical insects and the other the theory of prime numbers. The fact that some species
of cicadas (genus Magicicada) appear every 7, 13 or 17 years and that these periods are prime
numbers has been regarded as a coincidence [1, 2]. Without intending to argue in favour or against this statement, we will show here that a simple evolutionary predator-prey model yields primeperiodic preys. Moreover, this result will be used as a number-theoretical tool, namely to generate large prime numbers. Furthermore, we will demonstrate how a spatio-temporal extension of the model renders spiral waves being reminiscent of those observed in excitable systems, host-parasitoid systems and prebiotic evolution.
Also notice the following interesting works
Fogel in _1966_
Here did experiments with prime numbers
quote:
We can see in figure 1.16 that experiment 15 ended up predicting the prime numbers quite well towards the end, and we can see in figure 1.17 that it ended up with very few states.
This is easily understood when one notices that the higher we get into the environment string the less frequent prime numbers become.
and
quote:
Analysis of the results showed that the machines quickly “learned” to recognize numbers divisible by 2 and 3 as not prime, and some hints towards an increased tendency to predict multiples of 5’s as not prime.
Also in the Handbook of evolutionary computing
this experiment is described.
Not perfect but still not too bad either.
[ 09. June 2003, 23:31: Message edited by: Pim van Meurs ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 09. June 2003 23:45
Roger,
It's been asserted, but not specifically shown, that the evolved programs do not meet one or another aspect of Behe's or Dembski's definitions. You wrote quote:
If you don't have A single system composed of several well-matched, interacting parts that contribute to the basic function of the system, then knockouts can't indicate IC. You've thrown away a good part of Behe's pretty short definition, and yet claim to satisfy it. Sorry, that won't play.
Well, let's see, shall we? How does the evolved program I posted earlier in this thread fail to meet Behe's definition? Here are the main components of Behe's definition:
1. single system. Do you argue that the program is not a "single system"?
2. several parts. There are 60 instructions in the Case Study program. Do you argue that there are not "several" parts in that program?
3. well-matched parts. Fuzzy, but do you argue that the several instructions aren't "well-matched" in the face of the fact that if they are arranged in a particular configuration, they perform logic operations? Do you argue that they are not "well-matched" even though the primitive operations that the individual parts perform mesh together to accomplish a complex logic function?
4. interacting parts. Do you imagine that the various instructions don't interact, with the result of one instruction being an input to another?
5. parts contribute to the basic function. This is the focus of the operational definition. Recall that it was Behe who offered that operational definition, not me. Do you argue that the various instructions don't "contribute to the basic function" in the face of the fact that if we knock out any one of 35 instructions in the Case Study program it can no longer perform the logic operation EQU?
In what specific respect does that program fail to meet Behe's definition? Which of the five components of the definition is not satisfied by that program, and specifically how is it not satisfied?
**************************
A brief note on YZ2's posting. One part reads: quote:
-> All ICs are intelligently designed -- (2)
-> Some ICs are evolved ---- (3)
(or equivalently Some ICs are not evolved)
If the compound statement [(2) AND (3)] is the ID claim, then this effectively removes ICness as an indicator of design, and given that Dembski has claimed irreducible complexity as a special case of specified complexity, also weakens SC as an indicator of design. If all ICs are designed, and some are evolved, then the notion of irreducible complexity is vacuous from the point of view of detecting design. And that's the Dembski/Behe ID goal, remember?
I am still trying to figure out why the TESTPRIME function is relevant to any question of interest in evolution in either biology or ALife research. That is, I can think of no light that would be shed on any biological or evolutionary question of interest whether an evolutionary algorithm did or did evolve a solution to the TESTPRIME function. (However, cryptographers and electronic transaction security people would be highly interested, I imagine!)
RBH
P.S. to Jack: I understand now what you meant. Sorry for my obtuseness. I'll circle back to that when I have time.
[ 09. June 2003, 23:52: Message edited by: RBH ]
IP: Logged
|
|
|
Printer-friendly view of this topic
