|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
yersinia
Member
Member # 324
|
posted 14. May 2003 20:43
Regarding this,
quote:
Now, if we return to the flagellum, the alternative functions proposed are rather limited - protein secretion, virulence, adhesion. Yet removal of any ol' flagellar component, such that motility is lost, does not necessarily mean these subfunctions will remain. For example, you can remove one of the motor proteins (such as motA) and still retain something that could fulfill these functions. But if you remove flhA, all of these functions are likewise lost. Recall that there are about 20 structural components that make up the bacterial flagellum. In fact, this probably holds true for most of the flagellar components. In other words, the appeal to alternative functions and cooption only get us so far with regard to the flagellum.
Josh wrote:
quote:
--------------------------------------------------------------------------------
Examination of the paper shows that they attributed higher degrees of fitness to organisms that could perform more complex logic operations, with the "reward" being 2^n where n was the number of logic operations combined. The EQU function required 5 operations, so was rewarded with 32 points; but
intermediate rewards of 2,4,8,& 16 were also allowed for simple functions.
--------------------------------------------------------------------------------
And it's at this point that the appeal to protein secretion, virulence, and adhesion all fail. A glimpse of this can be appreciated from figure 1 in my fifth essay on the flagellum and IC (Here). Remember, the flagellum is not some abstract thing, but something concrete built from the products of fllhA, flhB, fliR, fliQ, fliP, fliF, fliG, fliN, fliE, etc.
But this (and your graphic) ignore several salient points of the biology that don't really have analogs in the mousetrap.
E.g.:
1) a number of the parts of the flagellum might all be modified copies of each other. You are implying that they all originated independently from outside the system
2) both secretion and adhesion systems can have varying levels of complexity and sophistication. You are assuming that each represents at best one stage, e.g. X components at the secretion stage and Y components at the adhesion pili stage.
3) it is not at all clear that the very first crudely function proto-flagellum would have had to have all of the complexity of the current flagellum. First, if the function is dispersal rather than taxis, you don't need the taxis-related components. Second, if the bacterium is spherical (which is, BTW, optimal for dispersal), then positioning the flagellum on the cell is irrelevant, as is having a universal-joint. A proto-flagellum system could logically get by with maybe 1 proto-rod and 1 proto-filament protein sticking out, not a bunch of rod proteins, plus a hook, plus two filament proteins. Might not be optimal, but in a world of flagella-less organisms the one with the crudely-functioning proto-flagellum is king.
So, assuming that flagellar function would only emerge once all 20 current parts arrived is also invalid.
yersinia
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 14. May 2003 20:58
Okay, one more brief reply and then I want to work on my flagellum essay, while I have the time.
Charlie thinks I am arguing that the flagellum has more components than EQU. That's not the point. The point is that the cooption scenarios invoking protein secretion, virulence, and adhesion break down and never truly explain the origin of the IC flagellum. In other words, the replies to Micah fail once we bring the biological data into the picture.
As for Nic thinking I am being inconsistent, I am simply trying to demonstrate that appeals to alternative functions may work well in a philosophical sense, but ultimately, we're trying to explain history. Behe himself makes it clear that the IC argument must center around physical precursors and not conceptual precursors. If we bring the biological data in into the picture, yes, it's safer to assume the existence of mice than paper.
Remember that I explicitly stated, "Of course, I have a different view of IC, as I don't view it as a general principle, but as a very helpful tool in a forensic analysis."
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 22:42
Mike G wrote quote:
Charlie thinks I am arguing that the flagellum has more components than EQU. That's not the point. The point is that the cooption scenarios invoking protein secretion, virulence, and adhesion break down and never truly explain the origin of the IC flagellum. In other words, the replies to Micah fail once we bring the biological data into the picture.
It is genuinely difficult to get across to people who have never worked with them how sneaky, powerful, and subtle evolutionary algorithms can be. To be blunt, they are much cleverer than their programmers in the domain in which they operate. If that domain is sufficiently high-dimensioned, say 3 or more dimensions, it's darned near impossible to stop the critters from finding a way to some adaptive behavior or another on a complicated multi-dimensional fitness landscape. This does not constitute evidence for whether the flagellum evolved, of course, but it is to say don't underestimate the evolutionary processes. A good deal of the skepticism about the ability of evolution to produce complex stuctures and processes lies in an extreme under-estimate of just how clever evolution is. Niiicholas has posted some remarks about that from Ofria, the second author of the Lenski, et al, paper, here. Read them and chuckle, but it's true in my experience as well: GAs and EAs are sneaky critters!
RBH
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 14. May 2003 23:19
quote:
Charlie thinks I am arguing that the flagellum has more components than EQU. That's not the point. The point is that the cooption scenarios invoking protein secretion, virulence, and adhesion break down and never truly explain the origin of the IC flagellum. In other words, the replies to Micah fail once we bring the biological data into the picture.
Oh, please Mike. You can argue about and desassemble all the components of EQU because you have all the pieces in their historical order, courtesy of the AVIDA simulation. If you just looked at one of the final programs, outside their historical context, you wouldn't know how to make heads from tail about how the pieces got together. In the flagellum, we don't have the benefit of hindsight, and yet we identify a bunch of subcomplexes (just at the level of protein complexes, I am not even going into individual proteins and their domains) that can arguably be, and in some cases actually are found to be, flagellum-independent from a functional standpoint.
Unfortunately, it is not me who is ignoring biological evidence, but those who want to draw rigid, imaginary lines in their description of a biological system (starting from its very function) to fit some arbitrary definition.
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 15. May 2003 00:22
Clarifications:
RBH, My argument/position is not rooted in any skepticism about the ability of evolution to produce complex stuctures and processes nor does it entail an extreme under-estimate of just how clever evolution is.
Charlie, I'm not drawing rigid imaginary lines to fit some arbitrary definition. As for the biological evidence, we'll get to that in time....
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 15. May 2003 06:37
Arguably the most interesting aspect of this topic is not the computer simulations generated, but the interpretation of the results. Even more interesting is the issue of the standards used to validate the interpretation.
To paraphrase, the simulations are being interpreted as supporting the claim "Process Xm in mathematical universe Ym creates or generates Zm from non-Zm which suggests process Xr in the real world Yr creates or generates Zr from non-Zr". In English this argument might be ‘Process ‘mutation- selection’ in a mathematical universe ‘unnamed’ creates or generates complexity from non-complexity suggesting that process ‘mutation-selection’ in the real world creates or generates complexity or non-complexity’.
As has been pointed out, this ‘claim’ has been offered 1)without providing definitions or complexity or non-complexity, 2) without explicitly showing relationships between the mathematical universe used in the demonstration and real world biological systems, and 3)without explicitly demonstrating the change process used in the mathematical universe has a direct counterpart or logical equivalent in real world.
Current scientific institutions have at least partially validated the claim of creating complexity by ‘passing pre-publication peer review’ and by not being rejected by the broader scientific community. As should be obvious, the claim of creating complexity would not be validated if validation required precise definitions of the terms and concepts of scientific claims. In other words, if you created a scientific institution to compete with the current academic institution, and if the alternative science imposed or re-imposed the traditional ‘precise definition’ standard on scientific claims, then the claims offered in this study would rejected as ambiguous and unsupported.
The question of whether mathematical systems can create complexity is an interesting one. As RBH points out, certain types of mathematical systems certainly appear capable of producing something which looks like complexity. However, you will not be able to resolve the issue until offer precise definitions of complexity and the capacity of mathematical systems to create complexity. I would go a step further, and suggest, the issue of biological systems creating complexity will never be resolved until you create a scientific institution which uses the precise definitions standard in evaluating and validating scientific claims.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 15. May 2003 07:08
RBH writes,
quote:
The notion that "informational gaps" in the configuration space of all objects is part of the definition of "irreducible complexity" for a given object (say the mousetrap, or a program that performs EQU) is brand new to ID, as far as I know. This criterion departs in a significant way from the definition of Behe and its later refinement by Dembski, in that it requires additional knowledge beyond the structure of the object being analyzed. In fact, it requires exhaustive knowledge of the configuration space within which the object resides.
If I understand the terms correctly, that is a rather stunning statement, since that concept is at the heart of what I always understood IC to be, from the time I first read Behe's book. Behe even decided to include the implications of that concept as part of a more quantative definition of IC in a later version:
quote:
An irreducibly complex evolutionary pathway is one that contains one or more
unselected steps (that is, one or more necessary-but-unselected mutations). The degree of irreducible complexity is the number of unselected steps in the pathway. (Behe 2001)
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 15. May 2003 07:37
RR:
even by Behe's second definition, certainly some, and possibly all, final EQU programs are IC. Some even had temporary decreases in fitness that turned out to be required to finally achieve EQU. This is all discussed in detail in the paper.
Note however that nowhere in that definition is there a concept of "informational gap", let alone any fixed, arbitrary value.
[ 15. May 2003, 07:40: Message edited by: charlie d. ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 15. May 2003 08:38
RBH,
The hard part is, that while there may be many who laud the sneaky power of GA's, it happens to often be those who are super siked about the power of natural selection in general.
It is worth noting that in my AI class, the point was made over and over (both in text, in articles, and by professor) that the word is still out on GA's (this is a 2003 text) and that while there was early excitement about their potential, that excitement has been waning of late. The reason? Genetic algorithms are turning out to be not much better (or even worse) than other search algorithms. In fact, a few papers have been written on how "genetic" algorithms can be optimized if we take the "genetic" part out of the algorithm.
Anyway, I see you claiming time and time again that we just don't understand your sneaky little critters. But having read some of the more recent literature on genetic algorithms by people who are disinterested in the implications for real evolution, I'm finding that a different tune is being played.
That's why I'm skeptical. I guess its a sort of guilt by association, kind of thing. When I see GA's being analyzed by those impartial to the power of natural selection, I see skepticism. So, when those who are siked about the creative power of natural selection go gung ho over GA's... I just don't bite.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 15. May 2003 09:00
Micah:
can you please post the actual references?
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 15. May 2003 10:32
Off the top of my head, no, I can't think of most them (I'm on my way to Rhode Island for my brother's graduation). But here are a few until I get back home:
http://citeseer.nj.nec.com/mitchell93when.html
http://citeseer.nj.nec.com/baluja95removing.html
The text is Russell and Norvig - Artificial Intelligence a Modern Approach 116-119
"At present, it is not clear whether the appeal of genetic algorithms arises from their performance or from their aesthetically pleasing origins in the theory of evolution."
I could say more here but I'd be violating a trust agreement. It is sufficient to indicate that some authorities on GA's feel that they are given too much credit because of their ties to Darwinian evolution and that for solving non-toy problems, they are not the algorithm of choice.
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 15. May 2003 10:33
An analogous account inspired by the film Space Odyssey, 2001:
Combining existing functions occurs frequently in life. Usually nothing imaginative can be discovered from this process. That is fine. Life goes on. However at this moment at the height of a climb on the fitness landscape, something amazing happens. After mixing and combining a number of NAND functions, at the summit of the fitness landscape, in a way nobody quite sure of, coincides with that of the selection function of a --- you guess it, an EQU! Despite its humble beginning, this new object has all the hallmark of a designed object. It is irreducible. Whether it is complex or not, I do not know. EQU is clearly much more design than the other constructed functions when combined. AND it is generated using a Darwinian process! Where did this designed EQU come from? Where is its origin? How could it happen? Should we change all the indicators of design to define it not to be designed? Or define every other function as designed? As I said, this happens when the fitness function meets the selection function of the designed EQU. When their requirements are both met, a designed object is born through a Darwinian process. This encounter changes a lot of things. The selection function signifies a designed object. That’s what gives it intention and what you call intelligence. The encounter between the Darwinian process and the selection function from the intention gives the EQU sign of intelligence. Do we know the origin of the selection function, and perhaps intelligence as well? Just as in the film, its origin was not told.
I hope you like the story.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 15. May 2003 10:35
Micah,
You wrote quote:
Anyway, I see you claiming time and time again that we just don't understand your sneaky little critters. But having read some of the more recent literature on genetic algorithms by people who are disinterested in the implications for real evolution, I'm finding that a different tune is being played.
The key phrase there is "disinterested in the implications for real evolution." I'll circle back to an implication of that phrase at the end of this posting.
Your skepticism is in part justified. I've seen a fair number of applications of GAs in which they are obviously not magic. They are not universal problem-solvers and can be misused in applied search problems.
Not infrequently, the failed applications I've seen are due not so much to the GA itself, but to misapprehensions about the algorithm's properties and its grounding in biology. I'm reminded of a guy from an engineering background I knew some years ago who decided a GA was just the ticket for his design problem. He bought the Evolver add-on to Excel, set the problem up in a spreadsheet, and fired Evolver up. He got no useful results. Looking at his setup, it was pretty obvious why there was no good result. He hadn't thought through what the GA needed access to in order to make the discriminations he was asking of it. In essence, he had plopped the GA down in a stimulus-deprived environment, defined a short binary gene string (two alleles per gene) that was incapable of adequately representing the problem space, 'trained' it on a fairly limited subset of the conditions that the system could encounter, and then asked it to make discriminations for which there was no relevant information available in the selective environment and that it couldn't have represented even had it had access to sufficient information. Not surprisingly, the GA didn't solve the problem in any useful way. This is not an extreme or infrequent case: It happens a lot out there.
On the other hand, when I plop a GA down in a stimulus-rich environment, give it 'sensory' apparatus sufficient to allow it to potentially sense and differentially represent relevant variations in the environment (even when I don't know what's relevant, as I normally don't), give it a long enough chromosome (enough representational power) to have a fighting chance to represent the problem space with some fidelity (or allow variable length chromosomes), give it several kinds of variation generators (several kinds of mutations and usually recombination as well), use a large enough population, think through what 'fitness' appropriately means in the problem under attack, and don't press it for too-rapid convergence on a single solution, I find it does the kinds of things one gets psyched about.
Our GAs have distinguishable groups of genes - subsections of their chromosomes - that semi-independently control (they evolve epistasis, so "independent" is not fully applicable) their sensory apparatus and internal representation of the environment, their cognition (information retrieval and aggregation, and decision-making), and their behavioural control outputs. All the genes have multiple alleles, ranging from 10 up to a couple dozen. We're not doing a simulation in order to do biological research, but we have borrowed as much of the power of biological evolution as we can in order to produce populations that can survive and prosper in the complex and dynamic real world environments that we pipe into the system.
Users' backgrounds are not irrelevant. It took me several years, for example, to 'train' one of my partners who comes from an engineering background to really grasp the notions that the unit of analysis is the population, not an individual member of it; that we are not seeking the (singular) solution but rather populations of solutions; and that what we are doing is 'satisficing' (in Herbert Simon's sense) rather than 'optimizing.' In other words, many of the problems in the application of GAs to search problems occur not because of the power (or lack thereof) of the algorithm itself, but because of the ignorance and/or misconceptions of users. The "disinterest" in real evolution evinces itself as ignorance of the fundamental nature of the algorithm.
BTW, what text are you using in that course?
RBH
[ 15. May 2003, 10:46: Message edited by: RBH ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 15. May 2003 11:48
Roger quoting Behe's variation on a theme of IC
quote:
An irreducibly complex evolutionary pathway is one that contains one or more
unselected steps (that is, one or more necessary-but-unselected mutations). The degree of irreducible complexity is the number of unselected steps in the pathway. (Behe 2001)
What about a pathway that contains one or more steps which involve neutral or even detrimental mutations? The paper points out some interesting detrimental mutations which later become instrumental.
But Behe's definition is all but circular. IC is that which cannot evolve. If that is what is required to save IC from falsification then fine.
[ 15. May 2003, 11:51: Message edited by: Pim van Meurs ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 15. May 2003 13:38
I agree, Behe's 2000 definition is a version of IC#2 (IC, defined as that which can't evolve, can't evolve...duh) -- although the Lenski et al. simulation acheives IC even on this definition because of hitchhiking and suppressor mutations of initially detrimental mutations (in some, not all cases, IIRC), both of which do occur in real life.
IP: Logged
|
|
|
Printer-friendly view of this topic
