Computer Simulations and Darwinism

_{GilDodgen
October 1, 2011

Intelligent Design

3}_{Categories
Intelligent Design}

Share: Facebook; Twitter/X; LinkedIn; Flipboard; Print; Email

Okay dudes, no more talk about my abandonment of atheism. Here’s some science and engineering talk.

I know something about computer simulations. In fact, I know a lot about them, and their limitations.

Search algorithms (and especially AI-related search algorithms) are a specialty of mine, as is combinatorial mathematics.

The branching factor (the average number of moves per side) in chess yields approximately 10^120 possible outcomes, but the number of legally achievable positions is approximately 10^80 — the estimated number of elementary particles (protons and neutrons) in the entire known universe. Compare this to the branching factor of nucleotide sequences in the DNA molecule. Do the math.

Finite element analysis (FEA) of nonlinear, transient, dynamic systems, with the use of the most sophisticated, powerful computer program ever devised for such purposes (LS-DYNA, originally conceived at Lawrence Livermore National Laboratory in the mid-1970s for the development of variable-yield nuclear weapons) is another of my computer-simulation specialties.

Dyna has been used heavily in the automotive industry for simulating car crashes, so that cars can be designed to produce the least damage to occupants.

In these simulations everything is precisely known and empirically quantified (the material properties of the components — modulus of elasticity, mass density, shear modulus, precisely calibrated failure modes, etc.).

In addition, the explicit FEA time step (the minimal integration time step determined by the software based on the speed on sound in the smallest finite element and its mass density, which is required to avoid numerical instability) is critical. In my simulations the time step is approximately a ten-millionth of second, during which partial differential equations, based on the laws of physics (F=ma in particular) are solved to compute the physical distortion of the system and the propagation of the forces throughout the system in question.

One learns very quickly with FEA simulations that even with all of this knowledge and sophistication one must empirically justify the results of the simulation incrementally by comparing the results with the reality it attempts to simulate.

One false assumption about a material property or any of the other aspects of a simulation can completely invalidate it. Worse yet, it can produce results that seem reasonable, but are completely wrong.

So, the next time someone tries to convince you that a computer simulation has validated the creative power of the Darwinian mechanism of random errors filtered by natural selection in biology, you should tell them to go back to school and learn something about legitimate computer simulations, and how difficult it is to produce reliable results, even when the details are well known.

Comments

OK, so you accept that natural selection (i.e. heritable variation in reproductive success) results in adaptation, but you argue that the variance must be directed? OK. But we see from GAs that undirected variance (variations in the virtual genomes are randomly selected) also leads to adaptation, so it appears that no direction at the level of the mutation is required for adaptation to occur. But then you seem to argue that direction at the level of the environment must occur. Well, in a GA, clearly, the environment is usually designed to solve a human problem, and so the fitness function criteria reflect the degree to which the output from the virtual organisms matches the target output. But that's no different from a natural environment, in which the "target output" is simply something that will enhance reproduction, for, example, better camouflage, or more effective foraging. Those "targets" are just as "built in" to the environment as they are to a fitness function, it's just that they are naturally ocurring targets. Hence the term "natural" selection. And I'm defining "fitness" as "reproductive success within the current environment". By the way: bear in mind that in a GA, the fitness function merely specifies the output from the virtual organism - it doesn't specify how the organisms must produce that output. That's the whole point, and that's why we use them - to evolve solutions that will give us the output we want.Elizabeth Liddle_{October 7, 2011
October
10
Oct
7
07
2011
01:38 AM
1
01
38
AM
PST}

6.2.1.1.12 Dr Bot, That the space has many dimensions is not why EAs work well. EAs are just a kind of local (non-systematic) search. They are as good as any other local search algorithm in general (e.g. tabu search, guided local search). In particular instances where we do not know much about the problem domain, they may be good because we don't know better. There are problem classes where hybrid search is better (non-systematic and systematic), i.e. where we can exploit our knowledge about the problem structure and efficiently propagate our knowledge as search progresses. You see, there are various types of search. Typically, people who work with one type tend to advocate it. I have seen people arguing for swarm intelligence, EAs, Tabu, you name it. In general, a search algorithm (as anything else) suffers from dimensionality not benefits from it because of combinatorial explosion: the neighbourhood size grows exponentially with the number of dimensions.Eugene S_{October 7, 2011
October
10
Oct
7
07
2011
01:35 AM
1
01
35
AM
PST}

OK, thanks for the clarification. You say:
I have a theory that there is no indirect Darwinian pathway from point x to y. I write a computer program, *building in the fact that there is no possible indirect pathway from x to y.* I run the program, and sure enough, none of my fake “organisms” make it from x to y through an indirect pathway. Would you say that the program has demonstrated the veracity of my theory? Or would you say that the program is an exercise in circular reasoning?
What I'd say is that it is very difficult to prove a negative, but you could say that the probability of a direct pathway, given your parameters, is very low. And if you'd selected the parameters deliberately in advance so that they would render the probability low, then what you would have done is confirm this. That's not necessarily circular - if you had simply hypothesised that, given your parameters, Y is very unlikely, your model is a test of that hypothesis. That's the point of stochastic models of non-linear systems. Unlike the pre-computer days when you could prove something with "pen and paper", we often do not know, until we run a model, whether our model fits the data,and the answer may be a p value rather than a "yes or no" answer. So I think we are not quite on the same page, although establishing this may nonetheless be useful! I am certainly not of the opinion that because a computer model is designed by a programmer, the answer must be known in advance, and conclusions from its output therefore circular. Anyway, I look forward to your critique of AVIDA :)Elizabeth Liddle_{October 7, 2011
October
10
Oct
7
07
2011
01:29 AM
1
01
29
AM
PST}

Oh, it refers to a mechanism, all right. It's just misleading to call "it" an agent. What I'm saying is neither an "admission" nor contrary to any "Darwinist Claim". Heritable variance in reproductive success (aka "selection") is the mechanism by which genotypic variants that confer reproductive success in the current environment become more prevalent in the population, i.e. is the mechanism by which populations adapt to their environment. What is less obvious, at least to Darwin, was the mechanism by which traits were inherited (we know know that it's mostly genetic), and the mechanisms by which novel sequences are introduced into the genome (there are many, point mutations, deletions, insertions, and recombination being the most important). So lots of mechanisms are involved at the level of generating heritable variance. Evolution occurs when this heritable variance results in changes in allele frequency in a population, and adaptation occurs when the heritable variance is reflected in reproductive success in the current environment.Elizabeth Liddle_{October 7, 2011
October
10
Oct
7
07
2011
01:06 AM
1
01
06
AM
PST}

Fake but accurate.material.infantacy_{October 6, 2011
October
10
Oct
6
06
2011
04:15 PM
4
04
15
PM
PST}

The Weasel program was intended to demonstrate the power of cumulative selection vs. a purely random search. And it does exactly that. I'm not sure why certain people got so hung up about it, but I think that says more about them than it does about the Weasel program and what Dawkins intended it to demonstrate.NormO_{October 6, 2011
October
10
Oct
6
06
2011
04:05 PM
4
04
05
PM
PST}

Elizabeth,
Firstly, Dawkins didn’t “go wrong” with Weasel. It was simply never intended to be anything other than a toy to demonstrate a very simple point. It is not a simulation of evolution in any sense but that simple point.
Dawkins acknowledged the limitations of Weasel and how it differs from evolution. But then why use the illustration at all while admitting that it doesn't fit? All it did was get people even more hung up on this half-crazy idea that random variation and selection can come up with something as simple as a coherent 28-letter sentence. Of course, this is coming from a guy who thinks monkeys really do write Shakespeare. Getting them into shape takes a few million years, and then the rest is easy.ScottAndrews_{October 6, 2011
October
10
Oct
6
06
2011
03:54 PM
3
03
54
PM
PST}

Mine too I've noticed. Perhaps fingers have a bad sense of context, or are a little lazy.material.infantacy_{October 6, 2011
October
10
Oct
6
06
2011
03:51 PM
3
03
51
PM
PST}

oops, site=sight. Why do fingers type homophones?Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
03:42 PM
3
03
42
PM
PST}

Firstly, Dawkins didn't "go wrong" with Weasel. It was simply never intended to be anything other than a toy to demonstrate a very simple point. It is not a simulation of evolution in any sense but that simple point. Secondly, "EQU" is not "programmed into" AVIDA. The fitness function rewards functions that perform "EQU" but that is based on the output of the evolving functions, not whether or not the functions themselves match a preset solution - indeed there is no "preset solution". The functions that evolve are different on every run. And of course EQU won't evolve unless it is "rewarded"! That's the Whole Point! In order for evolution to occur there needs to be heritable variance in reproductive success. A simulation that did not build this in would not be a simulation of evolution. In the case of AVIDA, what determines reproductive success is whether the critter can perform certain functions, just as, in nature, whether a critter can perform certain foraging functions will determine whether it thrives long enough to breed. The populations were not "coaxed towards EQU every step of the way". In fact, one of the reasons EQU is slow to evolve is that in many cases, both neutral and deleterious mutations are important precursors. EQU functions, in other words, tend to be "irreducibly complex". Nonetheless, functions that perform EQU regularly evolve! The idea that in evolutionary algorithms "information" is "smuggled in" is a fundamentally mistaken one. No "information" is "smuggled in". It is there, in plain site, in the fitness function, just as it is there, in plain site, in the natural environment. Hot sun "smuggles in" the information that organisms need to find a solution to "keeping cool"; "large seeds" smuggle in the information that organisms need to find a solution to "opening beaks wider". "EQU rewards" smuggle in the information that AVIDA organisms need to find a solution to "performing EQU". In other words, there is no "smuggling" at all. The environment is what it is, and critters evolve to thrive within what it offers in terms of resources and hazards, whatever they may be. And what they may be determines the functions they evolve to tackle. Just as in AVIDA.Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
03:03 PM
3
03
03
PM
PST}

Thanks, Elizabeth. We're getting closer to the heart of the matter, so let's keep going for a bit. Elizabeth: "No, the solution was not “programmed in from the beginning”. What was programmed in from the beginning was the problem, which happens to have a single solution. Posing a problem with a single solution is not the same as “programming in the solution”." You are correct that a problem with a single solution is not the same as programming in the solution. However, Weasel did *not* simply pose a problem. It contained the information for the output solution *in the program itself* as a target string. The "generations" of letters were then compared against this specified information *which was already in the program,* and accepted only to the extent that the letters approached the goal. Yes, there was only one solution, but that is not the problem. The problem is how it arrived at that solution -- the fact that the solution was slipped in the back door and the program was written to converge on the solution. We could have substituted any other target string, from something simple like "Hi there" to something more complex, like the Gettysburg address, and the program would have inevitably convereged on the solution. Elizabeth: "Or, if you insist that in the case of an evolutionary algorithm that poses a problem with a single solution that it is legitimate to say that the solution is “programmed in . . ." No. Again, whether there is a single solution is not the issue. It doesn't matter whether there is a single "correct" answer or whether there is a range of potential answers, or some stochastic result whereby we expect a certain percentage of the program runs to arrive at a solution. The problem is whether the solution is carefully slipped in the back door and whether the program is geared to converge on a solution -- again, that convergence could be inevitable (like Weasel) or stochastic (more like Avida). The key is that if the program drives the process toward a solution, then by definition, the information about the solution was contained in the program itself. This has nothing to do with evolutionary mechanisms; this is purely a question of programming, logic and information theory. Now we all can see what Dawkins did wrong. Does a similar problem beset Avida? Setting aside for a moment all the other significant problems with Avida, let's focus on this particular issue. EQU is a solution that is contained in the program itself. It is a known target, just like Weasel. You seem to think this is not an issue because, well, perhaps the program wouldn't have necessarily converged on EQU, you might say. But that is simply a misunderstanding of the program. The fact that Avida is more complex than Weasel, includes more steps, and is stochastic, rather than a single solution, does not change the fact that the target is set in the program from the outset and that the program is geared to move the populations inevitably toward EQU. It takes a bit more finesse to see what is going on with Avida, but in terms of containing a target sequence that the population is *programmed to converge on* it is fundamentally no different than Weasel. Indeed, when the authors tried to get EQU in the 50 populations they ran without the incremental rewards (think Weasel with the target sequence getting closer), the populations couldn't get to EQU. Thus, even in a relatively simple situation, the populations couldn't come up with the solution, *unless they were coaxed along toward the pre-determined and pre-programmed solution.* This is the real takeaway of the Avida program, and is consistent with the monkey-typewriter experiments and with Behe's review of functional gain from mutations. Now, there are lots of other issues with Avida, which I point to in my paper and which others have discussed, but we need to be very clear about this key problem. The EQU target was contained in the program from the beginning; the pathway to get to EQU was contained in the program from the beginning; the rewards to get to EQU were set up from the beginning; the populations were carefully coaxed toward EQU nearly each step of the way. All of the information content relevant to EQU and how to get there was in the program from the outset. Don't let the fact that there are lots of "organisms" fool you. Don't let the fact that there are "populations" rather than a single "entity" fool you. Don't let the fact that we are dealing with statistical curves and percentages and stochastic outcomes, rather than a single solution, fool you. The solution was programmed in from the beginning and the populations were driven inevitably toward it.Eric Anderson_{October 6, 2011
October
10
Oct
6
06
2011
02:37 PM
2
02
37
PM
PST}

As a GA, the Weasel program would consist of, in parts, a randomization procedure (a mutator), an evaluation routine (score keeper), and a fitness function (Hamming oracle). The fitness function would determine a Hamming distance between the target string and the mutated progeny, and feed that score to the evaluation routine, which would score each progeny, only allowing the most successful to reproduce. In this regard, the target string in the fitness function would be pre-programmed, unequivocally. Any target could be substituted, and the program would consistently converge on the solution without fail, given appropriate tuning for generation size and mutation rate. The output of the program would demonstrate the conservation of information; the information content of the final output would be less than or equal to the information content of the initial input. We could modify the fitness function to score against two target strings, say "methinks it is like a weasel" and "abcdefghijklmnopqrstuvwxyzab" and we would see convergence on either of the strings (at random) but nothing more. Keep adding target strings and we would see convergence on any one of them, but nothing more. The information input into the system would determine its output. String length, generation size, mutation rate, and target space would all determine how and if a solution was found, but the output would never wander beyond the constraints of the input. Do what you will to the fitness function. The output won't exceed the input (technically by more than 500 bits).material.infantacy_{October 6, 2011
October
10
Oct
6
06
2011
02:27 PM
2
02
27
PM
PST}

One additional source of confusion regarding Weasel is that in most evolutionary algorithms, the fitness function - the criteria that determines how likely a critter is to reproduce - does, unlike Weasel, simply match the critter against a "target". Instead, the critter performs some function, for example taking input and producing output, and it is the output that is matched against the "target" criteria. How that output is produced is "up to" the critter, and when you get a population of critters that regularly manage to match the output, you can peek inside, and see how it does it. In other words, the evolved critters embody the solution to your problem - the fact that you "programmed in" the problem does not mean you "programmed in" the solution.Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
02:24 PM
2
02
24
PM
PST}

In the special case of the Weasel program, problem, solution, and fitness function are the same thing - a specific letter sequence. If you want to call that "programming in the solution" I suppose you could, but it would be a non-generalisable observation, a bit like inferring that because the sqrt of 1 is 1, taking the sqrt of a number doesn't change it.Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
01:59 PM
1
01
59
PM
PST}

EL, are you suggesting that the Weasel program did not contain the target string "methinks it is like a weasel" and that the program did not test each generation against this target string? How was the problem modeled to have a single solution that didn't include comparison with a predefined target?material.infantacy_{October 6, 2011
October
10
Oct
6
06
2011
01:43 PM
1
01
43
PM
PST}

Elizabeth, please respond to the central question. Elizabeth: “In weasel, the problem is trivial and there is only one optimal solution (the sequence of letters that spell “methinks it is like a weasel). So it does not demonstrate that Darwinian algorithms can find unknown solutions to non-trivial problems.” Yes, these are problems with the weasel program, but you are skirting the primary issue. The question was whether the solution was programmed in from the beginning. This is important to the circularity issue we have been discussing from the beginning.
No, the solution was not "programmed in from the beginning". What was programmed in from the beginning was the problem, which happens to have a single solution. Posing a problem with a single solution is not the same as "programming in the solution". Or, if you insist that in the case of an evolutionary algorithm that poses a problem with a single solution that it is legitimate to say that the solution is "programmed in", then please do not generalise from that completely atypical example to any other evolutionary algorithm. The Weasel is a toy. Useful evolutionary algorithms pose problems to which there may be multiple solutions, none of which are known in advance. So they are certainly not "programmed in".
Do you or do you not acknowledge that the solution was programmed in from the beginning?
Only in the above, extremely limited and not generalisable, sense.Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
01:19 PM
1
01
19
PM
PST}

“Name an unconnected island” Organs of vision, for example. There is no known path of successive modifications that connect no-vision to vision. The famous diagram is too scanty to be able to show that in satisfactory detail. Behe points out that the curvature of the retina recess alone requires a complex protein-protein interaction to be fine-tuned.
Yes, there are known paths of successive modification to organs of vision. Not only that, but all stages of the path are actually found in extant living things. But let's not get on to too much of a tangent - my point is that the irreducible complexity argument is quite different from the Information Can't Be Created By Computer Simulations argument. And also quite different from the "Reaching Islands of Complexity is Highly Improbable" argument. Complex designs evolve within Evolutionary Algorithsm, so we know the sytem can produce complex novel solutions to the problems posed by the environment. Nobody argues that Darwin's theory accounts for "Reaching Islands of Complexity", so probability arguments don't work either. The only remaining argument is: Some things couldn't have evolved by stepwise increments. Which is simply an arguments from lack of a model, not an actual model. And, in the case of vision, we actually have a good model. Where we don't as yet have a good model is getting to a DNA-based life-form. But that's no reason to infer "Design" (nor of course, reject it).Elizabeth Liddle_{October 6, 2011
October
10
Oct
6
06
2011
12:23 PM
12
12
23
PM
PST}

6.2.1.1.14 "Name an unconnected island" Organs of vision, for example. There is no known path of successive modifications that connect no-vision to vision. The famous diagram is too scanty to be able to show that in satisfactory detail. Behe points out that the curvature of the retina recess alone requires a complex protein-protein interaction to be fine-tuned.Eugene S_{October 6, 2011
October
10
Oct
6
06
2011
11:33 AM
11
11
33
AM
PST}

Dr Bot, I am not an expert on ID. An even better example to show what you intended to show is synthetic life. Here is what I think about this. Obviously, information gets generated. As to how much and whether it can be described as CSI, I don't know. IN fairness, even if it is CSI, what remains to be seen is: 1. how you got the initial setup for the system to generate CSI. This is by no means the same as the problem of origins, so I am not switching the problem. The problem we are talking about is enabling optimisations within a (sufficiently small) set of allowed configurations. 2. Quantitatively, how much (CSI) information have been generated: - i.e. whether it is beyond the quantitative CSI threshold (500 bits) 3. Qualitatively, it remains to be seen whether the explanatory filter can give out the "design" answer. Dembski does address the criticism of Fibonacci series observed in nature, in the same vein: yes, natural systems produce Fibonacci series "naturally" (and not only Fibonacci, but other things like the golden ratio of parts to the whole, on mass). But what does that exactly mean? To me it would mean that design is involved at some stage here anyway.Eugene S_{October 6, 2011
October
10
Oct
6
06
2011
11:07 AM
11
11
07
AM
PST}

I think we're getting off in the weeds a bit. There are two primary points. (1) The Avida paper failed to do what it was intended to do, namely refute the concept of irreducible complexity. Here is what happened with the famous Avida paper: Behe says, in effect: "Darwinian processes are unlikely to produce a functionally integrated system unless there is a step-by-step incremental pathway with regular fitness rewards along the way." Lenski et al. say: "No! Behe is wrong! If there is a step-by-step incremental pathway with regular fitness rewards along the way, Darwinian processes might produce such a system." Anyone see an issue with their response to Behe? [Note, this isn't an evolution test, just a simple logic test.] ----------- (2) *All* of the key questions about what would make a Darwinian process actually work in the real world were *assumed* in the Avida program, not demonstrated by it. Simple fitness landscape, easy to traverse from one point to the next, regular rewards along the way, increasing rewards the closer you get to the goal, short pathway to a complex function, and on and on. There is a real, genuine, scientific, observational question about whether these things are true in the real world of biology. Avida simply assumes them all. People are easily impressed because it is a "program" and gee, it seems to work kind of automatically once you finish writing all the details and press "start". But we have to remember, anything a program can do, we can do with paper and pen. And what would that pen-and-paper description look like? Well, if we assume all the key points in question, then our theory will work. Sure. Of course. But that is *not a demonstration of the efficacy of our theory,* it is simply a restatement of it, in computer format, rather than with pen on paper. Avida is entirely circular. Indeed, the primary lesson of the Avida progam was that even with a simple function and an easy landscape, the program couldn't get to the complex function without regular rewards along the way. That is precisely what Behe said! Avida is a toy, just as Weasel is. Oh, it has a few more steps and a few nuances that have to be parsed through to see what is actually going on. And some people who have an a priori inclination to accept the result without asking the hard questions about how the result came about are easily impressed by the smoke and mirrors. But it is a toy, no more relevant to biology than Weasel and no more able to demonstrate the efficacy of the Darwinian creation theory.Eric Anderson_{October 6, 2011
October
10
Oct
6
06
2011
10:19 AM
10
10
19
AM
PST}

Elizabeth, please respond to the central question. Elizabeth: "In weasel, the problem is trivial and there is only one optimal solution (the sequence of letters that spell “methinks it is like a weasel). So it does not demonstrate that Darwinian algorithms can find unknown solutions to non-trivial problems." Yes, these are problems with the weasel program, but you are skirting the primary issue. The question was whether the solution was programmed in from the beginning. This is important to the circularity issue we have been discussing from the beginning. Do you or do you not acknowledge that the solution was programmed in from the beginning?Eric Anderson_{October 6, 2011
October
10
Oct
6
06
2011
09:58 AM
9
09
58
AM
PST}

DrBot:
Take a real example, evolving a logic circuit: One experiment used a genotype of 1800 bits, specifying the configuration of a 10×10 section of a field programmable logic array (FPGA). With a suitable fitness function the evolutionary algorithm was able to configure those 1800 bits, starting from a totally random population, to produce circuits that could discriminate between square waves of 1kHz and 10kHz.
Designed to evolve/ evolved by design. Thank you for proving our point.Joseph_{October 6, 2011
October
10
Oct
6
06
2011
08:53 AM
8
08
53
AM
PST}

And on top of that, the GA has the enormous benefit of not knowing or caring whether the antenna it "designs" ever gets manufactured. Real world evolution has to design and manufacture simultaneously. The GA is conveniently freed from solving that problem, passing it off to who? Intelligent designers.ScottAndrews_{October 6, 2011
October
10
Oct
6
06
2011
08:40 AM
8
08
40
AM
PST}

DrBot, Evolved antennas and evolved electronics are optimizations of designed functionality. Sure, GAs are handy way to quickly explore possible optimizations, but this has no bearing on the creative powers attributed to evolution. This example sums it up quite well.
Here we describe our work in using an evolutionary algorithm (EA) to automatically design an antenna for NASA's Space Technology 5 (ST5) mission. ST5 is one of NASA's New Millennium Program missions to launch multiple miniature spacecraft to test, demonstrate and flight-qualify innovative concepts and technologies in the harsh environment of space for application on future space missions.
What role does an evolutionary algorithm play in launching a miniature spacecraft? Does it identify the need for miniature spacecraft or the benefit? Does it design the spacecraft from raw materials? Does it invent a propulsion system? Does it determine that the spacecraft may need to send and receive signals? Does it conceive how such signals might be transmitted, received, and interpreted? Does it determine that sending and receiving such signals would require an antenna? None of the above. When the heavy lifting is done it shows up and optimizes the antenna after it's told what to do and what to look for. It's a good application of computer science, but it doesn't make a case that evolutionary processes can invent anything.ScottAndrews_{October 6, 2011
October
10
Oct
6
06
2011
08:33 AM
8
08
33
AM
PST}

The TARGET is programmed in. Elizabeth:
Well, the problem is programmed in. Not sure what you mean by the “target”. The “target” as in “a solution to the problem” is not programmed in, as you agree. So what are you calling the “target”?
The target is what it is you are trying to accomplish. That is the "problem" you are trying to find a "solution" to. How do I get X starting with something that isn't X.
We are talking about Darwinian evolution which pre-supposes a minimally complex self-replicating population.
But the origin of that directly impacts any subsequent evolution. And the resources and means to reach the target is also programmed in.
Well, the “resources” are simply part of the environment,
Not entirely. The resources also include the stuff necessary to reach the goal of the program. In biology this would be the genome/ whatever it is that makes an organism what it is- which we don't know what that is yet (yes we know what genomes are but there isn't any evidence that an organism is a sum of its genome).
What they do demonstrate is that given these two prerequisites, namely critters that replicate with modification and an environment that provides resources and hazards, adaptation (evolution of the population to fit its environment) will occur.
You are confused- ID is not anti-evolution. Rather it argues that the changes that lead to adaptation are directed via some internal programming. ALL genetic algorithms mimic ID not blind, undirected processes.
They mimic blind probes of nearby fitness space.
Fitness space? How are you defining "fitness"? And what GA does this and what is your evidence for that?Joseph_{October 6, 2011
October
10
Oct
6
06
2011
08:14 AM
8
08
14
AM
PST}

That's ok. In fact, yes I was :) By the initial conditions I mean not only the parameter values but the whole configuration for the experiment to run that, I believe, takes us to an island of functionality where a meaningful experiment is possible.Eugene S_{October 6, 2011
October
10
Oct
6
06
2011
08:06 AM
8
08
06
AM
PST}

The question is how this game was put together or how we got to that functional island.
Again, that is a different issue to the question: Can a GA produce CSI A GA clearly has CSI, but I would say there are tons of examples to show that they can generate more CSI. It makes no sense to demand that evolutionary mechanisms be used to explain the origin of evolutionary mechanisms!DrBot_{October 6, 2011
October
10
Oct
6
06
2011
07:59 AM
7
07
59
AM
PST}

6.2.1.2.1 DrBot Unfortunately I can't say at the top of my head whether CSI has been generated here. It is down to detailed analysis but the question is interesting. I accept that some information can be generated spontaneously, the question is how much. E.g. a double coincidental error introduction/correction in a message, however the original meaningful message has an intelligent cause in practice. In any case, this has been a very interesting discussion indeed, thanks to you and Elizabeth. I'll think of this question.Eugene S_{October 6, 2011
October
10
Oct
6
06
2011
07:59 AM
7
07
59
AM
PST}

Sorry, I just realized I misread when you referred to the example - thought you were responding to my example ;)DrBot_{October 6, 2011
October
10
Oct
6
06
2011
07:56 AM
7
07
56
AM
PST}

6.2.1.1.15 Dr Bot, Thanks for the example. Yes I know of capabilities of GAs. But that misses the target. Here GAs are working in an intelligently setup environment. This environment is not a simple sum of components sporadically put together. We are already in a target zone amenable to GAs. GAs have to have something that discriminates between "good" and "bad" and the other rules of the game, the fitness may be changed adaptively, etc. But again it is us who control the action of GAs. The question is how this game was put together or how we got to that functional island.Eugene S_{October 6, 2011
October
10
Oct
6
06
2011
07:52 AM
7
07
52
AM
PST}

Prev 1 … 5 6 7 8 9 … 11 Next