|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
Pim van Meurs
Member
Member # 541
|
posted 04. July 2003 14:46
Eric T Malroy asserts:
quote:
Given that the nucleotides are possibly coupled, the genes are coupled and the systems within the life-form are coupled, Darwinists are hard-pressed to explain the evolution of complexity in biological systems.
And now some helpful facts
Evolution of complexity requires:
- variation
- selection
Evolutionary Biology and Biocomplexity
We study fundamental properties of the evolutionary process, using theoretical and computational methods. Evolutionary theory has a claim of universality, in the sense that the theory does not make any reference to its instantiation, that is, how information is encoded. We therefore often use populations of self-replicating computer programs (also known as digital life) to perform simple evolutionary experiments. We believe that evolutionary theory can be treated just like any theory in physics, where theories inspire experiments, who in turn can be designed to validate of falsify theories. Below is a list of our publications in this are, with links to the Los Alamos archive or local PDF files.
ev: Evolution of Biological Information
What did Dembski have to say about Ev?
Well first of all "Schneider claims to have generated complex specified information for free. The No Free Lunch theorems, however, tell us this is not possible."
But the NFL theorems do not say this nor do the NFL theorems even apply to evolutionary algorithms. Lets listen to Wolpert, one of the authors of the No Free Lunch Theorems
quote:
Indeed, throughout there is a marked elision of the formal details of the biological processes under consideration. Perhaps the most glaring example of this is that neo-Darwinian evolution of ecosystems does not involve a set of genomes all searching the same, fixed fitness function, the situation considered by the NFL theorems. Rather it is a co-evolutionary process. Roughly speaking, as each genome changes from one generation to the next, it modifies the surfaces that the other genomes are searching. And recent results indicate that NFL results do not hold in co-evolution.
Lets move on then
"All such fine-tuning amounts to investigator interference smuggling in complex specified information."
So indeed Dembski seems to agree that CSI was generated but suggests (erroneously) that CSI was 'smuggled in'. Schneider addresses many of Dembski's complaints in the link provided above.
So rather than claim that 'Darwinists are hard-pressed to explain the evolution of complexity in biological systems.", it seems more accurate to state that "intelligent design theorists have been unable to show that complexity in biological systems cannot arise naturally"
[ 05. July 2003, 01:51: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 04. July 2003 18:17
Pim, you are dangerously close to getting banned again. When it comes to assertion, you are king.
Malroy brings up an interesting question (the coupling of variables) and rather than address it, you pull out your trusty ten line "put the stupid IDers in their place" handbook and cite Schneider. Well, if you take notice, Strachan has taken Schneider to task, and it is far from clear whether Schneider has shown much of anything. And please save us the moaning and whining about why Strachan is so obviously wrong. He's got a Ph.D. in the field and is an expert in both genetic algorithms and neural networks. He deserves to be taken seriously.
You are having the same exact psychological effect on me as Frances did. Be careful. We banned him and we won't hesitate to ban you too. We don't need people around here who view themselves as heroic firefighters. Either engage in meaningful discussion or disengage and leave our community.
[ 04. July 2003, 18:19: Message edited by: Moderator ]
IP: Logged
|
|
Rompecabezas
Member
Member # 814
|
posted 06. July 2003 20:38
I'm fascinated by the way the ostensible subject of this thread has been ignored. It seems obvious to me that IDC has to answer Nature's objections rather than debate the coding details of simulation programs.
By assuming that 'everything designed is evidence of intelligence,' IDC has been able to point to biological complexity and make the not altogether convincing parallel to mousetraps, outboard motors, and other human artifacts. Isn't the bacterial flagellum itself a challenge for IDC? Doesn't IDC have to model the possible/plausible design history of such biological complexity that its proponents claim that modern science cannot?
I'd be interested to hear how IDC explains the origin of the systems that are inaccessible to material mechanisms. I'd like to hear how IDC science will differ from reductionist science. What can we expect from IDC science that is missing from methodological naturalism?
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 07. July 2003 12:49
Rompecabezas,
Your posts are going to need to be more substantial and less polemical to continue at this forum.
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 09. July 2003 09:17
It is a misunderstanding to say that ID cannot infer the formation history of a biological object. There are differences:
1) An ID evolutionist infers the formation history based on a) the observed data, b) a design model of the object, c) all possible formation mechanisms (which may include RM&NS), whereas a Darwinian based on the observed data and RM&NS etc. Because the former can construct more possible hypotheses within the design model, it can result in an explanation with a higher level of specificity.
2) Furthermore, since the formation history is an inference rather than a part of the model for ID, the formation history can remain uncertain if the evidence does not convincingly justify such explanation. As comparison for Darwinian, the formation history is part of the model and requires a prior commitment whether there is evidence for its justification or not. Therefore the modeling process for Darwinian is more constrained.
[ 09. July 2003, 13:54: Message edited by: YZ2 ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 17. July 2003 02:43
For those still interested in the original topic of this thread (actually, this tangled ball of twine!), I've begun to do familiarization and callibration runs with both versions 1.3.0 and 1.6.0 of Avida, though 1.6.0 is having some problems on the Beowulf cluster - apparently there's some kind of buggy interaction with the particular implementation of Linux on the cluster. I got a workaround from the guys at CalTech and will try it soon.
Anyway, as part of getting familiar with the program and with editing the various configuration files, I've run a couple of very long evolutionary runs in 1.3.0 (the Windows version) with a flat fitness function. In contrast to Lenski, et al.'s graded fitness function, in which more complex (in the sense of requiring more nand operations) logic operations earned more reproductive resources (SIPs), I set up the config file to award equal SIPs for performing any of the eight logic functions, up to and including XOR and EQU which require a minimum of 5 nands.
In two very long runs of 500,000 cycles, within 320,000 cycles one of the runs evolved lineages capable of performing all eight of the logic functions, including the two that require a minimum of five nands, XOR and EQU. The other run (actually still in progress at Update #450,000; a 500,000 cycle run takes over 24 hours on a dedicated 2.4 GHz machine) has produced lineages that perform all the 1-, 2-, and 3-nand operations, along with NOR, a 4-nand operation, but has not (yet) produced lineages that perform the 5-nand operations. (Recall that the Lenski, et al., runs were 100,000 cycles maximum.)
The two runs differed in several respects, including initial critter size, initial randomization, and a couple of other variables, so I won't speculate on why one run produced lineages capable of performing all eight and the other hasn't yet done so. But a definite conclusion is that a graded fitness function is not necessary to evolve lineages capable of performing complex behaviors, at least not in the context of this simulation. This simulation is not merely 'flogging the population up Mt. Improbable' to suit the preconceptions of the experimenter.
Watching the runs, it is clear that there's lots of variability for evolution to work with. The run that evolved lineages to perform all eight had an average critter size of around 350 instructions when all eight finally appeared, of which an average of 180 instructions were actually executed, meaning that there were around 170 'junk' instructions, potential raw material for adding functions. (The 'junk' instructions, of course, can vary randomly as they mutate with no selective filtering, yielding different genotypes that had the same functional phenotype.) In the last stages of the run there were around 120 different "species" each composed of a number of different genotypes. In any case, even after 500,000 updates there was lots of variability in the soup for selection to work on.
In calibration runs with the same graded fitness function as Lenski, et al., the 5-nand operations appeared much sooner in the runs in which they evolved. The competitive 'strategy' for lineages evolving with a flat fitness function is different from that with the Lenski, et al. graded fitness function. With a graded fitness function, a lineage can prosper by replacing a simpler function with a more complex function, often coopting and thereby losing the simpler function, since the more complex function earns more SIPs. With a flat fitness function, though, a lineage can out-compete its neighbors only by adding a new function. A replacement yields zero change in fitness. In fact, with a flat extrinsic fitness function there is some slight implicit selective pressure against more complex functions. More complex functions require more instructions to perform, and the more instructions, the larger the target for mutations that are mostly disruptive. Hence to merely replace a less complex function with a more complex function is selectively negative to some slight degree. If a lineage coopts some existing function to perform a more complex function, losing the coopted function in the process, it is actually at a slight disadvantage relative to its ancestor (and its ancestor's unmutated descendants nearby). In order to compete successfully against its neighbors given a flat fitness function, a lineage has to add new adaptive functions, not replace less complex functions with more complex ones. Thus with a flat fitness function, in the competition in the Avida world there is selective advantage for cumulative complexity: lineages evolve to be able to do more and more things, some of them complex, rather than just evolving to do more complex things.
Since there is some slight selective pressure against longer instruction substrings, I'll be interested to see if there are lineages that evolve to perform the various functions more efficiently over evolutionary cycles, doing the same stuff with fewer instructions. Unfortunately, version 1.3.0 doesn't have all the data files dumped to disk that 1.6.0 does, so one can't trace the lineages to see just what was happening in them.
RBH
[ 17. July 2003, 03:05: Message edited by: RBH ]
IP: Logged
|
|
|
Printer-friendly view of this topic