|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
Micah Sparacio
Member
Member # 6
|
posted 06. June 2003 19:19
Jack, well stated. From the very beginning, the >50% success rate stuck out to me like a sore thumb.
And you are right, Behe acknowledges the possibility of the evolution of IC systems. His is not an "in principle" argument, but an "as the world turns" argument.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 06. June 2003 21:39
Great! We all agree then.
1. IC can evolve,
2. generally it evolves through circuitous routes, and
3. it can evolve easily enough as long as intermediates can exist that provide a selective advantage to the organism, even if this implies a change of fucntion between the intermediates and the final IC product.
This is exactly what evolutionary theory would predict, what the existing evidence suggests, and what evolutionary biologists have been saying all along, since DBB first came out. It is also what the AVIDA simulation demonstrates, to the satisfaction of many, including now, apparently, several ID advocates.
So what exactly makes IC a problem for evolutionary theory? Why should one infer (or even prefer) ID over naturalistic evolution when confronted with an IC system?
Quite frankly, I think we can all go home now. ![[Wink]](wink.gif)
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 22:19
Charlie may want to go home, but I've been writing this and I'm doggoned well going to post it! ![[Smile]](smile.gif)
Micah wrote quote:
Jack, well stated. From the very beginning, the >50% success rate stuck out to me like a sore thumb.
This is becoming amusing.
I'd sure like to hear a clear, reasonably detailed description of Jack's notion of "frontloading." Not a sentence or two, but detail. Is there more than this?
Platform
1. Start with a uniform population of replicators, none of which could do anything except replicate. The replicators were small (50 instructions) assembly language programs, and each run started with a population of 3,600 identical replicators.
2. Primitives in the form of instructions are available via mutations. Appropriately arranged, programs composed of those primitives can perform any computable function. (Turing-complete, remember.) The instructions are roughly equivalent to "genes," in the sense that each instruction performs a non-trivial operation. The sequencing of instructions in a program composed of such instructions is critical for it to perform any larger function: integrated sequences of instructions are necessary, not random aggregations.
3. Two evolutionary operators, mutations (3 kinds) and selection, operate on the replicators as they go about replicating.
4. There is limited 'space' in the Avida world, just room for 3,600 replicators. Hence replication efficiency is advantageous.
Experiment
Given this system, an experiment was designed to see whether programs that could perform a complicated logic operation could evolve. To provide comparative data and test relevant hypotheses, three conditions (fitness functions) were designed.
a. Experimental Condition: Members of a small set (NOT, NAND, AND, OR-N, AND-N, NOR, XOR) of all possible less complicated logic functions were differentially selectively advantageous, gaining a replication advantage if they were performed. The complicated logic function of interest, EQU, earned the most reward. This condition had 50 evolutionary runs.
b. Control 1: No less complicated logic function was selectively advantageous. Again, 50 evolutionary runs were made in this condition.
c. Control 2: In 36 repetitions of 10 evolutionary runs each, one or a pair of the set of less complicated logic operations was removed from the fitness function. All simple operations singly and all pairs of simple operations were tested for indispensability by being removed from the fitness function.
Performing the simpler logic operations earned 'energy' for the 'organisms' - more time in which to replicate - so performing those functions was selectively advantageous to the evolving programs because the population was limited to 3,600 programs and no more space for offspring was available: there was competitive pressure to replicate more efficiently. Appropriate apportioning of 'energy' according to genome length ensured that genome length was selectively neutral.
The question was whether under these circumstances, programs that performed a more complicated logic function, EQU, would emerge, given that it also was selectively advantageous.
Results
The answer, as we all know, is yes, programs that performed EQU did evolve. Moreover, by the operational definition of irreducible complexity, all those programs were irreducibly complex; knockout testing showed each had an irreducible core of instructions such that replacing any of those instructions with a null instruction eliminated the program's ability to perform EQU.
There were some subsidiary findings of interest. Among them was the finding that the evolution of a lineage (as evolutionary biologists have known for some time) is not an uninterrupted upward climb of the slopes of Mt. Improbable. At any given time a population is characterized by variability in the relative fitness of its members, and those of slightly (or even greatly) lower relative fitness may become the ancestors of a lineage that ultimately climbs higher on that mountain while their more (relatively) fit brothers and sisters are stranded on a local peak.
Another finding of interest was that no one simple logic operation and no pair of the simpler logic operations was critical for the evolution of a program that could perform EQU. The 360 control runs yielded 124 programs that performed EQU, with at least one such program evolving in every 10-run control condition. It also appears to be the case that at least some simpler programs are necessary. In 50 runs in which only EQU was selectively advantageous, a program that performed it never appeared, at least not in the 100,000 cycles for which the testing was run.
Another finding was that each program that evolved to perform EQU was different. The same functional outcome was reached by very different programs produced by very different evolutionary lineages.
Interpretations and redefinitions
A variety of arguments have been offered in several threads to attempt to minimize and even trivialize the implications of the study. Initially there were a number of attempts to 'refine' the definition of irreducible complexity, including some that redefined it as Behe's later evolutionary definition did. Those efforts are on-going: Micah has been hinting that IC applies to 3-D mechanical structures but may not be applicable to logical structures.
As I read him, John is mostly arguing that the simulation is so far removed from biology that it can tell us nothing about biological evolution. Further, he has argued that because the simulation has a 'ground' level higher than something analogous to base pairs in DNA, there is something fishy about drawing inferences about biological evolution from it. In another vein, John has suggested that since generating a program that performs EQU is 'just' shuffling commands around, and since there's a relatively high mutation rate in the simulation, maybe those programs (147 of them!) that perform EQU emerged sort of by chance. John has been reticent about pinning down how much chance is necessary to account for the results.
Micah made the (somewhat bizarre) suggestion that what would be convincing is showing that EQU (or some comparable IC structure) could evolve under selectively neutral ("unbiased") conditions. While he called that consistent with "Darwinian principles," of course it isn't.
Most recently, Jack Foster and Micah have seized upon the relatively high rate of occurrence of evolved programs that can perform EQU as being somehow suspicious, perhaps diagnostic of some sort of unspecified frontloading.
And that's where I end this posting:
If primitive building blocks are available, whether they are primitive assembly language instructions or are genes constructed by biochemistry;
and
if simple structures composed of those primitive building blocks are selectively advantageous, so they can emerge by random mutation, are preserved by selection, and are therefore subsequently available in an evolving population of multiple genotypes;
and
if a more complicated structure is also selectively advantageous;
then if that more complicated structure emerges in several different lineages through (among other things) the cooption and elaboration of simpler adaptive structures, does that tell us something potentially useful about biological evolution?
I think it does, particularly when (contrary to John's assertions) there are examples - analogs, if you will - in biology of all those things as they are instantiated in the Avida simulation. It suggests that the allegedly low probability of indirect pathways may well not be as low as imagined (hoped?) by IDists. It is amusing that the finding that the probability turns out not to be low seems to be interpreted (implicitly, at this point) by Jack and Micah to indicate that there's something fishy about the simulation. I want to hear explicitly what that is.
RBH
[ 06. June 2003, 22:33: Message edited by: RBH ]
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 07. June 2003 06:55
charlie d. writes:
quote:
it can evolve easily enough as long as intermediates can exist that provide a selective advantage to the organism, even if this implies a change of fucntion between the intermediates and the final IC product.
Allow me to change the "can" to "do". And then we are left with the distance between functional intermediates in terms of the blind search of random mutations. I think Dembski and Behe could both live with something close to that.
But where they may disagree is that we are at the point where we have strong evidence that this is the case. The Avida program demonstrates that functional intermediates (amongst other issues) make a significant difference in how "easily" a more complex structure (i.e. bigger) can evolve. It doesn't demonstrate that such intermediates exist in the biological world.
And there is nothing in the Avida experiment that has changed my mind at all. If you think that IDers were surprised by the results, I think you might need to consider that it was your view of ID that was deficient. The position of you and RBH that we are somehow in denial or defensive about the results, is just not true. It wasn't really very surprising.
[ 07. June 2003, 07:06: Message edited by: Roger R ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 07. June 2003 09:20
I was wondering from RBH and others who feel strongly that the Lenski system destroys the concept of IC, once and for all (the IC-bomb so to speak)...
Have you heard any good critiques of the system so far? If not, what do you think the critical aspect of the system is? If there were one or two things that we should all focus on, and try to settle on, what would they be?
IP: Logged
|
|
Argon
Member
Member # 276
|
posted 07. June 2003 11:49
Roger R writes:
quote:
The Avida program demonstrates that functional intermediates (amongst other issues) make a significant difference in how "easily" a more complex structure (i.e. bigger) can evolve.
Certainly. However, underlying this was the question of whether such complex structures actually could evolve and whether the tests (ICness, CSI, & etc.) reliably indicated design.
quote:
It doesn't demonstrate that such intermediates exist in the biological world.
Correct. But this leads to some difficulty: How does one demonstrate that intermediates steps exist or don't exist in any particular system? The advantage of the Avida simulation is that every step can be recorded and the conditions controlled at a level of resolution that is not nearly as easy to replicate in living systems. However, I'll note again that Lenski and others are running somewhat similar experiments on microbes in continuous culture. Lenski's group has been culturing several strains continuously for more than a decade, freezing samples of the cultures and testing how fitness variations arise and propagate through the populations. Perhaps within another decade (and more than a few grad students and post-docs) they'll be able to tease apart some interesting progressions. It will be tough going, trying to search for needles in haystack as they comb through genomes and correlate changes with variations in fitness, but I think the tools they will need to speed that work are starting to become available.
With regard to analogies between Avida and biology, I think the polyketide synthases and non-ribosomal peptide synthetases are biological systems that bear some similarity to the Avida simulation (referenced in Joshua Smart's thread). I think the streptomycin-resistance/resistance-cosuppressor double mutations characterized in lab experiments (referenced early in this thread) demonstrate a ratchet-like mechanism that can give rise to IC systems. I happen to like the strep-resistance example because it represents one of the simplest, most primitive case of how an IC relationship can arise and how shifting environmental requirements play a role.
[Aside: In the past, I've asked others to help list examples of the simplest, most-recently emerged biological IC systems of which they are aware. Because if one really wants to test the general question of whether IC systems can evolve as opposed to whether a particular IC system can evolve, then those newer, smaller systems are where most biochemists would look.]
Now, this is not to say that there aren't some IC systems that couldn't evolve, only that IC as a tell-tail marker of unevolvability has some doubt. This has no impact of ID itself, but simply consigns one hypothesis (ICness as a reliable hallmark of design) to the scrap heap. This doesn't prevent other ideas from being developed and tested. For instance, as noted by Jack Foster, CSI and IC are not co-joined. In fact the willingness to test ideas, discard hypotheses and develop new ones is a positive sign that a research discipline is not "out of steam" or played out.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 07. June 2003 11:56
Micah: Jack, well stated. From the very beginning, the >50% success rate stuck out to me like a sore thumb.
Why? What is so surprising that evolution can have a success rate of >50%. Surely that there is a succesrate of >50% in finding a solution to EQU which appears to be irreducibly complex is quite interesting.
Is EQU CSI? Well that depends on which one of the various definitions of complexity one uses? Does one calculate the probability based on a uniform probability function as suggested by the various examples in NFL? Richard Wein askes these very same questions and I do not believe that he received any relevant answers. But lets look at the flagellum example, surely Dembski's calculations of its probability could be helpful here? The tornado in the junkyard calculations suggest that the flagellum is CSI, what would the same calculations show for EQU? Certainly the example suggests that (complex) specified information can be generated. Is it a matter of time until the generation of information reaches a 150 bit threshold?
Or should we calculate the probabilities, which is after all what Dembski's measure of complexity is, based on the likelihood of EQU under the assumption of RMNS? In that case the numbers are much higher but then we can conclude similarly that ID cannot generate CSI either since the probability would be close to 1?
If anything, Lenski et al's paper has revealed that the concepts of IC, and CSI are in dire need of careful definition.
Micah: And you are right, Behe acknowledges the possibility of the evolution of IC systems. His is not an "in principle" argument, but an "as the world turns" argument.
Behe's argument was that IC systems could not evolve and although he does touch briefly on what he calls indirect routes, he considers them to be too unlikely to be taken seriously. Certainly Dembski argues that CSI is a reliable indicator of ID and also argues that IC is a 'special case of' CSI. "Ultimately what enables irreducible complexity to signal design is that it is a special case of specified complexity."
Source
May we conclude thus that CSI may also not be a reliable indicator of ID? And although Behe rejected indirect routes as unlikely, Lenski's experiments indicate that indirect routes may be what defines evolutionary pathways.
Some Behe quotes
quote:
Examples of irreducible complexity can be found on virtually every page of a biochemistry textbook. But if these things cannot be explained by Darwinian evolution, how has the scientific community regarded these phenomena of the past forty years?
Source
So what have we learned so far?
IC systems by most definitions seem to be able to evolve through RM&NS. IC may not be a reliable indicator of intelligent design.
Thus looking at a system like the flagellum and stating that it is irreducibly complex does not help us establish if it is intelligently designed, nor do "tornado in the junkyard" calculations for the flagellum seem to help us resolve these issues.
I would like to hear the responses from Behe and Dembski on these recent findings and to hear what they believe is the impact of these experiments on the concepts of IC and CSI.
The ISCID forums seem to be an excellent place for such.
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 07. June 2003 12:32
Hi RBH, all:
quote:
This is becoming amusing.
I'd sure like to hear a clear, reasonably detailed description of Jack's notion of "frontloading." Not a sentence or two, but detail. Is there more than this?
(and then later.)
It is amusing that the finding that the probability turns out not to be low seems to be interpreted (implicitly, at this point) by Jack and Micah to indicate that there's something fishy about the simulation. I want to hear explicitly what that is.
I'm not quite sure what's so amusing here. The point is that the evolvabity displayed in this system did not come about by accident. "Frontloading" isn't a dirty word; are you arguing that the Lenski organisms came about via abiogenesis? Lenski et al set out to evolve EQU. They designed an evolutionary system that they felt was up to the task.
I don't think there was anything "fishy" about the simulation. (Well, . . . except that this is not a simulation of anything of course. The organisms in this experiment are their own form.) I think it's a perfectly valid and valuable experiment. The success rate at evolving EQU is simply consistent with my frontloading view of this. If the frontloaders want to improve their success rate, there are things they can do:
*) increase the number of trial generations.
*) smooth the fitness terrain by increasing the number of intermediate rewards.
*) determine how the unsuccessful lineages are being trapped by the landscape, and tweak the 26 operators to prevent the trap or to allow for escape.
What might be a bit "fishy" is the tendency to overextrapolate from this. We could use the same evolutionary logic commonly employed to come to the conclusion that AVIDA and computers evolved via Darwinian mechanism! The authors themselves fuel that tendancy in the paper by quoting a Dennet putative truism:
quote:
As Daniel Dennett has emphasized, "evolution will occur whenever and wherever three conditions are met: replication, variation (mutation), and differential fitness (competition)".
But at the risk of mixing apples with oranges, I'll point out that you write:
quote:
If primitive building blocks are available, whether they are primitive assembly language instructions or are genes constructed by biochemistry;
and
if simple structures composed of those primitive building blocks are selectively advantageous, so they can emerge by random mutation, are preserved by selection, and are therefore subsequently available in an evolving population of multiple genotypes;
and
if a more complicated structure is also selectively advantageous;
then if that more complicated structure emerges in several different lineages through (among other things) the cooption and elaboration of simpler adaptive structures, does that tell us something potentially useful about biological evolution?
Well the number of conditions seem to be on the rise.
IDists should indeed take the Lenski paper as a lesson. IC remains an important focus for ID because of the probabilistic hurdle that it represents. In the Lenski case, the hurdle is not so high that it cannot be easily negotiated. IDists need to therefore place more energy formalizing the notion "degree of IC".
Another focus should be evolvability. The evolvability displayed in this system did not arise by chance. I don't think most biologists really understand evolvability, and I think that it is a fundamental key to ultimately understanding biological evolution.
regards,
P.S.
Earlier you wrote:
quote:
"EQU" didn't evolve; programs that perform the logic operation EQU evolved. That's an important distinction.
Not to me. When I write "EQU evolves", feel free to read "programs that perform the logic operation EQU evolve".
[ 23. June 2003, 21:16: Message edited by: Jack Foster ]
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 07. June 2003 13:52
Here's an example of what I mean by frontloading.
Let's take another familiar example of IC: the mousetrap. 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". Provide replication, and perform a Lenski-style AVIDA run; give the system a try. If you aren't getting close, then further constrain the phase space via tweaking of the operators or guide the evolution by providing intermediate rewards within the fitness function.
I have no doubt that some evolutionary system can be designed that can evolve a mousetrap. So does this mean that IC is a useless concept and that Intelligent Design is a sterile intellectual pursuit?
Well not to me. I think Darwinism and ID are in fundamental philosophical conflict. But I think it's possible to scientifically view their common subject matter without wearing philosophical lenses. Darwin was at least partially correct. That doesn't mean that IDists may not be partially correct as well.
[ 07. June 2003, 13:53: Message edited by: Jack Foster ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 07. June 2003 13:53
quote:
IDists should indeed take the Lenski paper as a lesson. IC remains an important focus for ID because of the probabilistic hurdle that it represents. In the Lenski case, the hurdle is not so high that it cannot be easily negotiated. IDists need to therefore place more energy formalizing the notion "degree of IC".
That'd be fine, but then an objective way has to be devised to establish that degree. For instance Behe, who first came up with that definition, also never bothered applying it to anything. So, what's the degree of ICness of the flagellum? Or of any of the EQU programs from the AVIDA simulation? I sense another definitional can of worms is about to be opened.
As for the relative simplicity of EQU and its ease of evolution, everything has to be kept in context. These programs evolved in a few tens of thosands generations in a population of 3,600 "individuals". What kind of complexity would be considered to be out of reach of 10^40 bacteria evolving for 10^10 generations? Without some serious quantitative analysis, I'm not sure who has the advantage here, the "programs" or the bacteria. quote:
Another focus should be evolvability. The evolvability displayed in this system did not arise by chance. I don't think most biologists really understand evolvability, and I think that it is a fundamental key to ultimately understanding biological evolution.
That's irrelevant. Biological systems are endowed with evolvability, so are the AVIDA programs. How it got there is not what is at issue here, but what it can do once it's there.
Similarly, your claim of front-loading is "weak" (in a technical sense). Either you are saying that everything that evolves was by definition evolvable by the initial conditions of the system, which is obvious but again irrelevant, or you are trying to say that the reaching of the "target" was in some way encoded in the initial conditions, in which case you have to show where and how.
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 07. June 2003 13:59
Charlie, I snuck in another reply at about the same time as yours that you might easily miss. (Just an FYI.)
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 07. June 2003 14:03
Roger, quoting charlie d, wrote quote:
charlie d. writes:
quote:
it can evolve easily enough as long as intermediates can exist that provide a selective advantage to the organism, even if this implies a change of fucntion between the intermediates and the final IC product.
Allow me to change the "can" to "do". And then we are left with the distance between functional intermediates in terms of the blind search of random mutations. I think Dembski and Behe could both live with something close to that.
But where they may disagree is that we are at the point where we have strong evidence that this is the case. The Avida program demonstrates that functional intermediates (amongst other issues) make a significant difference in how "easily" a more complex structure (i.e. bigger) can evolve. It doesn't demonstrate that such intermediates exist in the biological world.
Let us not forget that those functional intermediates, the subprograms that were coopted in the evolution of programs that performed EQU, themselves evolved. Each evolutionary run started with a population that could only replicate. It evolved those intermediate building blocks. They too were generated by the evolutionary process; they were available because they evolved.
Roger also remarks on his (and other IDists') requirement that we must "demonstrate that such intermediates exist in the biological world." Actually, that's a substantial red herring: those intermediates may well no longer exist. In the Case Study lineage that evolved to perform EQU, for example, the emergence of EQU coincided with the disappearance of the ability to perform one intermediate function. If one looks through the program that first performed EQU in that lineage, that intermediate is gone. The particular code arrangement that performed it no longer existed in the genome. If we didn't have the complete history of that lineage we would never know it had been there! On the relevant (Smart) Brainstorms thread I recently wrote quote:
Note, by the way, that once a program that performs EQU has evolved, the simpler logic operations can drop out of the competitive race - now be unrewarded - and the main function will persist in the population so long as it continues to be selectively advantageous. Way downstream, since they no longer confer selective advantage, those simpler functions may no longer be performed and thus won't be visible in the current population. We'd see no "precursors" of EQU in the extant population and we'd wonder where the heck our programs with the ability to perform EQU came from. And we could generate an "irreducibly complex programs that perform EQU can't evolve" conjecture, and we could challenge critics of our IC program conjecture to produce the exact pathway and tell us what those hypothetical precursors are. (And if they can't specify the precise pathway and the exact precursors, we could say "Neener neener neener!" )
When I wrote that I hadn't looked to see if it had happened in the Avida evolutionary runs. It did, in spite of the fitness function not changing. A precursor disappeared from the code, apparently as a consequence of the emergence of the new functional capability. The emergence of EQU wiped out one of its precursors. (Interestingly, that precursor reappeared later; since it still conferred some selective advantage, it re-evolved and then persisted in the lineage. Similarly, another "precursor" was lost two mutational steps before EQU emerged, and then re-appeared a few thousand cycles later, after EQU had emerged.)
Jack wrote quote:
I'm not quite sure what's so amusing here. The point is that the evolvabity displayed in this system did not come about by accident. "Frontloading" isn't a dirty word; are you arguing that the Lenski organisms came about via abiogenesis? Lenski et al set out to evolve EQU. They designed an evolutionary system that they felt was up to the task.
I note that Jack has not described just what was frontloaded. In fact Jack's description is a complete mischaracterization of the experimental enterprise. The experimenters set out to ascertain whether such programs could evolve and to find out what some of the conditions governing that process are. That is, they asked an experimental question. Speaking as a long-time experimentalist, I can tell you unequivocally that there is little verging on no interest at all in doing (say nothing of the lack of interest of editor of a prestigious journal in publishing) an experiment of which one knows the result before doing it. That's a waste of time, effort, money, and computer cycles. It's not worth doing. Once again, as he did some months ago, I think Jack massively underestimates the amount of time and effort that is necessary to run these kinds of studies. One doesn't whip them out in a weekend. If Jack would think about the level of effort his CD has required he'd be more in the ballpark.
Jack complains that "Well the number of conditions seem to be on the rise." In fact, that was a summary of the conditions necessary for evolving "complex" structures, not for evolution in general. And it is a summary of the Darwinian account of the evolution of complex structures, to boot. Those conditions long predate the Lenski, et al., paper.
Jack touched on an important notion when he said "Another focus should be evolvability. The evolvability displayed in this system did not arise by chance." He's absolutely correct. Evolvability is key. So one asks "What is required for evolvability to characterize a system?" Why, it's the Dennett "truism" he noted earlier: "... evolution will occur whenever and wherever three conditions are met: replication, variation (mutation), and differential fitness (competition)". I'd add one adjective: heritable variation. If those three conditions characterize a population, it will evolve.
It's quite true that the evolvability - the ability to evolve at all - that is displayed by the Avida simulation was built in by the simulation's authors. That's what one does when one creates an experimental model - a computer simulation, in this case - of a system in order to study it. (Someone - I don't recall who - complained about this in the early stages of this discussion!) One builds a model to represent the variables and structures that theory says are important in the processes the theory purports to explain. That's the function of models - to represent theories so we can manipulate the model as perhaps we can't the real world, test predictions of the theory in the model environment, draw tentative conclusions from it for theory, and take the implications and predictions of the model back out into the world to see if they match.
As we learn more, both about the model and about the real world, the mapping and the model are refined, modified so as to make the model a better and better representation of the essential variables operating in the world. That's the fundamental nature of the research enterprise, in fact. If you look at the recent history of simulation models for biological processes, evolution in particular, you'll see exactly that process occurring. Avida is the latest and most veridical (and useful) of the sequence of models of evolutionary processes, and is a superset of Tierra, the first major representative of this sort of simulation model, which in turn partly grew out of the classifier systems and GAs of John Holland starting in the mid-1970s. It will be far from the last. The scientific enterprise in any discipline is a continuous 5-way 'conversation' among theory, data, models, hypotheses, and research technologies.
Those remarks speak to John Bracht's complaint (and Roger's above) that the Avida model doesn't veridically map (some properties of) real biological systems. Avida doesn't map all the properties, relevant and irrelevant, of biological systems, and it isn't intended to. It is built to represent the core properties of evolving systems that are identified by evolutionary theory as relevant to the question at issue, not merely some specific biological system. The task to answer Bracht's complaint is to show that the relevant properties of biological systems are mapped by the model, and elsewhere yersinia and charlie, among others, have spoken to that question, as have I. Others will no doubt do so, too. Lenski, it has been noted, also works with organisms in a microbiology laboratory.
RBH
P.S. The reason I stress the evolution of programs that perform EQU as distinguished from the locution "EQU evolved" is that the latter masks the wide variety of programs that evolved in the experiment. That's analogous, e.g., to the wide variety of flagellar structures that have evolved to perform the motility function (among other functions) in various bacteria. That kind of variability is a hallmark of the operation of evolution. It's known as "degeneracy," and is ubiquitous in biology. If it doesn't make a difference to Jack, he's ignoring a critical feature of evolved systems.
[ 07. June 2003, 14:30: Message edited by: RBH ]
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 07. June 2003 14:57
Hi RBH:
quote:
Speaking as a long-time experimentalist, I can tell you unequivocally that there is little verging on no interest at all in doing (say nothing of the lack of interest of editor of a prestigious journal in publishing) an experiment of which one knows the result before doing it.
So are you saying that you're surprised by this result? I know that I'm not. I truly can't imagine that this experiment is particularly valuable or compelling for those within your field. I think Nature's interest in this paper is at least partially provided by the anti-ID propaganda value that can be had by it. The inclusion of Robert Pennock as co-author lends some credence to this view.
quote:
Jack touched on an important notion when he said "Another focus should be evolvability. The evolvability displayed in this system did not arise by chance." He's absolutely correct. Evolvability is key. So one asks "What is required for evolvability to characterize a system?" Why, it's the Dennett "truism" he noted earlier: "... evolution will occur whenever and wherever three conditions are met: replication, variation (mutation), and differential fitness (competition)". I'd add one adjective: heritable variation. If those three conditions characterize a population, it will evolve.
This is absolutely wrong, unless of course, you reduce the meaning of the word "evolve" enough to trivialize it. Evolution simply must have somewhere to go. There needs to be adequate data structures, which are typically provided in systems (that have evolvability) by their genotype to phenotype maps. Here from Wagner and Altenberg from their landmark paper "Complex Adaptations and the Evolution of Evolvability":
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.
For biology the "representation problem" has some unsettling implications. If, as evolutionary biology asserts, all adaptations are the result of mutation and selection, organisms have to be evolvable. But once one calls into question the inevitability of organisms being evolvable, one can ask, how and why did an evolvable genome originate in the first place? Is it a fortuitous consequence of physics, or of biochemistry, or a "frozen accident" from life's origin? Are the genetic representations of the phenotype a product of evolution? What, if any, are the evolutionary forces that shape the genotype-phenotype map?
regards,
P.S.
quote:
If it doesn't make a difference to Jack, he's ignoring a critical feature of evolved systems.
It isn't that it "doesn't make a difference" to me. It's that the distinction wasn't important to the points that I was making.
[ 07. June 2003, 15:13: Message edited by: Jack Foster ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 07. June 2003 15:19
Jack: So are you saying that you're surprised by this result? I know that I'm not. I truly can't imagine that this experiment is particularly valuable or compelling for those within your field. I think Nature's interest in this paper is at least partially provided by the anti-ID propaganda value that can be had by it. The inclusion of Robert Pennock as co-author lends some credence to this view.
I disagree, I think that the experiment and the publication has done what ID proponents so far have failed to do AFAIK, namely defining experiments to discover the limitations and possibilities of RMNS as it pertains to concepts such as IC and CSI. In this regard the findings that RM&NS can "evolve" information which seems to be IC is nothing new (See Adami or Tom Schneider for instance) but the concept of ICness is surely new.
Is this meant as anti-ID or pro-science? One may speculate the former based on the involved authors but why not focus on the scientific value of this paper. I bet'ya this one will become a classic.
Jack: 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.
In fact research suggests that evolvability as well as robustness are concepts that can evolve themselves. Quite fascinating. What interests me is the network structure of DNA/RNA. DNA and RNA can be characterized by few very common components and many very uncommon components. These common components seem to be distributed well throughout sequence space and in fact seem to be very close to each-other. It may be this inherent structure of protein/RNA space which makes RMNS such an efficient algorithm.
IP: Logged
|
|
GP
Member
Member # 570
|
posted 07. June 2003 15:27
In my opinion, it is a major concession from ID proponents to invoke "front-loading" as a cause in explaining evolutionary products -- one might just as well throw up their hands and plead ignorance. As I have explored on several occasions with the members of this forum, "front-loading" is a theoretically bankrupt concept, without any methodology for exploring its properties. The only proper criticism to "front-loading" is "So what?"
I apologize for this short post, but my frustration at reading these posts has momentarily surpassed my curiosity. Perhaps I will return tomorrow with a better mindset.
[ 07. June 2003, 15:32: Message edited by: GP ]
IP: Logged
|
|
|
Printer-friendly view of this topic
