Selling Stupid

I like Word Mutagenation. It nicely demonstrates cumulative selection. Indium

Bflockquote>Does the demo illustrate “the power of cumulative selection”? Unlike Weasel, Itatsi explores any and all of "functional space". It does not have a target. It ranks fitness not by closeness to a target, but by the sum of the fitness of genes. In this case, genes are letter pairs, triplets, etc. that occur in dictionary words. Although you don't need a target, it makes no sense to call something a GA if there is no way to assign fitness. What I find interesting is that Itatsi makes novel words. Words that are not in the dictionary, but which are obviously pronounceable and which look like words. And it can do it in any language that uses the 26 letter alphabet. If you watch a number of demo runs you will notice that it doesn't get stuck trying to reach a specific target. That is because an asteroid comes along occasionally and kills off the "most fit" individual. Making it possible to branch in unexpected directions. It's an exploration of functional space, and it makes it clear that at least some spaces have fitness gradients that are connectable by incremental steps. Petrushka

Nice. Does the demo illustrate "the power of cumulative selection"? Mung

To be more realistic we could perhaps examine the progeny for realistic english words and use the number of identified words as a measure of fitness.

http://itatsi.com Try the default demo. Petrushka

fallacy [?fæl?s?] n pl -cies 1. an incorrect or misleading notion or opinion based on inaccurate facts or invalid reasoning [ --> or emotional manipulation of perceptions or distraction etc] 2. unsound or invalid reasoning 3. the tendency to mislead 4. (Philosophy / Logic) Logic an error in reasoning that renders an argument logically invalid [from Latin fall?cia, from fallax deceitful, from fallere to deceive] Collins English Dictionary – Complete and Unabridged © HarperCollins Publishers 1991, 1994, 1998, 2000, 2003 --> If fallacies were not persuasive, we would not have to be warned against them! kairosfocus

Mung- In my version we shall assume, for the sake of argument, that the “intermediate” strings are a feature of the environment that offers a selective advantage to the closest matching string(s). . . IOW, as soon as you step out of Dawkins’s fictitious world into the real world there is no “power of cumulative selection” worth demonstrating. The string against which the "organisms" are judged represents the environment in the WEASEL program. That is the case whether you are using the intermediate strings you propose, or the distant string as in the original. If you change the string (environment) in random ways in every generation, then there will be no cumulative selection because the changes that get closer to one string will not necessarily be closer to the next string. If there were chaotic changes in the real world environment in a time frame less than a generation, then evolution would not be effective at creating anything. Walter Kloover

Targeted search that rewards nonsense — non functional — phrases on advances in mere proximity?

Yet another "important way" in which the the Dawkins program is misleading! In my version we shall assume, for the sake of argument, that the "intermediate" strings are a feature of the environment that offers a selective advantage to the closest matching string(s). Surely that is at least as "realistic" as the Dawkins program. To be more realistic we could perhaps examine the progeny for realistic english words and use the number of identified words as a measure of fitness. Wait, hasn't this all been tried before? By monkeys, no less? IOW, as soon as you step out of Dawkins's fictitious world into the real world there is no "power of cumulative selection" worth demonstrating. Mung

I intend to write my program in Ruby. http://www.ruby-lang.org/en/ It's a wonderful language that I actually enjoy (gasp!) using to write programs. You can download Ruby from the downloads page. http://www.ruby-lang.org/en/downloads/ If you're running Windows, I suggest using the Ruby 1.9.2-p180 RubyInstaller. It's free. Open source. Mung

Elizabeth Liddle:

Mung, I simply do not understand your version.

You're an intelligent person. You've written your own WEASEL program(s). I find it hard to believe that you don't understand the program I propose to write. Certainly, if you wanted to, you could write it yourself.

And it doesn’t remove the biggest dissimilarity with biology, because you still have a single target to which there is a single solution.

I intend to introduce multiple targets. A new "intermediate" target every generation. My #2. At no point would the string used as the parent for the next generation be chosen based on comparison to the final target phrase. This is a MAJOR difference from Dawkins. Don't you agree? Quoting Dawkins: "...the mutant ‘progeny’ phrases were judged according to the criterion of resemblance to a distant ideal target"

Biological evolution has multiple targets to which there are multiple solutions.

That's irrelevant, actually. But we can deal with that in later versions. Let's get over that first hurdle first. My version proposes a different target every generation. Each string in the population is a possible solution. But please get it out of your head right now that I am proposing to accurately model evolution. Dawkins's WEASEL does not do so and neither shall mine. (At least not at first.) All I'm concerned about is that mine model evolution at least as well as the Dawkins version, if not better. First, we're going to get rid of that oh so non-Darwinian criterion of resemblance to a distant ideal target that every mutant progeny is judged against. Do you just not understand how I propose to do that, or do you assert that I intended to still use that final target phrase as a criterion of fitness? The real debate is over "the power of cumulative selection." What happens to it when we dispense the final target phrase as the judge of "fitness"? Do we still get there in 43 generations? I'd bet the farm! While I retain the "final target phrase" it's no longer there as an ideal. It has no impact on whether a mutant progeny lives or dies. It has no effect on reproductive success. It's only there now to let us know if we've hit it. To be sure, Dawkins admitted his program "is misleading in important ways." He tells us what one of those ways is. You've identified yet another. So now let's test to see just how important those misleading aspects of the program are to his argument, and what happens to his demonstration when you dispense with the misleading bits of it. Interested?

For instance a target might be “be inconspicuous to predators” and solutions might include camouflage, disruptive patterning, movement habits, decoy patterns etc. But “attract a mate” might be another target, and some of the solutions to this target may conflict with solutions to the first.

Why is it, do you suppose, that Dawkins did not incorporate all these things into his program?

WEASEL is simple to a fault, in that it has one target to which there is one solution.

Simple to a fault as a figure of speech? Or do you mean it had only one fault, in spite of Dawkins's own admission that it is "misleading in important ways."

With the proviso of that vast oversimplification, the actual mechanism is the same as is posited in biology.

That cannot possibly be the case. The fact that you, as a very intelligent and educated person think so, highlights just how fraudulent the program really is. Mung

Targetted search that rewards nonsense -- non functional -- phrases on advances in mere proximity? kairosfocus

Mung, I simply do not understand your version. And it doesn't remove the biggest dissimilarity with biology, because you still have a single target to which there is a single solution. Biological evolution has multiple targets to which there are multiple solutions. For instance a target might be "be inconspicuous to predators" and solutions might include camouflage, disruptive patterning, movement habits, decoy patterns etc. But "attract a mate" might be another target, and some of the solutions to this target may conflict with solutions to the first. WEASEL is simple to a fault, in that it has one target to which there is one solution. With the proviso of that vast oversimplification, the actual mechanism is the same as is posited in biology. Elizabeth Liddle

hi kairosfocus, Most of those versions of WEASEL suffer from the same flaw as does Dawkins' version. They use their knowledge of the final target phrase to direct the search to the solution and find it without fail. As Dawkins says, that's misleading:

...it is misleading in important ways. One of these is that, in each generation of selective ‘breeding’, the mutant ‘progeny’ phrases were judged according to the criterion of resemblance to a distant ideal target, the phrase METHINKS IT IS LIKE A WEASEL. Life isn’t like that. Evolution has no long-term goal.

So I want to make one that is less misleading. Doing so will show the fraudulent nature of his version. Oh well. I hear that's how science works. The version I propose will not judge against the distant ideal target. So each new generation is going to get it's own target against which "fitness" will be measured. Perhaps I'll even make a version where the same target will persist for a few generations. But for now, simple is best. The problem is, without that distant ideal, how can we show that "cumulative selection" is all it's cracked up to be? I'm guessing we can't. Having that distant ideal target is precisely what is required to demonstrate "the power of cumulative selection." Mung

Check out here and click the simulations tab. Notice, it has been shown that a well tuned targetted, proximity reward search, suitably tuned, will give on at least some "good" runs a latching of the output letters once they go correct. Latching does not have to be explicit (partitioned search) to appear in the output on runs good enough to showcase. kairosfocus

mung, C++ mike1962

Fraud Weasel

Weasel isn’t just “a bit of a cheat”, it’s a total cheat.

Indeed. mike1962, what language(s) did you use to write your code? I'm considering writing a version of the WEASEL program to show how and why the Dawkins version is fraudulent. Version 1. Use the same target phrase from Shakespeare. Version 1a. Allow a user input phrase (because Version 1 would never find the target, lol). 1. So before we perform any selection I'd look for a phrase that matches the final target phrase. WEASEL = "Methinks it is like a weasel." 2. If none is found then generate an intermediate phrase having the same number of letters. 3. Compare all the strings in the current population to the intermediate phrase. 4. Find the best match. 5. Create the new population from the best match. 6. Apply the mutatation operator to each string in the new population. 7. Start over at #1. This should be simple enough to code. What this version does is remove the first major obstacle or objection to the Dawkins version of the WEASEL program, which is the final distant target the selection operator knows about and compares each string to resulting in the deterministic outcome. We do that by injecting an intermediary, which is the "current environment" for each generation. We keep the best match and move on to the next generation and it's current environment. Now how many generations will it take to find the same target phrase from Shakespeare? Can we show that this will perform better than a blind search? Are we still not using "cumulative selection?" And if not, why not? Weasel is a fraud. Mung

Liz: But if all you want is any solution to the problem of survival in a given environment, all you need is the environment.

This is the hypothesis, that a mechanism is robust enough to do exactly that in a "blind" way without any intervention of intelligence along the way.

"And the random generator/self-replicator are simply the prequisites for natural selection.

"Simply" she says :)

You may well question whether those could have arisen from non-selfreplicating entities, but that wouldn’t be a critique of Darwinism, but an OOL question."

Wrong. It is a critique of "blind" evolution. And the entire question is intimately tied to OOL. I'm not anti-evolution, mind you. I'm not a Bible thumping creationist who thinks "God done it" in six 24-hour days. What I oppose is the notion that it's a blind process, with no intelligent infusion(s) leading from the LUCA to man. The notion is folly and seriously lacking in real-world provenance. You know that. ;) mike1962

The fitness function has to be "intelligently crafted" if you want either a solution to a specific problem, or a specific solution. But if all you want is any solution to the problem of survival in a given environment, all you need is the environment. And the random generator/self-replicator are simply the prequisites for natural selection. You may well question whether those could have arisen from non-selfreplicating entities, but that wouldn't be a critique of Darwinism, but an OOL question. Elizabeth Liddle

And WEASEL is important for another reason: It counters the stupid “tornado in a junkyard”-calculations of some anti-evolutionists because it shows that the odds can be changed dramatically with cumulative selection

Right, it shows how the odds can be changed for hitting a target string using an intelligently crafted random generator and fitness function. A great example of ID. mike1962

Indium 1. Real biological mutations can happen for any gene. Since there is no target explicit latching is completely impossible in biology anyway.

Sidebar: I disagree that it is conceptually impossible to latch features in biological entities. One possible way to do it is with redundant checks that would "kill the baby" if certain features varied beyond a certain range. And the checks have checks too in such as way that any damage to any part would "kill the baby." Etc. At any rate, the characters in Weasel are mere analogies to genes. They are not genes or anything like genes.

2. For high mutation rates and low population sizes sometimes good characters can indeed be lost again, even for a program like WEASEL for which such an event is extremely unlikely because of the way the parent is chosen.

Right. But no explicit latch is required because the way the mutation and fitness algorithms work together it guarantees the outcome: a given character has a low change of mutation WHILE adding to the fitness score if it matches the parent. What "fitness score" is judging genomic variation? "The environment?" But "the environment" is blind, there is no target to compare it to. So Weasel not analogous at all to blind evolution. Remember, what Dawkins says, "it's a bit of a cheat. Weasel looks to the future. [Blind] evolution doesn't look to the future." But Dawkins doesn't make the point strong enough. Weasel isn't just "a bit of a cheat", it's a total cheat. He should have just use an example of sorting socks from a drawer by color, or cite how a building is constructed. Both have a target and a fitness function and accumulate small changes. And they're easier to grasp conceptually. (But they are too obviously "intelligent", I would suspect, for his purposes.) Weasel is a great example of ID.

Again, WEASEL is therefore more closely (and still only loosely…) related to evolution and can at least demonstrate a few additional effects and dependencies that are also known in biology.

The mutation may demonstrate a random variation but the fitness function kills any analogous value it may have to blind evolution. The mechanism taken as a whole is no demonstration of blind evolution, and is needlessly complicated for the intended audience of Blind Watchmaker.

3. Implementing explicit latching needs more code. Why do it when there is no reason for and several against it?

Explicit latching requires less code. I've written it both ways. I think this particular horse is beat to death. mike1962

Mike1962

The randomness part of Weasel is an unnecessary complication given that the method of random generation and the fitness function guarantees the outcome within a statistically predictable number of iterations. And might as well just latch the characters and get it done faster. The outcome is the same and it still demonstrates cumulative selection

One can guess several reasons why explicit latching was not part of WEASEL: 1. Real biological mutations can happen for any gene. Since there is no target explicit latching is completely impossible in biology anyway. 2. For high mutation rates and low population sizes sometimes good characters can indeed be lost again, even for a program like WEASEL for which such an event is extremely unlikely because of the way the parent is chosen. Again, WEASEL is therefore more closely (and still only loosely...) related to evolution and can at least demonstrate a few additional effects and dependencies that are also known in biology. 3. Implementing explicit latching needs more code. Why do it when there is no reason for and several against it? For me, it´s still a mystery why so many people make such a fuss about the WEASEL program. It has absolutely zero scientific value. The only purpose was the demonstration for a lay audience that there can be an extremely large difference between a random search and a more or less cumulative search with some form of selection acting on variable offspring candidates. So what? Mung There is nothing fraudulent about WEASEL. It just shows the simple fact that cumulative selection is (not very surprisingly) much more efficient than selection acting only on the full target string (random search). You may doubt that in every case there is a cumulative path available in real biological evolution. But this does not change the fact that if there is such a path (as in WEASEL) then cumulative selection is extremely efficient. And WEASEL is important for another reason: It counters the stupid "tornado in a junkyard"-calculations of some anti-evolutionists because it shows that the odds can be changed dramatically with cumulative selection. This of course means that antievolutionists have to work harder and invent more sophisticated tools like CSI, FDCSI or whatever. Interstingly, all these funny new information measures still in some way reference the basic tornado-in-a-junkyard probability... Indium

WR is clueless. Weasel is a fraud. It's handwaving. Don't watch what I am doing with my other hand. 1. The program was designed for a purpose. 2. That purpose was to show how evolution makes the appearance of highly improbable things fall easily within the realm of what is probable. 3. It fails to do that. Yet people like you continue to believe that it has not failed. That's what makes it fraudulent. Dawkins was not honest about it. Mung

Mung,

If I was using single-step selection I’d get it right the very first time, since I know what the target is

Indeed! I will write to Dawkins and let him know that he can just write the sentence out! There's actually no need to go to all the trouble of writing a program (latching or non latching) to do it.

The claim that Dawkins makes is that a search algorithm that uses “cumulative selection” performs better than a random search. But when does a search algorithm that uses “cumulative selection” perform better than a random search. And why?

All that and more has been covered in the links I have previously added to this thread. The mathmatics has been worked out to the Nth degree already. If you really wanted to know the answers then you easily could have.

These are the things we need to know to test the claim. These are the things we are not told.

What are you blithering about? What claim? What is it that you are not being told? I linked to the actual New Scientist article, you can read it for yourself. If you were after the truth rather then a talking point then perhaps you would have already read it. Then you would be as up to speed as the rest of us. Or read the book. Have you ever stopped to think that perhaps it's not a conspiracy, perhaps it's just that you don't know as much as you think you know? Why Elisabeth takes the time out to educate you I don't know. Good for her but I could not bear to do it, not climb that mountain. WilliamRoache

Yup, deniable, useful goons doing the Alinskyite dirty work not even realising they are cat's paws for a dangerous, destructive radical agenda -- did you see how they could not even apply the Categorical imperative to identify what is morally unsound? No need for a widespread conspiracy, just have a few sponsored agents of influence set up astroturf fever swamps for mobs of the ill adjusted and educationally lobotomised who are bewitched by the Plato's cave shadow shows that stand in for reality (why else do you think they want to control education and the media) to stoke up on rage, deceptive talking points and turnabout accusation tactics. Then, point them to latest targets to swarm down. But this time they picked the wrong target: my family. That's a nuke threshold, a point of no return. kairosfocus

kairosfocus, what you are bringing to light is a common strategy. I remember I challenged Dawkins on a youtube video and I got the same email deal for a week lol. Not from Dick of course, some minions. junkdnaforlife

Paragwin: Nope, while your remark is distractively tangential in this thread, I will give a one-shot response for record. I will take note of a thread-jacking here, too. "William Roache" has shown the significance of Plato's warning on the ruthless, amoral agendas and factions that evolutionary materialism lets loose in a culture when it is pushed and backed by powerful elites. It is very clear that this is a voice for the faction that has sought to hold my family hostage under a mafioso type threat, and is seeking to "justify" that by labelling me an insane bigot. By means of turnabout accusation. He has no hesitation to refuse to address the pivotal issue, the grounding of OUGHT, and the question, what IS can bear the weight of ought. So, what he is doing -- just as what is plainly his ilk have tried to do with my family -- is to try to hold language and ideas hostage. Words like"rights," "marriage" and -- in the wider context -- even "science," are being radically redefined and that redefinition is pushed on us through intimidation tactics. Ultimately, if you dare object, such intend to rob you of your livelihood [hence the expelled phenomenon], then -- as this case shows -- your family [you are a child abuser ("thank you" Mr Dawkins for that particular nasty slander . . . ) and are unfit to be foster then adoptive then eventually natural parents], and your conscience or soul [how dare you think or teach differently than we do on what we feel is our "right"]. He who would rob me of my daily bread, would rob me of my life. He who would rob me of my children, would rob me of my posterity. He who would rob me of my conscience, would rob me of my soul. Notice how, in his agenda, he refused to engage in the foundational issue of the is-ought challenge, and how he refused to take the steps to assess per the Categorical Imperative what is morally sound or unsound. Instead, he arrogantly and insistently demands, three times, that I answer yes or no to his artfully loaded questions that in this context are really accusations that beg a lot of questions. Yet another warning about what we are up against. So, the real question is whether we will learn from the lessons taught us by history and those who have had to reflect upon it. Let Mr "Roache" and ilk boast elsewhere as they will, the record is plain enough. On the focal topic for this thread, let us note the unanswered challenge that Gil posed in the OP:

One of my specialties in aerospace R&D engineering is guidance, navigation and control software. The task of designing GN&C algorithms and the associated hardware that would permit an ornithopter to land on a swaying tree branch in gusting wind is so far ahead of our most sophisticated human technology that we can only dream about such a thing. Yet, birds do this with ease. Darwinists want us to believe that this all came about through a process of throwing monkey wrenches in working machinery and introducing random errors into highly sophisticated computer code . . .

How do they try to get us to accept this as though such a practical impossibility were a "fact"? Why, by redefining science to confine it to a materialistic straightjacket, and to then intimidate those who would beg to ask inconvenient questions, of course. The same tactics, the same tactics. It is time for us to heed Plato's warning. GEM of TKI F/N: On the immediate topic in this thread, which is related to the main topic, M62 nails it well in 261. That is how a Weasel is designed, tuned and implemented. Whether or not it explicitly or implicitly latches, that is what it is doing, and how. In addition, as I warned by clipping already, it is all a sham, a distraction from the key question -- same tactics, yet again -- of arriving at islands of function in large config spaces, as it works by rewarding non-functional nonsense phrases on mere increments in proximity to target. Impose the criterion that it must provide a set of functional words from the dictionary, even if unrelated to make progress, and at once it would collapse. So, how much moreso, are we to see that the complex code in DNA for functional body plans cannot credibly be explained on accumulations of errors filtered by trial and error. But, as we have been seeing, that is what the new misleading icon, Weasel and kin, are being used to bewitch many into believing. kairosfocus

Kairosfocus, It seems WilliamRoache has matched your masterful baiting skills. paragwinn

DrBot @232:

Obviously I was just blowing smoke...

Obviously. And you can't cover up the fact that you were just blowing smoke by blowing yet more smoke. Citing references might help though. I'm not going to hold my breath though. I sense the sort of smoke you're blowing isn't fatal. Mung

The illusion begins by talking abut single-step selection rather than search. If I was using single-step selection I'd get it right the very first time, since I know what the target is. He's not really comparing different types of selection, he's comparing different search algorithms. The claim that Dawkins makes is that a search algorithm that uses "cumulative selection" performs better than a random search. But when does a search algorithm that uses "cumulative selection" perform better than a random search. And why? How much better does a search algorithm that uses "cumulative selection" perform than a random search algorithm? These are the things we need to know to test the claim. These are the things we are not told. Mung

What's Wrong with WEASEL The Weasel program was designed to make a point. That essential point was that there are better (more efficient) ways to find something than looking for it by using blind chance mechanisms. Supposedly, evolution offers just such a way. So while an eye may be extremely complex, and hard to find by a stochastic search process, evolution makes it infinitely more probable to find an eye than does using a blind search. The proof: WEASEL. 28 slots. 27 characters. What's the probability of choosing each character at random for each of the 28 slots and finding that our string matches the target phrase. Slim and none. How many monkeys would it take? How many typewriters? How much time? But if we use cumulative selection! Now let me ask a question. Let's say that cumulative selection offered a 10x improvement over blind search. How many generations would it then take? But I doubt that would have impressed Dawkins readers. Let's say that cumulative selection offered a 100x improvement over blind search. Impressive yet? So just how much better is cumulative selection than a blind search? Eh, Mr. Dawkins? The only thing Dawkins demonstrated with WEASEL is that people are gullible when faced with junk science that reinforces what they already want to believe. Mung

WR:

So two different characterizations of Weasel, two different outcomes. Both get there.

Not in dispute. See my comments on latching earlier in this thread.

What, biological evolution? It’s an example.

An example of what? Intelligent Design?

So we’re all agreed I think that cumulative selection is an unavoidable component of micro-evolution.

Well, the only example of cumulative selection I've seen has a distant target known in advance. So if that's what you mean... Mung

WilliamRoache to KF: Funny how “explicit” latching became “implied” latching.

No, it isn't funny. The way these Weasel programs are designed guarantee a nearly 100% latching for a given matching character since for any given iteration it A) usually avoids the (necessarily low) mutation rate and B) adds to the fitness score. In effect, the fitness function rewards the string if it contains more characters that would be latched if we were explicitly latching them. Through the backdoor or through the front, the effect is the same. The randomness part of Weasel is an unnecessary complication given that the method of random generation and the fitness function guarantees the outcome within a statistically predictable number of iterations. And might as well just latch the characters and get it done faster. The outcome is the same and it still demonstrates cumulative selection. mike1962

p.s. As near as I can tell the two links refer to the same program. Mung

WR:

I like this version of Weasel: 5. Base a new generation population of size n upon copies of the closest matching string or strings, where each position has a chance of randomly mutating, based upon a set mutation rate.

I was referring to the one written by Wesley R. Elsberry at this page: http://www.antievolution.org/features/ec/weasel102.html This array holds the strings, which themselves are an array of characters. gdwr.pop = new Array(); I'll let you guess what this array holds: gdwr.closest = new Array(); Decide whether we're going to mutate the string: gdwr.mutate = Math.random()*100.0; if (gdwr.mutate <= gdwr.mutrate) If we decide not to mutate use an exact copy of the closest match in the new generation. gdwr.pop[ii][jj] = gdwr.closest[jj];

Base a new generation population of size n upon copies of the closest matching string or strings...

A tad misleading that. Mung

WilliamRoache: The exact number of code loops used to to get there is a different and irrelevant matter.

Try it with one internal loop, i.e, one string mutation and see how long it takes per generation. Try it with 100. And see how long it takes. Again, here's what I challenged:

If you can show me a program that ends up with a target string after 43 stabs at the entire string with any mutation rate for each character that does not use character latching, I’ll eat my hat.

Granted, given a re-reading of his description I agree that his method didn't employ latching, and it did imply that each generation had multiple string mutation candidates. My bad. I hereby withdraw my challenge as irrelevant. OK, so how many was it per outer loop (generation)? If we divide 2485 by 43 (assuming that the 2485 string mutations were scattered over 43 out loops), that gives about 57 string mutations per outer loop.

That is why Dawkins’ said himself that it was not a great example. But if understood in context, with the point being illustrated in mind, it makes perfect sense.

I agree, and have never said otherwise. And furthermore, I agree with Dawkins who said it with regards to Weasel being analogous to blind evolution, that it "was a bit of a cheat", that his program "looks to the future", and that blind "evolution does not look to the future." Yep. Everybody seems to agree with that. As I've said repeatedly, the point of this whole Weasel business was simply to demonstrate cumulative selection on small changes. It does nothing more or less than this. Randomly picking colored socks from a drawer and stacking them by color demonstrates the same thing. (The randomness part of Weasel is an unnecessary complication given that the method of random generation and the fitness function guarantees the outcome within a statistically predictable number of iterations. Might as well just latch the characters and get it done faster. The outcome is the same and it still demonstrates cumulative selection.) mike1962

KF,

Weasel should be retracted and apologised for.

Perhaps you would care to write an example that illustrates the vast gulf between random noise and cumulative selection, in a format understandable to the lay reader? Then perhaps you also could write a letter to 1986 and let them know?

Weasel should be retracted and apologised for.

But why? Dawkins explains the limitations in the very text he uses to tell us about it! The point is that cumulative selection is real even when the targets are very broad. Like "survive longer". Many paths to that.

De confession:

Confession? Of what? That evolution has no long term target? No target other then in relation to the environment or population pressures etc? Congratulations. Once again, I'll let 1986 know. Apparently this is somehow news. Yet you crow over it like the whole "weasel" thing has ever done you any favors. All weasel has ever done for you is shown your inability to admit error. You posting of Dawkins' own words at this point as a "confession" is actually immensely illustrating of your self aggrandizing nature. Dawkins confessed his sins and KF has the quote to prove it! Funny how "explicit" latching became "implied" latching. Funny how "implied" latching is cumulative selection. Funny funny funny. WilliamRoache

mike

Show me his source code and let’s see if he doesn’t latch. Or show me anyone’s source code that can do it in 43 iterations without latching.

It's not 43 iterations it's 43 generations. Doing it in 43 iterations would be silly! Might as well get a monkey to bang on a keyboard instead. Dawkins said nothing of the number of code loops that had to run,

It’s 10350 iterations on the subject string however way you nest the looping.

nothing about how the looping is nested. Your objections are in fact spurious. All you need to be able to reproduce the program exactly as he had it is in the description given by Dawkins. If you do that you can get a run with 43 generations. The exact number of code loops used to to get there is a different and irrelevant matter. WilliamRoache

F/N: Rollin de tape . . . 1 WDL*MNLT*DTJBKWIRZREZLMQCO*P 2? WDLTMNLT*DTJBSWIRZREZLMQCO*P 10 MDLDMNLS*ITJISWHRZREZ*MECS*P 20 MELDINLS*IT*ISWPRKE*Z*WECSEL 30 METHINGS*IT*ISWLIKE*B*WECSEL 40 METHINKS*IT*IS*LIKE*I*WEASEL 43 METHINKS*IT*IS*LIKE*A*WEASEL De confession: >> I don't know who it was first pointed out that, given enough time, a monkey bashing away at random on a typewriter could produce all the works of Shakespeare. The operative phrase is, of course, given enough time. [[NB: cf. Wikipedia on the Infinite Monkeys theorem here, to see how unfortunately misleading this example is.] Let us limit the task facing our monkey somewhat. Suppose that he has to produce, not the complete works of Shakespeare but just the short sentence 'Methinks it is like a weasel', and we shall make it relatively easy by giving him a typewriter with a restricted keyboard, one with just the 26 (capital) letters, and a space bar. How long will he take to write this one little sentence? . . . . It . . . begins by choosing a random sequence of 28 letters ... it duplicates it repeatedly, but with a certain chance of random error – 'mutation' – in the copying. The computer examines the mutant nonsense phrases, the 'progeny' of the original phrase, and chooses the one which, however slightly, most resembles the target phrase, METHINKS IT IS LIKE A WEASEL . . . . What matters is the difference between the time taken by cumulative selection, and the time which the same computer, working flat out at the same rate, would take to reach the target phrase if it were forced to use the other procedure of single-step selection: about a million million million million million years. This is more than a million million million times as long as the universe has so far existed . . . . Although the monkey/Shakespeare model is useful for explaining the distinction between single-step selection and cumulative selection, it is misleading in important ways. One of these is that, in each generation of selective 'breeding', the mutant 'progeny' phrases were judged according to the criterion of resemblance to a distant ideal target, the phrase METHINKS IT IS LIKE A WEASEL. Life isn't like that. Evolution has no long-term goal. There is no long-distance target, no final perfection to serve as a criterion for selection, although human vanity cherishes the absurd notion that our species is the final goal of evolution. In real life, the criterion for selection is always short-term, either simple survival or, more generally, reproductive success. [[TBW, Ch 3 >> Weasel should be retracted and apologised for. Period. GEM of TKI kairosfocus

M62: A key to that run is it starts with 3 correct letters. Look for a single letter that reverts in the sample. Ent dere G kairosfocus

mike,

You’ve just disqualified yourself from serious discussion.

So let's say you are right. That when construed a particular way (your way) that the statement "43 generations" is inaccurate.

Target string plus a fitness function and mutation rates that guarantee the outcome by the programmer.

Yes. Precisely. That is why Dawkins' said himself that it was not a great example. But if understood in context, with the point being illustrated in mind, it makes perfect sense.

That’s laughable. It’s 10350 iterations on the subject string however way you nest the looping.

Yes, laughable. Yet if Dawkins' words are read at face value you can run an example of his program, created from his original description, and have it come out at 43 generations. If however you rule "latching" out, you rule cumulative selection out, thereby ruling out the entire point of the toy example in the first place. Congratulations. You should write to New Scientist in 1986. Let them know that if you gut the example of it's only reason for existing in the first place it stops working. WilliamRoache

Mung, I like this version of Weasel:

1. Use a set of characters that includes the upper case alphabet and a space. 2. Initialize a population of n 28-character strings with random assignments of characters from our character set. 3. Identify the string or strings closest to the target string in the population. 4. If a string matches the target, terminate. 5. Base a new generation population of size n upon copies of the closest matching string or strings, where each position has a chance of randomly mutating, based upon a set mutation rate. 6. Go to step 3.

From here: http://tinyurl.com/6xxfwry That seem about right to you?

If you ask me, that’s as good as explicit latching.

No, not quite as good. Check out the link I posted earlier. From the link in this post:

At higher mutation rates, it is clear that the Dembski-Marks "oracle weasel" has the advantage in performance due to "locking-in" of correct letters, which the accurate Dawkins' "weasel" does not do. But the graph clearly shows that even without "locking-in" of letters the accurate Dawkins' "weasel" does eventually get to the correct string, and takes only slightly over ten times as many candidates as does the Dembski-Marks "oracle weasel".

So two different characterizations of Weasel, two different outcomes. Both get there.

I use that one single string as the basis for the entire next generation.

Not necessarily.

There’s more than one way WEASEL isn’t like evolution.

What, biological evolution? It's an example. How many strings would it have to use as the basis for the next generation for it to become an accurate example of evolution? How many would satisfy you? I'm surprised you have not already written the program, it's got to be a trivial extension to the original. So if extending it like you say does not make it "more like" evolution then perhaps the point has been missed. He's using it as an example in comparison to the "monkey on a keyboard" example so beloved of some commentators here who think that biologists believe cells came together in a unlikely whirlpool of lucky chance and chemicals like a film of a cell being pulled apart but in reverse.

If you ask me, that’s as good as explicit latching.

Check the link. It's not quite as good. But so what, it's not supposed to be nor was it claimed to be an accurate example of "evolution". It was just used to point out the potential time difference between cumulative selection and random chance stumbling on the same thing. So we're all agreed I think that cumulative selection is an unavoidable component of micro-evolution. How could it not be, right? Explicit or otherwise, latching is real. WilliamRoache

Check the actual text. He talks about generations. Dawkins: And the target was finally reached in generation 43

Show me his source code and let's see if he doesn't latch. Or show me anyone's source code that can do it in 43 iterations without latching.

No, it’s how you get the effect without explicit (i.e. “intelligent”) latching.

All the non-latching versions of Weasel are "intelligent." Target string plus a fitness function and mutation rates that guarantee the outcome by the programmer.

What do you expect to happen exactly if some configurations outperform others against a given measure and breeding is based on that success?

I expect just what the programs are designed to do.

Congratulations, that’s exactly right.

Congratulations? I've never said otherwise.

What it illustrates is how shallow some ID supporters think evolution is. Debunk Weasel, evolution is debunked.

Good for them. That has nothing to do with me.

Mike1962, if you can show me a program that ends up with a target string after 43 stabs at the entire string with any mutation rate for each character, I’ll eat my hat. WilliamRoache: I already did! Just wrap the code up into another function that re-runs the code until you get the generation number 43.

"Just wrap"? That's laughable. It's 10350 iterations on the subject string however way you nest the looping. You've just disqualified yourself from serious discussion. mike1962

mike, http://tinyurl.com/6japrus Check the actual text. He talks about generations.

Dawkins: And the target was finally reached in generation 43,

Mike, you said:

This is a backdoor means to the same effect.

No, it's how you get the effect without explicit (i.e. "intelligent") latching. What do you expect to happen exactly if some configurations outperform others against a given measure and breeding is based on that success?

In other words, Weasel is not analogous to blind evolution.

Congratulations, that's exactly right. It's a toy example designed to illustrate a single point and instead it's become an obsession of the ID movement. What it illustrates is how shallow some ID supporters think evolution is. Debunk Weasel, evolution is debunked. Hardly.

If you can show me a program that ends up with a target string after 43 stabs at the entire string with any mutation rate for each character, I’ll eat my hat.

I already did! Just wrap the code up into another function that re-runs the code until you get the generation number 43. Per generation. WilliamRoache

So let's say I have a population size of 100 strings. I choose one that most closely matches the target phrase. I use that one single string as the basis for the entire next generation. 100 strings just like it form the next population prior to applying the mutation operator. If you ask me, that's as good as explicit latching. There's more than one way WEASEL isn't like evolution. Mung

William Roache, I meant to say on #244 that: If you can show me a program that ends up with a target string after 43 stabs at the entire string with any mutation rate for each character that does not use character latching, I’ll eat my hat. mike1962

William Roache: So what’s your actual point?

Also, that the Weasel program describe in Dawkin's book did use latching if the actual number of attempts on the string were really 43. Does it matter? No. The latching or non-latching issue is irrelevant to the purpose of Weasel, which is to demonstrate cumulative selection. The way all the non-latching algorithms are written, *guarantees* that an implicit "fuzzy latching" will occur a great percentage of the time simply because a given character will not be altered while at the same time contribute to a higher fitness score for that string. This is a backdoor means to the same effect. mike1962

That Weasel is a fraud? That Dawkins misrepresented it?

No. Dawkins has never (to my knowledge) used Weasel in a fraudulent way. His purpose was to demonstrate the accumulation of small changes contra the "junkyard 747" idea of the necessity of large scale change all at once. To that end, it serves it's purpose. It serves no other purpose. As Dawkin's acknowledged in that video, Weasel is "a bit of a cheat. It looks to the future, whereas [blind] evolution does not look to the future." In other words, Weasel is not analogous to blind evolution. And the mechanism certainly is nothing like biological systems. Weasel is no more analogous to evolution than is someone pulling random socks out of a drawer and stacking them in piles according to color. Nobody disputes that small changes can accumulate if there some targeted selection going on. mike1962

WilliamRoache: So what’s your actual point?

See #152: https://uncommondesc.wpengine.com/intelligent-design/selling-stupid/comment-page-3/#comment-387178 mike1962

WilliamRoache: Are you now a believer in the power of non-latched cumulative selection now then?

Nope. Not for 43 attempts. That Python example has 10250 attempts with a 1% mutation rate per character. (My own experiments suggest 3% is a better figure using cryptographic grade random number generation.) If you can show me a program that ends up with a target string after 43 stabs at the entire string with any mutation rate for each character, I'll eat my hat. mike1962

mike1962,

That is not 43 generations as Dawkins claimed in his book. So either it’s a misprint, or he’s using different program here to generate the results than the one he used for his book.

So what's your actual point? That Weasel is a fraud? That Dawkins misrepresented it? That perhaps two programs were written, or one program at different stages in it's development were used or two different outputs from the same program were used in different formats? That generations and individuals might be different numbers? That you can write the program yourself using the description he provided in a paragraph or two and come up with the same results, no latching required? Please clarify! There are of course other possibilities I am yet to enumerate. I could try to take more wild stabs in the dark but help a guy out here, please! WilliamRoache

WilliamRoache Am I missing something? The code here: http://austringer.net/wp/index.....th-part-2/ Includes an example run taking “only” 41 generations. And that’s under 43.

Yes, you are apparently missing something. That Python example makes 10250 attempts. Multiply the outer loop count (41) by the number of generations (250). That's 10250 just like the output indicates. mike1962

WilliamRoache: Here is a video of Dawkin’s Weasel, as written originally. http://austringer.net/wp/index.....-evidence/

Take a look at that video at 9:47. Notice how "Darwin" reaches the goal at 2485 attempts. That is not 43 generations as Dawkins claimed in his book. So either it's a misprint, or he's using different program here to generate the results than the one he used for his book.

However even video evidence will not convince those who do not want to be convinced.

Convince me of what? mike1962

mike1962,

I would like to see anyone to write a program that can get to Dawkin’s target string in 43 iterations without using explicit latching

Am I missing something? The code here: http://austringer.net/wp/index.php/2009/04/20/the-weasel-saga-with-math-part-2/ Includes an example run taking "only" 41 generations. And that's under 43. Now you've seen what you'd like to see (and more) now what? Unless you can point to the line (there are only 34 of them) of code that does the latching you've seen what you wanted to, right? Are you now a believer in the power of non-latched cumulative selection now then? :P Or perhaps it's that it was Dawkins who made the claim he did it in 43 and by definition anything he says is impossible or just plain wrong? But that can't be it, can it? WilliamRoache

Non latching (real) Weasel: http://www.antievolution.org/features/ec/weasel102.html WilliamRoache

mike1962, excellent point. It takes about 26 queries for a deterministic search. Mung

mike1962,

I would like to see anyone to write a program that can get to Dawkin’s target string in 43 iterations without using explicit latching:

That sounds like the original Dawkin's version to me. If you think otherwise please provide a quote from Dawkins that shows he intended letters to be fixed in place once correct. I offered $10,000 to KF for such a quote previously but he never managed to come up with one. Here is a video of Dawkin's Weasel, as written originally. http://austringer.net/wp/index.php/2009/03/15/dembski-weasel-and-video-level-evidence/ However even video evidence will not convince those who do not want to be convinced. You can see the mathematical differences between explicitly latched and unlatched versions here: http://austringer.net/wp/index.php/2009/04/18/the-weasel-saga-with-math-part-1/ There's not that much of a difference really. WilliamRoache

DrBot, While it is certainly possible to write Weasel programs without explicit latching, I would like to see anyone to write a program that can get to Dawkin's target string in 43 iterations without using explicit latching: Generation 01: WDLTMNLT DTJBKWIRZREZLMQCO P [2] Generation 02: WDLTMNLT DTJBSWIRZREZLMQCO P Generation 10: MDLDMNLS ITJISWHRZREZ MECS P Generation 20: MELDINLS IT ISWPRKE Z WECSEL Generation 30: METHINGS IT ISWLIKE B WECSEL Generation 40: METHINKS IT IS LIKE I WEASEL Generation 43: METHINKS IT IS LIKE A WEASEL mike1962

Mike1962 “The scientific methodology you apparently are happy to hang you hat on excludes design a priori” Liz: no it doesn’t. Ask any forensic scientist

Right. Forensic scientists don't have an a priori commitment against intelligent causation. If you don't either, then that's great. mike1962

DrBot:

What would a non-latching system look like?

One that lacks either an explicit or an implicit latching mechanism.

There seems to be a difficulty understanding the difference between mechanism and the effects of mechanism. If the word "latching" means anything at all, it must mean that "correct" letters cannot mutate. This would require an elaborate mechanism within the mutation mechanism. Not only would the "best" of a generation not show mutations at the correct locus, but none of the offspring would show such mutations. This is quite clearly not the case in any practical GA. What you call implicit latching is simply the effect of selection. What a GA without latching or implicit latching would be what real GAs look like. Random variation, fecundity, and selection. The appearance of implicit latching simply reflects the rarity of simultaneous compensating mutations. Petrushka

mung: Was this in the segment of code you posted earlier?

Yep. From the website you provided. mike1962

DrBot:

What would a non-latching system look like?

One that lacks either an explicit or an implicit latching mechanism. Mung

Do you have a reference for that? Or were you just blowing smoke?

Obviously I was just blowing smoke, if you look through the literature you will see that when people refer to a targeted search, as compared to a general search or an un-targeted search, they mean that an un-targeted search and a general search are the same as a targeted search and by using a targeted search instead of a general search or vice-versa they are saying that one type of search was more useful for their research than the other type of search, which was the same type of search as the one they were using before, but is better even though it is the same because a targeted search is a search for some things, and so is an un-targeted or general search, which is why they add the term targeted to some instances of the word search but general or un to others, or just use the word search - because they all mean the same thing! DrBot

Latching at the individual level, where you look at an individual and say “don’t change that” I would call explicit latching. I think that’s what Dawkins has been accused of doing.

Which would have been a mistake on Dawkins part if he did. He doesn't describe any mechanism in WEASEL for locking letters when they are correct and if you implement WEASEL without such a mechanism it works fine.

Then you need to properly define what you mean by implicit.

Basically what you described as selection. Latching via population level mechanisms.

So explicit latching is a non reversible change whilst implicit latching is reversible - anything that can change, then change back is implicit latching. So any system that can change is latching. What would a non-latching system look like?

In the case of WEASEL it’s clearly non-Darwinian and Dawkins admits it.

Yes, because it is a targeted search (there is an explicit target) rather than a general search to match multi variable criteria. Like biological evolution there is no latching mechanism.

I think that’s true of all GA’s.

You are wrong. DrBot

DrBot:

...why don’t you have a go at finding and reading them, it might help you understand the subject a bit better!

And why don't you go read up on targeted searches. Then we can share what we learn. DrBot @28:

Generally speaking any search has to have a target or it isn’t a search, but ‘targetted’ means something more specific – pre-specified and unchanging (I believe – but I could be wrong)

But the next day you were apparently much more sure of yourself. DrBot @74:

The phrase ‘targeted search’ is more specific and refers to having an explicit, pre-specified goal, rather than a set of dynamic criteria to match. Think about it this way – if all searches have to, by definition, have some form of ‘target’ in the loosest sense then arguing that they are therefore all ‘targeted searches’ is pointless semantics – the word ‘targeted’ is redundant because they are all searches and by definition there have to be targets of some kind. The inclusion of the word ‘targeted’ is there to indicate reference to a subset of search evaluation criteria, namely those that are explicitly defined and fixed in advance.

Do you have a reference for that? Or were you just blowing smoke? Mung

DrBot:

Then you need to properly define what you mean by implicit.

Basically what you described as selection. Latching via population level mechanisms. Latching at the individual level, where you look at an individual and say "don't change that" I would call explicit latching. I think that's what Dawkins has been accused of doing.

So if you decide that selection is actually ‘implicit latching’ then would you consider selection to be non Darwinian?

It depends on why the selection is taking place. In the case of WEASEL it's clearly non-Darwinian and Dawkins admits it. I think that's true of all GA's. Mung

I would assume it’s because the changes accumulate. From which it follows they are retained. So the question is, what is the mechanism by which they are retained, and is that mechanism non-Darwinian.

It's called selection - in the case of WEASEL, strings with a higher fitness have a greater probability of being copied, and consequently the correct letters that give them high fitness get copied (preserved) but mutation is random with respect to fitness so any letter, even a good one, can be mutated.

I take the position that “latching,” whether explicit or implicit is not non-Darwinian.

Then you need to properly define what you mean by implicit.

To me what makes a process non-Darwinian is why the latching takes place.

So if you decide that selection is actually 'implicit latching' then would you consider selection to be non Darwinian? DrBot

The question was, why does Dawkins call it cumulative selection. I would assume it's because the changes accumulate. From which it follows they are retained. So the question is, what is the mechanism by which they are retained, and is that mechanism non-Darwinian. I take the position that "latching," whether explicit or implicit is not non-Darwinian. To me what makes a process non-Darwinian is why the latching takes place. Mung

1. You don’t need a selection mechanism. 2. The fitness function does not have to be designed. 3. You can generate your fitness function randomly. What are you talking about?

Evolutionary algorithms, models of biological evolution, and evolution its self. It's quite obvious, and there are tens of thousands of research papers available on these subjects that you could look at if you wanted to - why don't you have a go at finding and reading them, it might help you understand the subject a bit better! DrBot

mung

2. The fitness function does not have to be designed.

they can be just the result of existing in the world - but I suspect you might confuse 'designed' with 'exist' ;) DrBot

Why then does he call it cumulative selection?

Ever done any hill walking? "Cumulative height gain is a measure of the total height climbed along the route ignoring the downhill sections." A cumulative measure can be one of movement in one direction, ignoring movement in another direction, so cumulative selection does not imply unidirectional movement (i.e. no latching). DrBot

Elizabeth Liddle @166:

No. You don’t need a “selection mechanism”. All you need is a fitness function, and as is said, you can generate that randomly as well. But obviously if you do in fact want to solve a problem for your own benefit (like writing a nice bit of useful code) then you will design your fitness function. But it doesn’t actually have to be designed.

Hi Elizabeth. Lest I misunderstand you, please put the above quote in context for me. Thanks 1. You don't need a selection mechanism. 2. The fitness function does not have to be designed. 3. You can generate your fitness function randomly. What are you talking about? Mung

Elizabeth Liddle:

If Dawkins’ original latched, then he made a silly mistake, because obviously biology doesn’t latch, and he’s never claimed that it does.

Why then does he call it cumulative selection?

If Dawkins’ original latched, then he made a silly mistake, because obviously biology doesn’t latch, and he’s never claimed that it does.

I don't think it's at all obvious that biology doesn't latch. Without latching, how do you get nested hierarchies? HOw do you get consistent embryological development? Mung

Elizabeth Liddle:

You certainly don’t need to latch to get a program that will produce produce Weasel in a matter of minutes, and you don’t need to manipulate the mutation rate either.

The mutation operators have to be tuned to get the desired results. You really should know this. It's just like Dembski's rejection region ;) The WEASEL string contains 28 characters. If you mutate every one of the 28 characters every generation what is the chance that you're going to find a match? So you have to adjust the rate downwards to a certain level or you have no realistic hope of finding the target. We can of course put this claim to the test using one of the available programs. http://evoinfo.org/weasel Mung

That reply was with regards to the “random generator”, not latching.

ok, thanks.

Biology doesn’t reduce the mutation rates either the closer it gets to a fitness goal, as the Rosetta Code does, thereby effecting a latch “through the backdoor”, as it were.

Was this in the segment of code you posted earlier? Good catch. Mung

Liz: "Not only that, but the solutions, so far from being pre-specified, actually surprised the people who set up the GA."

I could write a simple program on my PC that would spit out random words of four characters without any latching (whether explicit or implicit) and print out the results. I may not have specified "MIKE" in any fitness function in the program, but if I let it run for a few days I would probably see "MIKE" output in the results. How surprising! I "never expected" that. But that same program would never output "METHINKS IT'S A WEASEL" in a million years. The search space is small enough to produce "MIKE" (even if the program isn't rewarding advances toward it) but odds are that it's never going to produce the Weasel string. Search space is too large to reasonably expect that. Moreover, I can provide a gap-free history of how "MIKE" was produced by the algorithm. And provide math to show how within the timeframe, how likely it was that "MIKE" would be produced. Now, what would *really* surprise me is if some of the output characters were Chinese Kanji, since A) my program was not capable of producing them, and B) my printer was not capable of rendering them. Now, if I came across this computer and printer, and knew little of the algorithm, and someone handed me a print out with Kanji characters on it and told me that the program and printer were producing these, I would have to be skeptical. You get my point? That antennae producing GA program you mention probably produced lots and lots of crappy results too. And also some good ones and very good ones. Some of the good ones "surprised" the engineers. But they programed the algorithm in the first place with certain parameters and with a quantifiable search space. They may be "surprised" when the see certain result (since they weren't looking for them), but the results never exceeded the predetermined search spaces or the capabilities of the GA system. And they actually ran empirical tests! If one engineer asked another to prove that a particular design came from this GA system s/he could demonstrate that it did, and provide a gap-free history of it's production. And the math to back it up. How does that compare with what we know about biology Is the known DNA replicator, and the known sources of stochastic variations + natural selection capable of covering the search space required to make a human eye at some point? Nobody knows. mike1962

Liz: On the other hand, programs that produce good radio antennae are very like biological evolution because they actually result in a functional piece of kit that is even directly analogous to the sense organs sported by some organisms. Not only that, but the solutions, so far from being pre-specified, actually surprised the people who set up the GA.

Indeed. So, are the known sources of stochastic variations + natural selection (the "fitness rewarder") capable of covering the search space required to make a human eye? Nobody knows. mike1962

Liz: If Dawkins’ original latched, then he made a silly mistake, because obviously biology doesn’t latch.

Biology doesn't reduce the mutation rates either the closer it gets to a fitness goal, as the Rosetta Code does, thereby effecting a latch "through the backdoor", as it were. mike1962

Mung, The full quote is this:

Liz: “Well, a powerful enough self-replicator. The random generator doesn’t have to be intelligently designed.” Me: The one you use and the one Dawkins uses is, and is fitting for the kind of fast result your program is designed to produce.

That reply was with regards to the "random generator", not latching. At any rate, I'd like to see Liz's code. I suspect it implicitly latches in the same way as the code on that page you cited. mike1962

You certainly don't need to latch to get a program that will produce produce Weasel in a matter of minutes, and you don't need to manipulate the mutation rate either. If Dawkins' original latched, then he made a silly mistake, because obviously biology doesn't latch, and he's never claimed that it does. It would be a weird mistake, though, because, as Mung says, it's not only unnecessary, but more complicated to do. And, as we can see from the published versions, many don't and still work. I won't bother to re-write one, because I think the point is made. And yes, it's not a good model of incremental functionality, and was never intended to be. My own, which "rewards" pronouncable strings, whole words and grammatical word combinations is somewhat better, but language based algorithms are a very poor model of biology because proteins are not very like text, and evolutionary process are very little like the processes by which we produce it. On the other hand, programs that produce good radio antennae are very like biological evolution because they actually result in a functional piece of kit that is even directly analogous to the sense organs sported by some organisms. Not only that, but the solutions, so far from being pre-specified, actually surprised the people who set up the GA. Cheers Lizzie Elizabeth Liddle

Mung: I think you need to recognise the point above, that the evidence of the 1986 runs is that the O/P was latched. The evidence in hand is that this can be achieved by writing in latching code explicitly, or by so manipulating the rates and generation sizes etc that on "good" runs you get a latched o/p; even absent specific explicit latching code. Dawkins' published runs in the literature in 1986, show IIRC 200+ "advances" and no reversions. He doubtless selected what at the time looked like "good" runs, and published them. That latching of the o/p whether achieved explicitly or implicitly, serves to underscore the REAL problem, that the code rewards increments in mere proximity to target, regardless of the non-functionality of the "nonsense" phrases. This is not a good model of differential success on improved functionality achieved by chance variation, and inadvertently points to the problem of getting to the shores of islands of function in large config spaces. By side=tracking into a debate over explicit vs implicit latching, we are allowing Darwinists to get away with begging that very big question. How big is it? let's just say that he concept of complex, specified information pivots on the issue of finding special, narrow and separately describable zones on searching large config spaces without intelligent direction of the search. GEM of TKI kairosfocus

I prefer to not argue with design proponents, but at times I think it is prudent to do so. We need to have integrity. I try not to have a double standard. Mung

Me: mike1962's accusation was that Lizzie’s version incorporated latching. mike1962: No, actually, I said Dawkins’s does. So let's review: mike1962 @160:

Dawkin’s “weasel” string is highly specified, and the latching mechanism is very simple compared to the search string, i.e, the “context” of selection.

Elizabeth Liddle @163:

Ah. If there was a “latching mechanism”, then I agree, it wasn’t Darwinian.

Elizabeth Liddle @163:

But was there? I’ve written one, and I certainly didn’t latch. They are dead easy to write. It wouldn’t be much fun if you latched, anyway.

Elizabeth Liddle @166:

But I don’t think Dawkins’ one latched, and all the copies I’ve seen didn’t, and the one I wrote myself didn’t.

mike1962 @197:

Let’s see your code. The one you use and the one Dawkins uses is, and is fitting for the kind of fast result your program is designed to produce.

Mung

Mung: The issue of latching is therefore moot.

I disagree. Whether it's done explicitly, or by subterfuge (in these cases, but decreasing the mutation rate) the effect is the same. They effect latches by decreasing the mutation rate. A latch is a latch thru the front door or the back. Leave the mutation rate at the initial value and see how far it gets. mike1962

kairofocus: 3: THE PUBLISHED RUNS C 1986 SHOW WAS IT 200+ ADVANCES AND ZERO REVERSIONS.

Yep. This is why I said Dawkins's version latched. (Even though I've never seen the code.) I didn't say Lizzie's latched. Although, if it's analogous to the example Mung pointed us to, it does latch in a fuzzy way, still subject to more vanishing probabilities of mutation. mike1962

mung: How many of those versions of the “weasel” program incorporate latching?

It's not done directly, but via the "backdoor" by decreasing the mutation rate, and whenever a random value falls under the rate, which is more and more does as the mutation rate decreases, notice how the parent (*parptr) character is copied. This in effect is a "fuzzing" latching, I would say. void mutate(TgtString kid, TgtString parent, float mutateRate) { char *cp; char *parptr = parent; for (cp = kid; *parptr; cp++, parptr++) { *cp = (frand() < mutateRate)? randChar() : *parptr; } *cp = 0; } mike1962

mung: mike1962?s accusation was that Lizzie’s version incorporated latching.

No, actually, I said Dawkins's does. mike1962

mike1962's accusation was that Lizzie's version incorporated latching. He has no evidence that this is in fact the case. The failure of Lizzie to present her version for inspection proves what? Latching is not required for a "weasel" style program to succeed in finding the target phrase. The issue of latching is therefore moot. See my link @201. How many of those versions of the "weasel" program incorporate latching? Mung

Mike1962: 1: Explicit latrching is of course a well known implementaiton of Weasel. 2: Implicit latching is a demonstrated fact. 3: THE PUBLISHED RUNS C 1986 SHOW WAS IT 200+ ADVANCES AND ZERO REVERSIONS. 4: best explanation: these were "good" runs, which implicitly latched. 5: In any case, Dawkins long since confessed to the root problem, targetted search on mere proximity for NONSENSE phrases, the latching debate was a side-issue meant to be the red herring led away to the ad hominem soaked strawman. As usual. 6: UD long ago showed what seems to have been a credible more or less original Weasel and it could latch implicitly. GEM of TKI kairosfocus

Yes, it would take more lines of code to latch than not.

And I don't know what I'm talking about ;) (I'm not making reference to a claim you made.) Mung

As a cognitive neuroscientist, I’m very interested in how intelligent design works, so I certainly don’t exclude it a priori!

http://www.amazon.com/gp/product/1592535879 http://www.amazon.com/gp/product/0750660775 http://www.amazon.com/gp/product/0750664487 http://www.amazon.com/gp/product/1847886361 Mung

Exactly Mung. And thanks :) Yes, it would take more lines of code to latch than not. Elizabeth Liddle

There's even a matlab version! Mung

OMG! Lizzie, my link trumps yours any day. ;) http://rosettacode.org/wiki/Evolutionary_algorithm Mung

p.s. I am willing to take your word for it! Mark this day! Mung

Lizzie, I don't care whether your code latches or not :) Say a string matches more closely than any other the target phrase. That string will be preferentially selected and more copies of it will appear in the next generation. Unless the mutation rate is too high, chance favors that the matching letters will not be changed. So one can control the "appearance of latching" (haha) by choosing an appropriate mutation rate. Frankly I think it takes more programming effort (offering my opinion as a programmer) to introduce a latching mechanism than not. So I consider it more likely that there isn't one in your program nor was there one in Dawkins' program. He's not a programmer. Mung

Mike1962: I had a look for mine, but I must deleted the folder. However, there are lots here: http://like-a-weasel.blogspot.com/ Re your last point: "The scientific methodology you apparently are happy to hang you hat on excludes design a priori" - no it doesn't. Ask any forensic scientist :) Or me. As a cognitive neuroscientist, I'm very interested in how intelligent design works, so I certainly don't exclude it a priori! Elizabeth Liddle

Elizabeth Liddle:

"But I don’t think Dawkins’ one latched, and all the copies I’ve seen didn’t, and the one I wrote myself didn’t."

Let's see your code.

"Well, a powerful enough self-replicator. The random generator doesn’t have to be intelligently designed."

The one you use and the one Dawkins uses is, and is fitting for the kind of fast result your program is designed to produce. The ones found in nature are not analogous. And it is unknown when they are powerful enough along with natural selection to produce novel cell types, tissue types, organs or body plans.

"No. You don’t need a “selection mechanism”. All you need is a fitness function,"

What's the difference?

"you can generate that randomly as well."

Does nature have such a thing powerful enough to produce novel cell types, tissue types, organs or body plans. Nobody knows. You can believe it does, or wish it does, or "assume it does for the sake of parsimony, Occam's Razor" or whatever else. But the hypothesis is undemonstrated. I'm still not getting on that airplane.

"But simple changes are all we need. No-one is postulating radical changes, except as the accumulation of simple changes over time."

However, it is unknown if the known "simple" variations that exist in nature plus natural selection is powerful enough to produce novel cell types, tissue types, organs or body plans. Do the chemical/mechanism pathways exist for this sort of blind incremental change to produce the aforementioned? Nobody knows. It's an open question.

"The reason is a simple scientific principle, which is parsimony, otherwise known as Occam’s razor."

Occam's Razor is a guiding principle of science. Occam's razor is not a license to avoid demonstrating your hypothesis. The Modern Synthesis is the best anti-design materialist hypothesis going. Nobody is arguing with that. But that's not saying much at this point.

"It has nothing to do with “materialistic faith”. It is simply to do with scientific methodology: "

The scientific methodology you apparently are happy to hang you hat on excludes design a priori, and seems to be happy to extend the inference into very deep uncharted waters when the hypothesis has only been proven in the most shallow of puddles. No sale. mike1962

William, I have no idea what you're talking about. I offered a clear test that can be done. You're saying that I might move the goalposts and therefore my proposed test doesn't prove anything? Mung

p.s. I could be mistaken about whether that provides a reproductive advantage. That may not change the ratio of sip/instruction. So it could be that they award the additional sip so that it won't be out-reproduced by organisms with shorter genomes. But isn't that still stacking the deck? So here what we can do, if that is the case. Don't reward sips based on genome length. That would ensure differential reproduction wouldn't it? Also, my original test was that we would only disable the rewards for the simpler functions, not all logic functions. So if they evolved a complex function they would still out-compete. Mung

Mung,

not that they can evolve them by a strictly Darwinian process?

What, you mean that the computers themselves need to breed and evolve themselves first? Next you'll be saying that you can only simulate parachute drops by dropping the computer running the simulation out of a plane! If you can't tell the difference between "designed to evolve" and "evolved" does that not tell us something striking about the very idea of "designed to evolve"? About how very unnecessary it is? So is it your position that only the first replicator was designed and all biological diversity is natural from that point on? WilliamRoache

Am I mistaken in believing that this is the same as saying removing the advantage of possibly beneficial mutations in a cell?

The avida organism is given two means by which it can attain additional processing resources. Finding a logic function is only one of them.

Digital organisms compete for the energy needed to execute instructions. Energy occurs as discrete quanta called ‘single-instruction processing’ units, or SIPs. Each SIP suffices to execute one instruction. By executing instructions, a digital organism can express phenotypes that enable it to obtain more energy and copy its genome. In Avida, organisms can acquire energy by two mechanisms. First, each organism receives SIPs in proportion to its genome length. Second, an organism can obtain further SIPs by performing one- and two-input logic operations on 32-bit strings (Supplementary Information).

So all an organism needs to do to gain a reproductive advantage is evolve a longer genome. And even what you say were true, doesn't that still make my same point, that they were designed to evolve complex functions, not that they can evolve them by a strictly Darwinian process? Here's another point to consider. Does having a longer genome increase an organisms chance to evolve the complex logic function? If we remove that part of the simulation and left everything else as is, what would happen? Mung

UTE: Go read Dawkins' own weasel words, Correct letters are being preserved because they move nonsense -- non-functional -- phrases closer to a pre-loaded "distant target." He himself admits this is misleading. Indeed, Weasel is a case of intelligent design. Other EAs and GAs have similar but subtler failings. GEM of TKI kairosfocus

*removing = remove ute

Mung: '2. Remove the rewards for “evolving” simpler logic functions.' Am I mistaken in believing that this is the same as saying removing the advantage of possibly beneficial mutations in a cell? ute

Doesn't the search need to be targetted in some ways to mimick natural selection. So for example, in the weasel program, a letter is locked in because it is mimicking a beneficial mutation that becomes fixed. TBH, I have no idea how avida works, but doesn't the same idea apply? ute

DrBot Quote-Mine DrBot:

Generally speaking any search has to have a target or it isn’t a search …

Mung:

Thank you, thank you, thank you. I can’t believe how often I have to argue over this very simple fact (MathGrrl comes to mind).

DrBot:

I said: Generally speaking any search has to have a target or it isn’t a search, but ‘targeted’ means something more specific – pre-specified and unchanging

Actually, you took yourself out of context, did a little quote-mining as it were. What you fully wrote (@28) was:

Generally speaking any search has to have a target or it isn’t a search, but ‘targetted’ means something more specific – pre-specified and unchanging (I believe – but I could be wrong)

Oh my. Mung

On the Evolution of Complexity via a Darwinian Process Elizabeth Liddle:

But the argument that once Darwinian process have got going they can’t generate further functional complexity is simply falsified by programs like Avida.

I say the claim is not falsified by Avida. Other programs like Avida I am not aware of. Avida:

The handwritten ancestral genome was 50 instructions long, of which 15 were required for efficient self-replication; the other 35 were tandem copies of a single no-operation instruction (nop-C) that performed no function when executed.

Elizabeth Liddle:

What we can do is test the hypothesis that Darwinian processes, can, for example, generate novelty, complexity, function, etc, given self replicating entities that reproduce with variance, and where that variance results in differential reproduction.

I propose such a test here: 1. Remove the 35 nop-C instructions we'd still have a functional Darwinian replicator. 2. Remove the rewards for "evolving" simpler logic functions. Would Avida still evolve a complex logic function? I seriously doubt it. I say no. There's no reason my claim cannot be falsified. Yet Darwinian processes would still be in operation. What does this tell us about Darwinian processes? What does it tell us about what role Darwinian processes actually play in Avida? Here's what it says to me, an admitted critic. Avida fails to demonstrate that a Darwinian process, operating alone, can evolve complexity. The only reason that it currently does so, is because it is designed to do so. My claim can be falsified. Mung

Final post of the night:

I do not see why a “purposeless, mindless process” should not produce purposeful entities, and indeed, I think it did and does. - Elizabeth Liddle

I assume by "a purposeless, mindless process which can and did produce purposeful entities" you're speaking of the same "Darwinian process" we're talking about in this thread? I guess that rules out computer programs which are written with a purpose in mind. Like Weasel. Like Avida. Mung

Are GA's "Darwinian?" Elizabeth Liddle:

I’d say it’s the very limitations of Darwinian processes (lack of foresight; inability to transfer solutions from one lineage to another) that make its most powerful differential predictions compared to what you might expect from an Intentional Designer (as exemplified by human designs).

EA's which employ crossover transfer solutions from one lineage to another. Therefore it follows that EA's which use crossover are non-Darwinian. The chromosome/genome in most EA's don't self-replicate. See above. Mung

Is Dawkins' Weasel Program "Darwinian?" Elizabeth Liddle:

I use “Darwinian processes” to denote, strictly, processes in which self-replicators replicate with variants, and in which the variance confers different degrees of reproductive success.

The strings in Dawkins' Weasel program do not self-replicate. If follows that according to the definition given Weasel is non-Darwinian. Mung

Is "latching" non-Darwinian? We are told that Darwinism predicts a nested hierarchy. For this to be true, don't certain features have to be "latched" or retained and then shared in common by future descendants? Isn't a "latching mechanism" a requirement of the theory? I think the answer is yes. It follows that the claim that the presence of a "latching mechanism" would be non-Darwinian is not true. If a process fails to produce a nested hierarchy, does it follow that it is non-Darwinian? Mung

Interesting...

Even Intelligent Design could be Darwinian, if the Designer merely pre-screened (or designed) the variants, and let the resulting differential reproduction run its course. Darwin did not actually propose a variance-producing mechanism. However, I think this would be stretching things a bit.

Unless Darwinist are now attempting to redefine themselves yet again after a century of failures, this twist makes little sense. Having just seen Nullasus's response, my original would be mostly a repeat. That "assumed" Darwinian mechanisms are easily utilized by Designers. Nullasus said,

In other words, the options here are not ‘evolution or design’, because evolution itself is entirely capable of being subsumed under design in whole or in part. It’s not even ‘selection and variation versus design’, because intelligent agents can both select, and orchestrate variation.

Well said and agreed. Darwinism has always been an unguided process. If not, then there is no purpose for the theory. But Design can have subsets of variation, to the point of extinction.

I use “Darwinian processes” denote, strictly, processes in which self-replicators replicate with variants, and in which the variance confers different degrees of reproductive success.

That is a limited definition of Darwinism w/o explanation of micro steps to macro complexity. Reproductive success does not indicate growing complexity like Dawkins Weasel program, which created specified complexity. The end goal, the target compared to by a search, does not exist in true Unguided processes and is not recognizeable. Survival is not a target for a blind process. Reproductive success is not a target. No matter how much a Darwinist wants to make up just so stories. Survival and fitness have no reality in blind processes.

"Weasel is Darwinian. It just has an extremely simple problem to solve and that problem has only one solution."

Wrong, it is an extreme account of disinformation to the public at-large. Weasel is not Darwinian, it is Design by an intelligent mind. What we observe today is variation, drift, and entropy. We do not observe creation of higher complexity, by unguided gradual processes. We do not observe whale tales or bear tales by Darwin, or any new fictionalized accounts told by Dawkins. But, Genetic Entropy is being confirmed... Dr. John Sanford's Genetic Entropy DATCG

Elizabeth Liddle:

But the argument that once Darwinian process have got going they can’t generate further functional complexity is simply falsified by programs like Avida.

I hope you realize that if Avida is non-Darwinian then it follows that the claim has not been falsified. Also, if the ability to generate functional complexity is not inherent in the "Darwinian process" itself that is used by Avida, again, the claim is not falsified. Mung

WEASEL - A targeted goal, compared by an intelligently designed program to reach that end goal is "Darwinian." Who knew? Since when did Darwinist accept programmatic Design by intelligence as Darwinian? Dembski, et al., dealt with Dawkins program here... Designed Outcomes Dawkins improved upon Darwins orignal bear to whale story. A fiction quickly pulled by Darwin after embarrassment. The contemporary version is still fictionalizing science. Turning it into a prophetic movement, more than scientific inquiry for public understanding. As for AVIDA, thats a failure in logic as well. Introducing a target no matter how complex a derivation scheme is still intentional. Therefore intent modifies the outcome, not blind processes. It is still using Active Information... AVIDA Stair Step Active Information to intelligently guide an outcome. Simulating a prophesied inefficiency does not prove a theory. It is a more complex Weasel program designed by programmers better than Dawkins, albeit, inefficiently. It does not simulate any observed Macro mechanism in nature that we know of, only fictional accounts that Darwinians think may have happened based upon assumptions, fictional guesses and a theory that has continued to produce failed predictions, such as vestigial organs and "junk" DNA. What AVIDA shows is Design works. DATCG

I think design is a perfectly legitimate domain of scientific enquiry, and in principle detectable in patterns through well-designed (heh) hypothesis testing. That's nice. I disagree in part, but I'll put that aside for now. But what I'm saying here is not that 'inferring design is a legitimate domain of scientific enquiry'. I'm saying that inferring non-design, non-purpose, non-guided - clearly to the extent it has been in contemporary evolutionary biology - is scientifically unsupported, not demonstrable, and largely the result of metaphysics rather than science. Further, my point wasn't merely that some organisms are 'designed'. It's that evolutionary mechanisms, including many supposedly 'Darwinian' mechanisms, are not only entirely capable of being used by an intelligent agent, but already have been shown to be in particular cases. Evolutionary mechanisms are yet more tools for intelligent agents. In other words, the options here are not 'evolution or design', because evolution itself is entirely capable of being subsumed under design in whole or in part. It's not even 'selection and variation versus design', because intelligent agents can both select, and orchestrate variation. nullasalus

Sorry, my post above was to Mung. Nullasalus: I should make it clear that I have no problem with Design as a hypothesis. As you point out, some organisms we know are, partly "designed", e.g. by Monsanto, or Craig Venter, or even John Sanford! I think design is a perfectly legitimate domain of scientific enquiry, and in principle detectable in patterns through well-designed (heh) hypothesis testing. I just thought I'd make that clear. Anyway, it's past my bedtime - see you guys later :) Elizabeth Liddle

Good question: They are "conserved" because they replicate consistently better than any variant of them. In other words, they are very un-robust - all but slight changes have disastrous consequences. So they are still vulnerable to mutations, but the mutated versions don't propagate through the population. Other sequences are far more robust to changes - many polymorphisms function just fine, and indeed provide the population with the allelic diversity that enables it to adapt to changing environments. But as you can imagine, something like a hox gene will be highly conserved (and is) because if you mess up which end is head and which end is tail you aren't likely to get a viable organism! Elizabeth Liddle

Well, in Fisherian hypothesis testing that’s exactly what it means. Wonderful. Inappropriate in this context. Exactly. Formulating your null is absolutely crucial. Without a clearly formulated null, we cannot make any successful inference. Except the 'unguided and purposeless' position is not clearly formulated, and not all inferences are scientific inferences besides. No, it certainly is not demonstrated. Indeed. What we can do is test the hypothesis that Darwinian processes, can, for example, generate novelty, complexity, function, etc, given self replicating entities that reproduce with variance, and where that variance results in differential reproduction. Except A) we're unable to test the most crucial part of the 'Darwinian processes' under discussion here - the lack of guidance or goal, either in past or present evolutionary scenarios, and B) insofar as we can demonstrate the capabilities of those processes, we can likewise demonstrate the capability for intelligent input at each and every stage to generate desired results as well. To put it another way, whether or not "unguided and purposeless processes" in the relevant sense even exist is outside of science, and empirically undemonstrable. But the existence of guided and purposeful processes, at least on a certain level, is not only demonstrable, but their ability to make use of evolutionary processes is also demonstrated. So that means we have a candidate mechanism for the generation of novelty etc in a system that reproduces with variance etc. And the candidate mechanism isn't the concern here. The mechanisms in question can be either guided or unguided, purposeful or non-purposeful. But we only have evidence of the purposeful and guided existing. The unpurposeful and unguided? It's metaphysical assumption, and superfluous. And we cannot conclude that those gaps are not filled with something extraordinary like a God. However, nor can we conclude that they are. A) The problem here isn't merely the gaps, but the conception of the mechanisms, period. Attempts to fill in the gaps with the unguided and unpurposeful is an exercise in metaphysics, not science. B) The intelligent agent need not be God, and what is or is not extraordinary is in large part that of opinion anyway. There are, for example, salmon who grow faster now. Determining that the salmon did not come about by an unguided, purposeless process does not require positing God - it requires positing Monsanto. C) We can certainly infer that Monsanto had a role. In principle, we can infer other intelligences as well. Can we demonstrate it beyond a doubt? No, but that's not necessary for science anyway. Will there be scientists who disagree? Probably, but consensus isn't necessary either. And what's really not necessary is the assumption 'it's unguided and without purpose'. nullasalus

Elizabeth Liddle:

If there was a “latching mechanism”, then I agree, it wasn’t Darwinian.

First, I don't doubt that your program did not have a latching mechanism and I think someone can write a "weasel" program that functions without one. But second, why would such a thing be non-Darwinian? What is the explanation for conserved sequences other than that they have been "latched?" Mung

Probably, somewhere. It's in MatLab. I'll see if I can find it. I wrote a cooler one, though, that evolved sort of daft English sentences, by rewarding pronouncable syllables and English words from a lexicon. I even had some grammar in there - extra reward for noun verb noun and article-noun. I'll try to dig it out. Elizabeth Liddle

Elizabeth Liddle:

But I don’t think Dawkins’ one latched, and all the copies I’ve seen didn’t, and the one I wrote myself didn’t.

Hi Lizzie, do you still have a copy of the program you wrote? Mung

Nullasalus:

“Retaining the null” is not the same as saying “science can’t determine or infer an answer here”.

Well, in Fisherian hypothesis testing that's exactly what it means. However, Fisherian hypothesis testing isn't the only way of testing a hypothesis.

In the latter case, there is no ‘null’ – there’s no “scientific position on what science can’t determine”. And just what the null should be in science, particularly in this context, is controversial itself.

Exactly. Formulating your null is absolutely crucial. Without a clearly formulated null, we cannot make any successful inference.

And it goes without saying that “blind and purposeless” is itself not demonstrated. It is, at best, an assumption – and a metaphysical one at that.

No, it certainly is not demonstrated. To demonstrate it you'd have to set it up as an alternative hypothesis to something, and that would be very difficult to do, because you'd have a heck of lot of trouble formulating your null. That's why scientific hypothesis are generally far more specific. They may arise from theories, but they refer to a very specific aspect of that theory. For instance, we can't hope to demonstrate that the whole of life can be accounted for by Darwinian evolution. What we can do is test the hypothesis that Darwinian processes, can, for example, generate novelty, complexity, function, etc, given self replicating entities that reproduce with variance, and where that variance results in differential reproduction. So that means we have a candidate mechanism for the generation of novelty etc in a system that reproduces with variance etc. And so on. But it's an iterative process, and no scientific theory is ever complete. There will always be gaps :) And we cannot conclude that those gaps are not filled with something extraordinary like a God. However, nor can we conclude that they are. Elizabeth Liddle

That’s why we often “retain the null”. Retaining the null doesn’t rule out something else, it just means we don’t consider the something else has been demonstrated. "Retaining the null" is not the same as saying "science can't determine or infer an answer here". In the latter case, there is no 'null' - there's no "scientific position on what science can't determine". And just what the null should be in science, particularly in this context, is controversial itself. And it goes without saying that "blind and purposeless" is itself not demonstrated. It is, at best, an assumption - and a metaphysical one at that. nullasalus

Nullasalus:

And in cases where science is incapable of making a determination either way, the proper result is not to go with what you personally think is simpler – but to say “science can’t determine or infer an answer here”.

I absolutely agree :) That's why we often "retain the null". Retaining the null doesn't rule out something else, it just means we don't consider the something else has been demonstrated. Elizabeth Liddle

Well, no. The reason is a simple scientific principle, which is parsimony, otherwise known as Occam’s razor. Occam's razor is a philosophical principle first and foremost, not a "scientific principle". Its application in science in general, much less this particular situation, is debated in numerous ways - and whether it's simpler to believe in blind, unguided processes (the very existence of which, in a relevant sense, is not demonstrated by science nor can it be) versus inferring intelligence (the existence of which, broadly, is known) is a subject of debate itself. And in cases where science is incapable of making a determination either way, the proper result is not to go with what you personally think is simpler - but to say "science can't determine or infer an answer here". nullasalus

Mike1962:

The putative process is “compatible” or “consistent” with what is observed thus far. Sure.

Cool :)

And what you mean is that there is no contradiction. Whoopee. But it’s explanatory power is practically nil for all except some very simple changes to earth’s biological objects.

But simple changes are all we need. No-one is postulating radical changes, except as the accumulation of simple changes over time.

Go ahead and make the leap of faith if you want. I won’t, because there is no proof of concept to the scale of functional complexity that exists. It’s an open question. The only reason in the world why anyone would “take a side” is because of an a prior philosophical commitment to non-intelligent causation.

Well, no. The reason is a simple scientific principle, which is parsimony, otherwise known as Occam's razor.

Sorry, no sale. I’m not such a bigot as that.

I'm sure you aren't, but then nor am I.

We understand your position on this. Have at it. No one is going to burn you at the stake. But it’s not persuasive to those of us who have no desire to “drink the coolaid” of materialistic faith that people such as you exhibit.

It has nothing to do with "materialistic faith". It is simply to do with scientific methodology: do not multiply entities unnecessarily; an explanation should be as simple as it needs to be, but no simpler. That doesn't prevent even the scientist having faith in something beyond the simplicity of an explanation that works. Elizabeth Liddle

PS: the "Can you?" is Mike's. Sorry! Elizabeth Liddle

oops, missed the last part of your post: <blockquote:At best, as I said before, Dawkin’s Weasel program shows that small changes can be accumulated if you have a powerful enough (intelligently designed) random generator Well, a powerful enough self-replicator. The random generator doesn't have to be intelligently designed. Better if it isn't. If you can feed in atmospheric static, that's best of all.

and a powerful enough (intelligently designed) selection mechanism

No. You don't need a "selection mechanism". All you need is a fitness function, and as is said, you can generate that randomly as well. But obviously if you do in fact want to solve a problem for your own benefit (like writing a nice bit of useful code) then you will design your fitness function. But it doesn't actually have to be designed.

that are matched for a certain purpose (matching to the search string.) Everyone knows this. The weasel program makes no case for a blind evolution system unless you can FIRST demonstrate that the randomization source and the selection source of earth’s life is NOT intelligently designed.

The blind part is the random trial part. That's it's important not to latch. If you latch, it's not random, and your mutations can only get better, not worse. That's not "blind" - it's biased in favour of what works. But I don't think Dawkins' one latched, and all the copies I've seen didn't, and the one I wrote myself didn't. Can you? Elizabeth Liddle

Elizabeth L: We have a process that is compatible with what we see, so we need not postulate any other.

The putative process is "compatible" or "consistent" with what is observed thus far. Sure. And what you mean is that there is no contradiction. Whoopee. But it's explanatory power is practically nil for all except some very simple changes to earth's biological objects. Go ahead and make the leap of faith if you want. I won't, because there is no proof of concept to the scale of functional complexity that exists. It's an open question. The only reason in the world why anyone would "take a side" is because of an a prior philosophical commitment to non-intelligent causation. Sorry, no sale. I'm not such a bigot as that. We understand your position on this. Have at it. No one is going to burn you at the stake. But it's not persuasive to those of us who have no desire to "drink the coolaid" of materialistic faith that people such as you exhibit. mike1962

Well, it is a low bar, Nullasalus. That’s why the Design inference is very weak (at least with regard to the Origin of Species, not with regard to the Origin of Life). I agree that the inference to evolution being blind and purposeless is very weak indeed. The inference to it not being blind and purposeless is quite a bit stronger. We have a process that is compatible with what we see, so we need not postulate any other. And this process is entirely compatible with evolution being directed, guided, and purposeful. Indeed, that general - and broad - claim is far more supported than 'blind' or 'non-purposeful'. The only support present, and even possible, for such is (metaphysical) assumption. And those assumptions aren't necessary. We could, but we need not. 'Need'? 'Need' rarely shows up in science - you can always draw epicycles if you wish, or appeal to merely needing more time to properly explain what we see, etc. As it stands, a variety of factors - some philosophical, some empirical - give great warrant to concluding the evolution is, after all, a guided, purposeful process. That someone can nevertheless insist that it's blind and purposeful doesn't keep me up. If the desperate need to believe in blindness and lack of purpose to keep their warm fuzzies, they're welcome to. ;) nullasalus

Mike1962

Elizabeth L: So yes, if you call “solving the presented problem of survival” as “having foresight”, then Darwinian processes have foresight.

Dawkin’s “weasel” string is highly specified, and the latching mechanism is very simple compared to the search string, i.e, the “context” of selection.

Ah. If there was a "latching mechanism", then I agree, it wasn't Darwinian. But was there? I've written one, and I certainly didn't latch. They are dead easy to write. It wouldn't be much fun if you latched, anyway. But if you are talking about a latching program sure. I agree.

If you think this is analogous to biological life you’re going to have to demonstrate that the biological randomization mechanism(s) of earth are sufficient to generate enough trials such that the context of the environment of earth is capable to “matching” it to it’s “predetermined goal” of “fitness.”

Well, no, it doesn't have to be "predetermined" at all. In life, of course, it is constantly changing. I've done sims like that (very simple ones anyway) where the fitneess criteria fluctuate randomly, and your evolving population have to keep adapting to the new environment. You can even do them so that the evolving population itself is part of the environment (by introducing competition, for instance).

Of course, you can’t.

Can't what? Make life-sized model? No, you are right, we can't. But that's not the purpose of a model.

Something to remember: Dawkin’s simulation has a PRE-DETERMINED GOAL THAT WAS SET BY AN INTELLIGENCE

But that is no difference to a very simple natural scenario in which there is a single solution to survival.

PLUS A RANDOM GENERATOR DESIGNED BY AN INTELLIGENCE FITTING TO THE GOAL.

Yes. But now you are talking about abiogenesis not Darwinian evolution. Darwinian evolution can't, by definition, account for the simplest organism capable of Darwinian evolution.

Do you think earth’s biological system is analogous to that? If so, demonstrate it.

No, and I've said why. But I do agree that Darwinian evolution doesn't latch. At best, as I said before, Dawkin’s Weasel program shows that small changes can be accumulated if you have a powerful enough (intelligently designed) random generator and a powerful enough (intelligently designed) selection mechanism that are matched for a certain purpose (matching to the search string.) Everyone knows this. The weasel program makes no case for a blind evolution system unless you can FIRST demonstrate that the randomization source and the selection source of earth’s life is NOT intelligently designed. Can you? Elizabeth Liddle

Well, it is a low bar, Nullasalus. That's why the Design inference is very weak (at least with regard to the Origin of Species, not with regard to the Origin of Life). We have a process that is compatible with what we see, so we need not postulate any other. We could, but we need not. The null, as I keep explaining to Mung, is retained :) Elizabeth Liddle

To continue... Edit last sentence: The weasel program makes no case for a blind evolution system AT ALL. mike1962

Elizabeth L: So yes, if you call “solving the presented problem of survival” as “having foresight”, then Darwinian processes have foresight.

Dawkin's "weasel" string is highly specified, and the latching mechanism is very simple compared to the search string, i.e, the "context" of selection. If you think this is analogous to biological life you're going to have to demonstrate that the biological randomization mechanism(s) of earth are sufficient to generate enough trials such that the context of the environment of earth is capable to "matching" it to it's "predetermined goal" of "fitness." Of course, you can't. Something to remember: Dawkin's simulation has a PRE-DETERMINED GOAL THAT WAS SET BY AN INTELLIGENCE PLUS A RANDOM GENERATOR DESIGNED BY AN INTELLIGENCE FITTING TO THE GOAL. Do you think earth's biological system is analogous to that? If so, demonstrate it. At best, as I said before, Dawkin's Weasel program shows that small changes can be accumulated if you have a powerful enough (intelligently designed) random generator and a powerful enough (intelligently designed) selection mechanism that are matched for a certain purpose (matching to the search string.) Everyone knows this. The weasel program makes no case for a blind evolution system unless you can FIRST demonstrate that the randomization source and the selection source of earth's life is NOT intelligently designed. Can you? mike1962

mike1962, but it does not demonstrate *Darwinian* (blind, purposeless) evolution. It's not possible to demonstrate that, which is part of the problem here. At least not in a meaningful way, with regard to ID, TE, etc. You can demonstrate a general process, sure. You can identify mechanisms of both variation and selection. That the actual instantiated process is purposeless? Not demonstrable by science. Blind, in the sense of the variation and selection was not intended or guided by any mind? Again, not demonstrable. You can, however, at times demonstrate when, how, and to what proximate degree the process is guided. (Talk to the programmer, make inferences about such and such being programmed or purposeful, etc.) Doubly so in the context of simulations, where you can do everything from bias the mutations to intervene directly in the process with your eye on influencing the results. (I suppose these would be called miracles, eh?) The best you can get on the 'blind, purposeless' front is compatibility: "What I see is compatible with...!" But logical compatibility is a ridiculously low bar. nullasalus

I'm so sorry, Mung! If you said that, you were absolutely right IMO :) I hope it will be the first of many occasions when you, DrBot, and myself agree! Cheers Lizzie Elizabeth Liddle

As DrBot said, all searches are “targeted” or they wouldn’t be searches.

Credit where credit is due please. DrBot was just parroting something I wrote in a prior thread. Of course, I don't know what I am talking about, so ... Mung

Mike1962:

Weasel is Darwinian. It just has an extremely simple problem to solve and that problem has only one solution.

No it isn’t. Darwinian mechanisms have no “goal” or “foresight.” Targeted searches do. Weasel is a targeted search. You could say it’s demonstration of intelligent evolution.

As DrBot said, all searches are "targeted" or they wouldn't be searches. Of course Dawkins had a "goal" when he wrote WEASEL. In addition, the only solution to the problem of survival posed by the digital organisms in WEASEL is to spell out "methinks it is like a weasel". If there were two solutions, for example if both: "methinks it is like a weasel" and "I think she is looking east " then that would still be a "targeted search" in that as in the original, the "target" is survival in the current environment, but there are two solutions. If the solution to survival is any pronouncable string of pseudo words, then there are many solutions, but it is still a targeted search. If the solution to survival is a piece of code that forms a NAND gate, then there are many solutions, but it is still a targeted search. If the solution to survival is not being eaten by a predator, there are a vast number of solutions, but it is still a targeted search. So yes, if you call "solving the presented problem of survival" as "having foresight", then Darwinian processes have foresight. But what is usually meant by that phrase is that it is entirely trial and error - it doesn't pre-select likely solutions. It simply throws the stuff at the wall to see if it sticks. The wall is the target. What sticks is the solution. Now, what really would be a "targeted search" in a manner that would NOT, at least as far as we know it, resemble biological evolution, would be a search in which instead of "blind" trial and error, each partly successful solution was scrutinised by some additional process that analysed it for where it had gone wrong before devising the next one. This of course is how human designers do it. But with Darwinian processes (including Avida) this doesn't happen. Mutations are as likely to make things worse as better, and, in fact, more likely as the evolving "solution" gets better, and there are more ways to worsen it than improve it. It is in this sense that both WEASEL and Avida are NOT targetted. Both are as likely to throw worse solutions at the problem as better. In fact it's sort of frustrating watching WEASEL programs, because you get nearly all the letters there, and then, dammit, just as it gets the final one, it loses one it already has! Fun, though. Have you ever written one?

That’s why Dawkins wrote it.

He wrote it to demonstrate the idea of the accumulation of the selection of small modifications contra the idea of a whirlwind blowing thru a junk yard producing a 747, etc. It does demonstrate that,

Yes, it does.

but it does not demonstrate *Darwinian* (blind, purposeless) evolution.

Yes, it does :) See above. Elizabeth Liddle

Mung:

Hi Lizzie. Thanks for your post. This could explain some of the misunderstanding and give us something to work with. I asked: “Are all evolutionary processes Darwinian? Can an evolutionary process be non-Darwinian?” Is your answer YES to both those questions? Or can I take it that you were saying YES to the second question and that therefore your answer to the first question is NO? cheers

The second thing. Evolutionary processes can be non-Darwinian. An example would be genetic engineering. Drift is an important non-Darwinian mechanism, although you could argue that it is closely related, in that it involves differential reproduction - however the differential reproduction is independent of the phenotypic effects of the inherited genotype. Artificial selection is Darwinian, it's just that the environment that the critters have to survive in includes opinionated human beings with their own purposes. Even Intelligent Design could be Darwinian, if the Designer merely pre-screened (or designed) the variants, and let the resulting differential reproduction run its course. Darwin did not actually propose a variance-producing mechanism. However, I think this would be stretching things a bit. Elizabeth Liddle

Hi, ciphertext!

Per Elizabeth Liddle:

Often that code is creative, innovative and useful. But I’ll keep trying. For clarity, no I don’t mean that computers write their own code in general, but that it it is possible for code to write code, and for that code to be extremely creative, efficient, and functional.

What is meant by the terms “creative”, “innovative”, “useful”, “efficient”, and “functional”?

I meant them in the following senses: creative: uses solutions that, in a human being, we might call "lateral thinking". DrBot had a nice example above, about a faultchecker that checked its own faults. innovative: uses solutions that hadn't previously been proposed. useful: solutions are put to some practical human purpose, such as receiving better radio signals, or running a better transport system (I think that one's been done). efficient: uses economical code (e.g. few command lines). functional: does something. Cheers Lizzie Elizabeth Liddle

Per Elizabeth Liddle:

Often that code is creative, innovative and useful.

But I’ll keep trying. For clarity, no I don’t mean that computers write their own code in general, but that it it is possible for code to write code, and for that code to be extremely creative, efficient, and functional.

What is meant by the terms "creative", "innovative", "useful", "efficient", and "functional"? ciphertext

Weasel is Darwinian. It just has an extremely simple problem to solve and that problem has only one solution. No it isn't. Darwinian mechanisms have no "goal" or "foresight." Targeted searches do. Weasel is a targeted search. You could say it's demonstration of intelligent evolution.

That’s why Dawkins wrote it.

He wrote it to demonstrate the idea of the accumulation of the selection of small modifications contra the idea of a whirlwind blowing thru a junk yard producing a 747, etc. It does demonstrate that, but it does not demonstrate *Darwinian* (blind, purposeless) evolution. mike1962

Hi Lizzie. Thanks for your post. This could explain some of the misunderstanding and give us something to work with. I asked: "Are all evolutionary processes Darwinian? Can an evolutionary process be non-Darwinian?" Is your answer YES to both those questions? Or can I take it that you were saying YES to the second question and that therefore your answer to the first question is NO? cheers Mung

DrBot, Do you have a reference to where MathGrrl and I debated or discussed Avida? No comment on Joseph's post #26? You deny Lizzie was responding to the comment in the OP about computer programs not writing themselves (maybe she was quoting it by mistake and thinking of something completely different)?

Mung, I see little point in carrying on conversations with you.

Evidence that I do know what I'm talking about is ignored. Maybe that could have something to do with it? Mung

Mung:

You regard the processes within Avida as Darwinian. I don’t. No surprise there I suppose. :) How can we resolve the difference of opinion, or at least discuss it in a meaningful way? Do you at least see why in one case, when you say “evolutionary processes” when what you really mean by that is Darwinian processes that I might think you were equivocating?

I try to be careful when I use each precisely so as not to equivocate. I use "evolutionary processes" to denote processes that include factors like drift (and sometimes some others). I use "Darwinian processes" denote, strictly, processes in which self-replicators replicate with variants, and in which the variance confers different degrees of reproductive success. If I used the wrong term at any point, it was accidental, but I do try to be careful.

Are all evolutionary processes Darwinian? Can an evolutionary process be non-Darwinian?

Yes.

Think Weasel.

Weasel is Darwinian. It just has an extremely simple problem to solve and that problem has only one solution. That's why Dawkins wrote it. Avida has more difficult problems for its organisms to solve, and there are a possibly infinite set of solutions to each. So it's more like biology. But both are Darwinian. Elizabeth Liddle

DrBot:

I see no evidence of this either!

I take it you just ignore it when I quote from standard texts in the field. Is it because they refute what you say or do you have as better reason? Think back to the FEA thread and the subject of representation. You still think the chromosome/genome in a GA doesn't need to be designed? We can actually test that claim in a real GA if you want. You recall me explaining that you were talking about how a GA works and I said I was talking about what it takes to get a GA to work? There's a difference, don't you know. Mung

Elizabeth Liddle:

But the argument that once Darwinian process have got going they can’t generate further functional complexity is simply falsified by programs like Avida.

You regard the processes within Avida as Darwinian. I don't. No surprise there I suppose. :) How can we resolve the difference of opinion, or at least discuss it in a meaningful way? Do you at least see why in one case, when you say "evolutionary processes" when what you really mean by that is Darwinian processes that I might think you were equivocating? Are all evolutionary processes Darwinian? Can an evolutionary process be non-Darwinian? Think Weasel. Mung

Ah, well, that makes two of us, Mung. We should get on fine :) Now, about that null hypothesis... Elizabeth Liddle

Elizabeth Liddle:

And I’m sorry if I offended you.

I haven't taken offense. Not even what DrBot wrote offended me. So no worries. Heck, look what StephenB had to say about me, lol. I think that if I take offense at something I need to look at myself and ask why. Why did that offend me. I'm not easily offended, and it it happens I easily forgive. Do unto others, and all that. ;) Mung

I’m interested in honest debate.

If you are interested in serious debate then I would be very happy - I see no evidence of this so far though.

it’s not because I don’t know what I’m talking about

I see no evidence of this either! DrBot

Yes,I would agree that not all GAs etc write code. I wasn't making a broader claim. I was merely pointing out that some can. One Black Swan and all that. Yes, I know what a DSL is. And I'm sorry if I offended you. The thing is I don't usually have this much difficulty in making myself understood, and I'm finding it frustrating when you seem to simply misread my words (simple for simplest, for example) and then partially quote me in a manner that makes it look as though I meant something quite different from what I did mean! I am honestly trying to bullet proof my posts before posting, but communication does require effort from both ends, and I just didn't get the impression you were putting it in. But I'll keep trying. For clarity, no I don't mean that computers write their own code in general, but that it it is possible for code to write code, and for that code to be extremely creative, efficient, and functional. That seems to me to be a serious impediment to any argument that, once you have a Darwinian set up complexity can't emerge. Yes, the challenge is to address the issue as to how the Darwinian set up got going in the first place (you can't, as I have been reminded a few times) account for something by invoking itself to explain it. But the argument that once Darwinian process have got going they can't generate further functional complexity is simply falsified by programs like Avida. IMO. And it hasn't been changed by any argument made on this thread! Elizabeth Liddle

Elizabeth Liddle:

Quibbling about the difference between EAs and GAs doesn’t alter my point, unless you are determined not to get my point. Which it is starting to seem like, tbh.

It could be seen as quibbling. I thought about it earlier but did not say anything because you used a broader category and you said "some." But lately it appears like you're making global claims that apply to all EA/GAs. So I brought it up. You would agree, wouldn't you, that not all GA's write code? And that not all EA's write code? And that not all EA's and GA's evolve the code they write? Do you know what a DSL is? Mung

Elizabeth Liddle:

It is possible to write code-writing-code. That’s what a GA is.

The sad fact is, you're completely serious. Almost as sad, DrBot is encouraging this sort of thing. Your first statement above is not in dispute and has never been in dispute. I can write code that writes code, and which even executes the code it writes, in the Ruby programming language, without the use of a GA. But to assert that that is what a GA is? That a GA is "code-writing-code"? No. And then to argue that this demonstrates that computer programs can and do write themselves? Again, no. And then I get accused of having no interest in (or perhaps no ability to engage in) serious debate, of learning anything at all, and sometimes coming across as having a total lack of understanding of these subjects. Sad, truly sad. Mung

DrBot:

avida is not a targeted search.

MathGrrl and I never discussed Avida. And the person claiming Avida was a targeted search was Joseph, in #26, not me.

Mung, I see little point in carrying on conversations with you.

It would help if you get your facts straight.

...when viewed as a whole the system (software) writes its own code (more software) It is obvious and only someone willfully trying to distort and poison the debate would call it false.

Go back and read what Elizabeth was claiming. There is a difference between a program writing code on it's own and a program writing it's own code. In response to the claim in the OP:

Computer programs don’t write themselves

Elizabeth responded:

Well, yes, they do. Or can. That’s exactly what evolutionary algorithms do – evolve their own code.

That is not "exactly what EA's do." And in context it's clear that she meant what she wrote as a rebuttal to the claim in the OP. It clearly amounts to a claim that EA's write themselves.

Mung, I see little point in carrying on conversations with you. I get the impression you have no interest in (or perhaps no ability to engage in) serious debate, of learning anything at all, and you sometimes come across as having a total lack of understanding of these subjects (which may just be obfuscation).

The FEA thread comes to mind. In what sense is the genome or chromosome in a GA designed? I really don't care if you carry on a conversation or not, but it's not because I don't know what I'm talking about, or can't learn, or am not interested in serious debate. I'm interested in honest debate. Now, what claim would you like me to retract on the basis that it is false, or mistaken, or misguided? The one that all GA's and all EA's are not GP's? The one that Lizzie started out talking about EA's but shifted to GA's when what she really has in mind are GP's? The one that EA's don't write their own code? Mung

Mung:

…we can’t invoke Darwinian processes to account for something that is the simplest self-replicator that can evolve by Darwinian processes.

Life, as we know it, does not consist of simple self-replicators. A simple cell is stunningly complex. Hello.

I'm not sure why you are doing this, Mung. It makes communication extremely difficult. Likewise, what I said about programs evolving their own code was loose, but not false. It is possible to write code-writing-code. That's what a GA is. Quibbling about the difference between EAs and GAs doesn't alter my point, unless you are determined not to get my point. Which it is starting to seem like, tbh. I say this reluctantly, and hope to be proved wrong. :( Lizzie Elizabeth Liddle

mung, you said

And you’re confusing GA’s with Genetic Programming. You started out talking about EA’s, not GA’s.

Then your link goes to this

GP starts from an ooze of random computer programs, and progressively refines them through processes of mutation and sexual recombination, until solutions emerge. All this without the user having to know or specify the form or structure of solutions in advance.

where a GA is described, It's right there, on the first page, a description of a GA, yet you said it was different to a GA and use that as evidence.!!! GP is a specific approach to using GA's for creating programs - its right there, on the page you linked to! surely only someone willfully trying to distort and poison the debate would miss something like that, or perhaps you just don't get it ;)

It wasn’t a little loose, it was flat out false. So it is hardly the case that it was “essentially true.” Sheesh.

It was badly phrased, but not false - when viewed as a whole the system (software) writes its own code (more software) It is obvious and only someone willfully trying to distort and poison the debate would call it false.

What are you talking about. MathGrrl never discussed search algorithms. He avoided it like the plague.

avida is not a targeted search. Remember those bits? I think you do, don't you (otherwise it would indicate that you didn't read the discussions, making what you just said a lie?) Mung, I see little point in carrying on conversations with you. I get the impression you have no interest in (or perhaps no ability to engage in) serious debate, of learning anything at all, and you sometimes come across as having a total lack of understanding of these subjects (which may just be obfuscation). Perhaps, like Elizabeth suggested, you a lawyer, one who doesn't understand science? DrBot

DrBot:

Which is where your argument with mathgrrl comes from – she was using the correct terminology in the context of discussions about search algorithms.

What are you talking about. MathGrrl never discussed search algorithms. He avoided it like the plague. He wouldn't even engage on the subject of what a target was. Mung

Elizabeth Liddle:

And while I agree that it might have been a little loose to say that “GAs write their own code” it’s essentially true: A GA (ie. Virtual environment, self-replicators and fitness function) can result in the writing of code that was not written by a human designer. Often that code is creative, innovative and useful.

It wasn't a little loose, it was flat out false. So it is hardly the case that it was "essentially true." Sheesh. And you're confusing GA's with Genetic Programming. You started out talking about EA's, not GA's. http://www.gp-field-guide.org.uk/

While it is common to describe GP as evolving programs, GP is not typically used to evolve programs in the familiar Turing-complete languages humans normally use for software development. It is instead more common to evolve programs (or expressions or formulae) in a more constrained and often domain-specific language.

Mung

Let me quote the OP:

Life is fundamentally based on information and information processing — a software computer program and its associated, highly functionally integrated execution hardware. Computer programs don’t write themselves, and they especially don’t write themselves when random errors are thrown into the code.

Mung

...we can’t invoke Darwinian processes to account for something that is the simplest self-replicator that can evolve by Darwinian processes.

Life, as we know it, does not consist of simple self-replicators. A simple cell is stunningly complex. Hello. Mung

Elizabeth Liddle:

The second is the one I am addressing here, and was advanced by Gil, and others – that evolutionary processes (in other words replication with variation that results in differential reproduction) can’t result in complex functions. It can.

Well Elizabeth, there you go equivocating again. Is it that you just can't help yourself? Please provide your absolute best example of an evolutionary process bringing about complex functions. Even better, and more to the point, provide one from nature in which the selection mechanism is blind, unguided, purposeless, non-teleological, unintentional, etc. and the source of variation is likewise. Mung

Mung:

Elizabeth Liddle:

There is plenty of evidence, from both biology and GAs, that Darwinian evolution works once you’ve got the ball rolling. The question is, does it require an Intelligent Designer to roll that initial ball?

Well, in the case of GA’s, we know the answer is yes. So if you can use GA’s to assert the power of chance + selection, why can’t we use GA’s as evidence for the presence of design? Do you think GA’s indicate that there is design in life Lizzie? Or can they only be used when they support a particular view.

Since I've been posting here, I've encountered two arguments for ID. One I think has potential validity. The other I don't think does. The second is the one I am addressing here, and was advanced by Gil, and others - that evolutionary processes (in other words replication with variation that results in differential reproduction) can't result in complex functions. It can. And while I agree that it might have been a little loose to say that "GAs write their own code" it's essentially true: A GA (ie. Virtual environment, self-replicators and fitness function) can result in the writing of code that was not written by a human designer. Often that code is creative, innovative and useful. So the argument that we must infer ID because Darwinian processes can't result in complex innovative functions I think is falsified. However, the argument that that a Darwin-capable self-replicator is sufficiently complex that it must be designed by a designer, has, potential merit, because we can't invoke Darwinian processes to account for something that is the simplest self-replicator that can evolve by Darwinian processes. I think it's important to separate these two - it's simply the difference between the Origin of Species and the Origin of Life. I think Darwinian evolution is good for the Origin of Species - no ID argument is required. But Darwin himself did not attempt to solve the Origin of Life problem. Actually there's one other argument that I should mention, that perhaps has a little merit, IMO: that Darwinian natural selection works fine, but that the mechanisms of variance must be designed - random variation isn't good enough. I think this also fails though, because we know that most mutations, at least in a well-adapted species, are neutral or slightly deleterious, and we also know that even when this is true in an GA, where the variants really are randomised, evolution still occurs. So from my PoV, you guys still have the OOL argument :) But I have it in my sights.... Elizabeth Liddle

Indium:

your question “Do genetic algorithms write themselves?” is, well, not even wrong.

The answer is not that the question is not even wrong, the answer is no, GA's do not write themselves. If you scroll up in the thread you would see it claimed that EA's evolve their own code. Start with post #12 please. Mung

Well, there you have it. The power of chance + necessity.

Those choosy snowflakes! Mung

Elizabeth Liddle:

There is plenty of evidence, from both biology and GAs, that Darwinian evolution works once you’ve got the ball rolling. The question is, does it require an Intelligent Designer to roll that initial ball?

Well, in the case of GA's, we know the answer is yes. So if you can use GA's to assert the power of chance + selection, why can't we use GA's as evidence for the presence of design? Do you think GA's indicate that there is design in life Lizzie? Or can they only be used when they support a particular view. Mung

I love Lizzie's analogies :) Consider a snowflake. Tossed by the winds. Gently dropping to earth due to the inexorable pull of necessity (gravity). It lands. Many snowflakes land. They spell, "Welcome to Aruba, kairosfocus." Well, there you have it. The power of chance + necessity. Mung

Elizabeth Liddle:

Well, Avida incorporates “blind undirected processes” in that the mutations are random, and the only criteria that decides whether or not an organism “breeds” is whether it performs some function.

This is wrong. Did you even read the paper? All organisms in the initial population start out able to replicate but not performing any of the logic functions.

Experiments began with an ancestor that could replicate but could not perform any logic functions.

Mung

DrBot:

We are talking about evolutionary processes which require replication with variation where the variety affects replication rates.

As a point of fact, Elizabeth was talking about code that writes itself. Mung

Indium:

Somehow I get the impression that neither Mung nor kf understand the concept of a “model”.

Latecomer are you? The subject of GA's as evolutionary models or simulations has already been discussed in a previous thread. We're beyond that now. Mung

Well, I’d certainly like to know what kind of barriers you envisage. Can you give an example? Of a transition you think is impossible, or at least, very unlikely?

Possible candidates are, of course, chemical/mechanical barriers. Let me give you an example by analogy. If I handed you a brand new Rubic's Cube and told you to solve it, assuming you are a person of normal intelligence, you would be able to do it sooner or later. (Maybe you would have to read a book and learn some tricks.) However, if I gave a Rubic's Cube for which I rearranged the colored dots such that it can never be solved, and you happened not to notice my little gag, why then, of course, you would never be able solve it. There would be no pathway to the standard solution. What chemical barriers exist in nature with regards to blind evolution of the particular known DNA replicator? Nobody knows. In engineering we have to prove our concepts from start to finish, especially when lives are at stake. We have to build models and test them. That's when you're liable to find holes in your design.

So what candidate barriers do you see as preventing the steady incremental adaptive evolution, and the intermittent divergence, of populations over time from accumulating large scale differences between the extant populations and their remote ancestors or cousins?

Nobody knows if chemical/mechanical pathways exist for the kind of large scale blind evolutionary change Again, if you want to convince this engineer, the onus is on you, not me, to demonstrate that the known sources of variation + natural selection are sufficient to generate novel organs, body plans, cell and tissue types, not to mention the wiring of nervous systems and brains. The concept is not proven. To say that currently known mechanisms are sufficient is a very large leap of faith. I wouldn't trust engineers who employ that sort of philosophy of design. Extraordinary claims require extraordinary evidence. The blind evolutionary narrative is an extraordinary claim. mike1962

Mike: Ah, I understand what you are saying, thanks:

: You can think whatever you want, but it wouldn’t fly in the real world of engineering. You see, in the real world, sloppy unproven thinking like that can cost lives. I have no reason to accept that what we know about micro-evolution would scale to the macro world of novel organs, body plans, cell types, and tissue types. Just because you can drive a car to L.A to New York, doesn’t mean you can drive it to London. There are extra mechanisms needed to surmount the problem of the Atlantic.

Two points: One is that we are not talking in terms of dimensional scale here, but over time. Nobody is saying that because a bacterium can, in one step, mutate to have antibiotic resistance, then so can a land-based artiodactyl mutate in one step to a whale. I'm sure you know this, I just want us to be clear what we mean by "scale" here. We are instead saying: if we can drive from LA to Las Vegas, then we can drive from LA to New York as long as we allow more time. Second is: fair point re the Atlantic. However, that's what I mean by the onus - essentially what you are postulating is there is some "Atlantic" that must surely get in the way between our land-based artiodactyl and the whale (well in that case it might BE the atlantic, I guess!). So what candidate barriers do you see as preventing the steady incremental adaptive evolution, and the intermittent divergence, of populations over time from accumulating large scale differences between the extant populations and their remote ancestors or cousins?

The onus is on evolutionary biologists to demonstrate that the known mechanisms of stochastic “blind” micro-evolution are adequate for the production of “macro” level biological features. As of yet, no sale.

Well, I'd certainly like to know what kind of barriers you envisage. Can you give an example? Of a transition you think is impossible, or at least, very unlikely? Elizabeth Liddle

Elizabeth: However, I think the onus is on those who balk at “macro evolution” to explain why, when we observe both micro-evolution within a lineage, and speciation of populations into two lineages, why this process can’t continue to happen.

You can think whatever you want, but it wouldn't fly in the real world of engineering. You see, in the real world, sloppy unproven thinking like that can cost lives. I have no reason to accept that what we know about micro-evolution would scale to the macro world of novel organs, body plans, cell types, and tissue types. Just because you can drive a car to L.A to New York, doesn't mean you can drive it to London. There are extra mechanisms needed to surmount the problem of the Atlantic. The onus is on evolutionary biologists to demonstrate that the known mechanisms of stochastic "blind" micro-evolution are adequate for the production of "macro" level biological features. As of yet, no sale. mike1962

Mike1962:

Elizabeth Liddle: Well, we can observe micro-evolution and speciation, and multiply

Too bad most evolutionary biologists are not engineers. If they were, they might understand why a lot of engineers are skeptical. I would never get a plane designed by someone like you Elizabeth.

Probably an excellent idea, Mike :) However, I think the onus is on those who balk at "macro evolution" to explain why, when we observe both micro-evolution within a lineage, and speciation of populations into two lineages, why this process can't continue to happen. What stops it, in your view? F/N (see, I'm getting the lingo here :)): I did in fact do a bit of structural engineering as part of my training in architecture. I've sized the odd beam fairly successfully, and designed the odd truss :) I don't think anything has fallen down yet. Elizabeth Liddle

Joseph: well, let's not argue about terminology - the important thing is that I know what you mean, so thanks for clearing that up. Indeed genetic drift is an important factor in evolution.

But what is important is NS is a result- an after market assessment.

Yes indeed.

Then with sexual reproduction there isn’t any guarantee that any of the offspring will get the heritable beneficial trait.

Well, it depends on the number of offspring of course, but yes, a new mutation, even if potentially beneficial, is at its most vulnerable until a few members of the population have it. Then its future starts to become more secure. There are fancy equations in population genetics that describe these probability functions, as you are probably aware.

So one could out-reproduce the rest in one generation but the next generation could be altogether different.

Yes, there are important stochastic factors in selection. However, the larger the proportion of the population that has the new mutation, the more likely it is to go to fixation. So yes, the early stages are the "riskiest".

Fastest can also mean the first to the ambush- IOW it ain’t that black and white.

You are absolutely right. Again, this is why the probability of net beneficial mutation increases as the proportion of the population that bear it grows. But if it is very marginal (running away from bears isn't that much more common than running into them) then it may be stopped by drift effects anyway.

Ya see there are competing traits- better sight, stronger, better hearing-> and then there is just plain ole luck and cooperation.

Yup. The fitness landscape is both multidimensional and stochastic.

As for “random” perhaps you should read “Not By Chance” by Dr Lee Spetner.

Thanks for the recommendation. I have serious issues with the word "random" - it's used in lots of different ways, some of them deeply contradictory.

As for genes and genotype- genes influence traits and influence development. They do not determine what is going to develop.

Well, again, it's stochastic rather than deterministic. We are talking about changing probabilities here (biasing rather than determining what happens). And some genes are more deterministic than others. Relatively few phenotypic effects are determined by a single gene. Cool - we seem to agree on most of that :) Elizabeth Liddle

First natural selection is a result of three processes- (1)differential reproduction due to (2)heritable (3)random variation. Elizabeth:

Well, that would usually be lumped together as “Darwinian evolution” rather than “natural selection, which is the differential reproduction part” but, yes, that’s it, basically.

That is natural selection, period. Darwinian evolution is also OK with random genetic drift. But what is important is NS is a result- an after market assessment. Then with sexual reproduction there isn't any guarantee that any of the offspring will get the heritable beneficial trait. So one could out-reproduce the rest in one generation but the next generation could be altogether different. Elizabeth:

The differential selection part is obviously stochastic in one sense (there are factors other than heritable ones that affect your reproduction chances) but the heritable factors are what they are – if you can run faster you can run faster, and if that helps you catch prey/escape from predators, then it will help you survive to reproduce.

Fastest can also mean the first to the ambush- IOW it ain't that black and white. Ya see there are competing traits- better sight, stronger, better hearing-> and then there is just plain ole luck and cooperation. As for "random" perhaps you should read "Not By Chance" by Dr Lee Spetner. As for genes and genotype- genes influence traits and influence development. They do not determine what is going to develop. Joseph

Elizabeth Liddle: Well, we can observe micro-evolution and speciation, and multiply</blockquote? Too bad most evolutionary biologists are not engineers. If they were, they might understand why a lot of engineers are skeptical. I would never get a plane designed by someone like you Elizabeth.

mike1962

Joseph:

Elizabeth:

OK, so your issue is not with selection, but with the processes that result in variance?

First natural selection is a result of three processes- (1)differential reproduction due to (2)heritable (3)random variation.

Well, that would usually be lumped together as "Darwinian evolution" rather than "natural selection, which is the differential reproduction part" but, yes, that's it, basically. Very neatly done :)

Next there isn’t any methodology which demonstrates the known evolutionary mechanisms are stochastic/ spontaneous, ie blind and undirected (unguided).

Well this is where parsing it out is important. The differential selection part is obviously stochastic in one sense (there are factors other than heritable ones that affect your reproduction chances) but the heritable factors are what they are - if you can run faster you can run faster, and if that helps you catch prey/escape from predators, then it will help you survive to reproduce. The interesting question concerns the mechanisms of variance production, as you say. We tend to call them "random" because they mostly appear orthogonal to survival (a mutation is at least as to be deleterious as beneficial, and, in fact, in a well-adapted population, will almost certainly be so. So it doesn't look as though whatever processes govern mutations have any particular end in view (ditto with GAs). However, what may well be the case is that mutation rates, and mutation types are well suited to successful adaptation. This would be interesting, and would either suggest Intelligent Design, or selection above the level of the individual (differential persistence of populations rather than differential reproduction of individuals).

And still no one knows if any amount of variation can take some populations of imperfect replicators and produce the diversity of living organisms. Can’t simulate it because no one knows what makes an organism what it is.

Well, we can observe micro-evolution and speciation, and multiply :) As for your second point, we do know quite a lot now about how genotypic variation produces phenotypic variation, but it's in its infancy still. Elizabeth Liddle

“How did stupid atoms spontaneously write their own software, and where did the very peculiar form of information needed to get the first living cell up and running come from? Nobody knows” (New Scientist, September 18, 1999, volume 163, page 2204). bornagain77

Elizabeth:

OK, so your issue is not with selection, but with the processes that result in variance?

First natural selection is a result of three processes- (1)differential reproduction due to (2)heritable (3)random variation. Next there isn't any methodology which demonstrates the known evolutionary mechanisms are stochastic/ spontaneous, ie blind and undirected (unguided). And still no one knows if any amount of variation can take some populations of imperfect replicators and produce the diversity of living organisms. Can't simulate it because no one knows what makes an organism what it is. Joseph

What I keep pointing out is that until you get to the root the tree of life has no root.

Yes it does, we just don't understand what it is yet - chemistry and physics or intentional agent. We can study the tree and understand how it works. We can also try and discover what the seed was like. They are two related but separate lines of enquiry - the tree does not change when we discover the seed that produced it.

And until you can empirically warrant your claims what you have is a priori materialist metaphysics force-fitted on science

What we have are questions that need answers. Inserting 'intentional agent' into the gaps you perceive does not answer the question. It may yet turn out to be the right answer, but we do not have a complete understanding of the entire system yet so simply stepping back and proclaiming that you can see a gap, and therefore the only way it can be filled is by intentional agents is bad science. The scientific approach is to say - we don't know, lets try and find out. Not, we don't know, therefore some unknown intentional agent did it. DrBot

PS: What I keep pointing out is that until you get to the root the tree of life has no root. Similarly, until you can SHOW branching at body-plan level, incrementally, you have no branches. What you have is a forest of small bushes in effect. And until you can empirically warrant your claims what you have is a priori materialist metaphysics force-fitted on science, not real science free of censorship; exactly what Newton warned against in 1704 in Opticks, Query 31. kairosfocus

Indium: Not when you are moving up tot eh origin of a novel body plan. The claim of a smoothly branching tree of life, after 150 years, is till without empirical support worthy of the name. When it comes to AVIDA, I am pointing out that it is modelling something that happens long after you get to an island of rather specific and complex integrated function [hardware NAND gates and comparables, and riding on that software with sufficient sophisticated design to have NANDs as elements], so at best it is a loose analogue of micro-evolution, sufficient to say compare to the varieties of dogs and their wild kin. GEM of TKI kairosfocus

kf It is quite easy to find an "island of funtionality" if you already start on one, as self replicating organisms do. Since you keep discussing the origin of these self replicating agents (and NAND gates and whatever) you don´t seem to understand what AVIDA is supposed to model. For example, your question "Do genetic algorithms write themselves?" is, well, not even wrong. Indium

Joseph:

Elizabeth: So, Joseph: what would you regard as evidence for Darwinian processes being instrumental in the evolution of new body parts? First you would need to demonstrate all mutations are genetic accidents/ errors/ mistakes. Good luck with that beyond demonstrating blind, undirected chemical processes can produce a living organism from non-living matter. And if you did that I wouold accept darwinian processes can construct just about anything wrt biology.

OK, so your issue is not with selection, but with the processes that result in variance? That's sort of interesting, because of course Darwin had nothing to say about that either (in fact at one point he suggested something Lamarckian). However, I would point out, if that is indeed your position, that Avida is somewhat relevant, as, in Avida, the mutations are in fact random, and I think are drawn from an equiprobable distribution IIRC. So we know that even random mutations to virtual genotypes can produce useful selectable phenotypes. However, I take your point that the same may not be true of biological genotypes. Elizabeth Liddle

Thanks, Joseph, that is helpful. Elizabeth Liddle

I simply footnote: 1 --> I am pointing out the empirical characteristics of chance, necessity and choice. Necessity is not father to high contingency, and contingency comes from chance and/or choice. 2 --> that is empirically warranted, I have made no ontological assumptions about the roots of choice, only empirical inferences which establish that choice is possible and that it leaves signs. Such as the FSCI in posts in this thread. 3 --> the attempt to twist what I stated into a claimed assumption that choice cannot be reduced to chance and necessity is as turnabout, burden of proof shifting tactic. 4 --> We know, on empirical warrant that choice can source FSCI, but have no empirical warrant that can stand the light of day, that chance and necessity can do so. 5 --> In fact the real point is that we have signs in origins contexts that would -- absent a priori materialism whether overt or disguised as methodological naturalism -- point to choice as best candidate for explanation for origin of cosmos and of life and body plans, i.e FSCO/I etc. 6 --> to prop up this question-begging a priori, we see the claim that science is to be redefined in such terms, but this is neither epistemologically warranted nor historically justified. 7 --> Not to mention, that question begging is a logical fallacy. 8 --> instead, so long as choice is POSSIBLE at the relevant points as a candidate explanation, the empirical signs should be allowed to speak for themselves without interference from materialist censors hiding in lab coats. 9 --> In this context, some species of random walk in a config space is a reasonable model of chance. Notice, again trial and error filters out failures, it does not create new configs. necessity does not create contingency, it plays out form the dynamics of initial conditions. 10 --> It is accident of initial condition or injected noise along the way that have to explain contingency, once we are ruling out choice as the relevant candidate. 12 --> Laplace's demon's calculations on initial conditions start with the initials as a brute given, and he now has to deal with chance injections along the way, whether bubbling up from quantum behaviour [Schodinger's cat and that potentially fatal alpha particle . . .] or by the clash of un-correlated causal chains of necessity [a die drops and hits table, tumbles and settles] or both acting together. 13 --> So, if you lock out choice, you are left with chance to explain high contingency. (And note, physico-chemical laws embrace both; deterministic cause-effect chains on initial conditions, perturbations, quantum effects, noise, stochastic patterns, fluctuations etc. None of these -- singly or in combination -- is on OBSERVATION sufficient to explain sophisticated language, codes, algorithms, complex functionally integrated and specific structures, etc.) 14 --> So [while there is a lot of bun], "Where's the BEEF?" GEM of TKI kairosfocus

Elizabeth:

So, Joseph: what would you regard as evidence for Darwinian processes being instrumental in the evolution of new body parts?

First you would need to demonstrate all mutations are genetic accidents/ errors/ mistakes. Good luck with that beyond demonstrating blind, undirected chemical processes can produce a living organism from non-living matter. And if you did that I wouold accept darwinian processes can construct just about anything wrt biology. Joseph

Elizabeth:

Also, could you tell me what you mean when you say that ID is not “anti-evolution”?

Sure: Defining “evolution”:

Finally, during the evolutionary synthesis, a consensus emerged: “Evolution is the change in properties of populations of organisms over time”- Ernst Mayr page 8 of “What Evolution Is”

Biological (or organic) evolution is change in the properties of populations of organisms or groups of such populations, over the course of generations. The development, or ontogeny, of an individual organism is not considered evolution: individual organisms do not evolve. The changes in populations that are considered evolutionary are those that are ‘heritable' via the genetic material from one generation to the next. Biological evolution may be slight or substantial; it embraces everything from slight changes in the proportions of different forms of a gene within a population, such as the alleles that determine the different human blood types, to the alterations that led from the earliest organisms to dinosaurs, bees, snapdragons, and humans. Douglas J. Futuyma (1998) Evolutionary Biology 3rd ed., Sinauer Associates Inc. Sunderland MA p.4

Biological evolution refers to the cumulative changes that occur in a population over time. PBS series “Evolution” endorsed by the NCSE

Biological evolution, simply put, is descent with modification. This definition encompasses small-scale evolution (changes in gene frequency in a population from one generation to the next) and large-scale evolution (the descent of different species from a common ancestor over many generations) UC Berkley

In fact, evolution can be precisely defined as any change in the frequency of alleles within a gene pool from one generation to the next. Helena Curtis and N. Sue Barnes, Biology, 5th ed. 1989 Worth Publishers, p.974

Evolution- in biology, the word means genetically based change in a line of descent over time.- Biology: Concepts and Applications Starr 5th edition 2003 page 10

Intelligent Design is NOT Creationism (MAY 2000)

Scott refers to me as an intelligent design "creationist," even though I clearly write in my book Darwin's Black Box (which Scott cites) that I am not a creationist and have no reason to doubt common descent. In fact, my own views fit quite comfortably with the 40% of scientists that Scott acknowledges think "evolution occurred, but was guided by God."- Dr Michael Behe

Dr Behe has repeatedly confirmed he is OK with common ancestry. And he has repeatedly made it clear that ID is an argument against materialistic evolution (see below), ie necessity and chance. Then we have: What is Intelligent Design and What is it Challenging?- a short video featuring Stephen C. Meyer on Intelligent Design. He also makes it clear that ID is not anti-evolution. Next Dembski and Wells weigh in:

The theory of intelligent design (ID) neither requires nor excludes speciation- even speciation by Darwinian mechanisms. ID is sometimes confused with a static view of species, as though species were designed to be immutable. This is a conceptual possibility within ID, but it is not the only possibility. ID precludes neither significant variation within species nor the evolution of new species from earlier forms. Rather, it maintains that there are strict limits to the amount and quality of variations that material mechanisms such as natural selection and random genetic change can alone produce. At the same time, it holds that intelligence is fully capable of supplementing such mechanisms, interacting and influencing the material world, and thereby guiding it into certain physical states to the exclusion of others. To effect such guidance, intelligence must bring novel information to expression inside living forms. Exactly how this happens remains for now an open question, to be answered on the basis of scientific evidence. The point to note, however, is that intelligence can itself be a source of biological novelties that lead to macroevolutionary changes. In this way intelligent design is compatible with speciation. page 109 of "The Design of Life"

and

And that brings us to a true either-or. If the choice between common design and common ancestry is a false either-or, the choice between intelligent design and materialistic evolution is a true either-or. Materialistic evolution does not only embrace common ancestry; it also rejects any real design in the evolutionary process. Intelligent design, by contrast, contends that biological design is real and empirically detectable regardless of whether it occurs within an evolutionary process or in discrete independent stages. The verdict is not yet in, and proponents of intelligent design themselves hold differing views on the extent of the evolutionary interconnectedness of organisms, with some even accepting universal common ancestry (ie Darwin’s great tree of life). Common ancestry in combination with common design can explain the similar features that arise in biology. The real question is whether common ancestry apart from common design- in other words, materialistic evolution- can do so. The evidence of biology increasingly demonstrates that it cannot.- Ibid page 142

And from one more pro-ID book:

Many assume that if common ancestry is true, then the only viable scientific position is Darwinian evolution- in which all organisms are descended from a common ancestor via random mutation and blind selection. Such an assumption is incorrect- Intelligent Design is not necessarily incompatible with common ancestry.- page 217 of “Intelligent Design 101”

That is just a sample of what the Intelligent Design leadership say about biological evolution- they are OK with it.

Joseph

And again if the first living organisms were designed then it is a very safe inference that they were designd to evolve, ie evolved by design, just as GAs do with their originally designed populations. Joseph

So, Joseph: what would you regard as evidence for Darwinian processes being instrumental in the evolution of new body parts? What would you need to observe to be convinced? Also, could you tell me what you mean when you say that ID is not "anti-evolution"? Because I am still unclear how you are using that term. Elizabeth Liddle

Elizabeth:

There is plenty of evidence, from both biology and GAs, that Darwinian evolution works once you’ve got the ball rolling.

That is the propaganda. Unfortunately it doesn't have any evidentiary support. Joseph

Elizabeth:

I referred you to a paper that showed you how natural selection acting on variation (i.e. evolutionary processes) could produce new body parts.

Except they have no idea if natural selection didit. And they have no idea if the variation was random. Other than that- fine reference. :) Joseph

Can we get one thing very clear: Nobody is claiming that Darwinian processes (i.e. replication with variation that affects replication rate) explains the origin of the first self-replicators. All we are claiming is that GIVEN those first self-replicators, complex functional systems (including actual usuable computer code) can evolve, using no more than the Chance (random variation of the genomes) and Necessity (whatever the physics and chemistry of the system are, virtual or otherwise, designed or otherwise). That is easily demonstrated by the existence of GAs; it is also demonstrated in the field, in the lab, in biology, and speciation is not only observed both in the field and in the lab but is extremely well attested in the fossil record. But none of this tells us how the thing got set up in the first place, and no-one here is trying to tell IDists that it does. It's a complete straw man. What IDists need to do, it seems to me, to demonstrate support for their hypothesis, is to do as kairosfocus has done in the past, i.e. to demonstrate that the minimal self-replicator required for Darwinian evolution is too complex to have arisen by chance. That isn't even an attack on Darwinian evolution, it is an attack on the as yet unsupported idea that the first Darwinian-capable self-replicators could have arisen by mere Chance and Necessity. It doesn't seem "obvious" to me that this is impossible, nor "stupid" of me to think that it might be. But to demonstrate that it is plausible, I (or we, I guess) need probably to demonstrate that the postulated simplest Darwinian self-replicator is simpler than that proposed, and that it could have arisen through non-Darwinian processes. We haven't done that yet. That's our Achilles heel. But by tilting at Darwin, you are tilting at windmills! IMO. There is plenty of evidence, from both biology and GAs, that Darwinian evolution works once you've got the ball rolling. The question is, does it require an Intelligent Designer to roll that initial ball? Elizabeth Liddle

Elizabeth, sees a post entitled 'Selling Stupid', and does her utmost to sell away. bornagain77

F/N: It seems that I have to repeat, again and again and again . . . , that material causes break down into chance and/or mechanical necessity.

Well, in a sense, yes. I don’t think they are clearly separate, but I’ll buy it.

In that context, necessity — as Newton pointed out 300 years ago and more — does not account for high contingency, but for natural regularities. That is a frequent target of scientific investigation: identify regularities and deduce covering laws then embed into theoretical models. And yes, a theory is an explanatory model. High contingency — wide range of possible outcomes under materially similar initial conditions — is accounted for by chance and/or choice. If choice is ruled out ex hypothesi, then that leaves only chance. Random character generation is then what we are left with to account for digital, coded information. A white noise source feeding a random walk in a configuration space is a good model for that. In effect, monkeys at keyboards or the equivalent.

kairosfocus, it appears to me that you are inadvertently assuming your conclusion - you assume that “choice” cannot itself be accounted for by “chance” and “necessity”. Take the simplest “choice” you can imagine: a sand grain, drifting along a sea bottom may drop to the bottom and lodge there, or continue to drift on the current. We can discuss what happens to it as a combination of “chance” (the eddies in the current, the topography of the sea bottom) and “necessity” (gravity, water resistance, terminal velocity whatever). But let’s say that the bottom starts as a perfectly flat plane of sand. At first, the only factors that affect whether a given sand grain drops are the “chance” eddies in the current. However, over time, because the process is stochastic, and because the eddies have a spatial frequency, despite being stochastic (the distribution is not flat, in other words), the sea bottom becomes uneven – it acquires a more varied topography. Now, the topography itself interacts with the current – where there is a peak, the current will be slowed , and be more likely to drop sand; where there is a trough, the current will sail over the top, carrying its sand with it. So we have a positive feedback loop, resulting, over time, in patterns of ridges, or ripples, in the sand of the sea bottom. Now, it’s a very simple pattern, and very easily explained, and it would be absurd to consider it a “choosing” system in the same breath as, say, a human being “chooses” whether to sit down on a bench or walk on to the next one. Nonetheless, what the system embodies is a feedback system of contingencies. I suggest that where that feedback system acquires depth (more contingent branching) and more feedback loops (so that what happens next is contingent on what happened before) we have the beginnings of a real “choosing” system, based, nonetheless, simply on Chance and Necessity. And Darwinian evolutionary processes are exactly such a system; not only that, but when our brains choose, an analogous “neural Darwinism” works in just the same way, so it is no surprise (to me) that the products of Darwinian evolution should bear a family resemblance to the products of human choice. Elizabeth Liddle

If choice is ruled out ex hypothesi, then that leaves only chance. Random character generation is then what we are left with to account for digital, coded information. A white noise source feeding a random walk in a configuration space is a good model for that. In effect, monkeys at keyboards or the equivalent.

If intention is ruled out then what you have left is physics -> chemistry, electromagnetism, gravity etc. A random walk is not a model of any of these things, for example it rules out nested contingencies - one set of reactions will lead to another set, and another, all contingent on previous ones. You don't know if OOL depends on contingencies like this yet you explicitly rule them out of your model. you are not proposing a model that relates to reality.

More sophisticated GAs and their kin go wrong when they present a system that is moving in a carefully constructed and constrained zone of function and hill climbs as though this does not beg the question of how one can get to a zone of function to begin with.

This is exactly what GA's do, be they models of biology or engineering design tools. The evolutionary processes rely on an existing functional replicator as a starting point. The question of how you get to this proto function is not one that the theory of biological evolution, or experiments with GA's for engineering, is addressing. You can yell about question begging all you like, what you fail to understand is that it is a different question than the one of evolution. In the case of biology, it is the origin of life question. In the case of GA's for engineering - we observe people designing them so no question is begged.

Your valid model of body plan origin evolution is

Growth and development? I believe we have already been over this in other threads, and Elizabeth has pointed to actual research into how novel body plans can evolve. I'm an engineering scientist who uses GA's to solve design problems so it is not a question I get into much, when I do it tends to be from the perspective of designing GA's that will evolve better solutions - but I would say that taking inspiration from biological developmental processes is quite a good approach if you want to generate phenotypic diversity.

So, we see the bait and switch from yet another angle: at best a model of micro-evo enough to plausibly explain varieties of dogs and kin or red deer and American Elk or maybe varieties of trout and salmon, is being extrapolated beyond all reasonable limits as though it explains origin of vertebrates, arthropods, worms, plants, fungi etc.

Only if you ignore all the research demonstrating that it is not beyond reasonable limits! DrBot

F/N 3: notice that little slip: "programmed cell death." kairosfocus

F/N 2: More on HOX funcitonality:

Hox genes act at many levels within developmental gene hierarchies: at the "executive" level they regulate genes that in turn regulate large networks of other genes (like the gene pathway that forms an appendage). They also directly regulate what are called realisator genes or effector genes that act at the bottom of such hierarchies to ultimately form the tissues, structures, and organs of each segment. Segmentation involves such processes as morphogenesis (differentiation of precursor cells into their terminal specialized cells), the tight association of groups of cells with similar fates, the sculpting of structures and segment boundaries via programmed cell death, and the movement of cells from where they are first born to where they will ultimately function, so it is not surprising that the target genes of Hox genes promote cell division, cell adhesion, apoptosis, and cell migration.

Chicken-egg loops again and again, in a context of regulation and control based on information stored in a coded, digital system. This SCREAMS "sophisticated design." kairosfocus

F/N: Wiki clip on hox genes, as FYI:

Hox genes are a group of related genes that determine the basic structure and orientation of an organism.[1] Hox genes are critical for the proper placement of segment structures of animals during early embryonic development (e.g. legs, antennae, and wings in fruit flies or the different vertebrate ribs in humans). Hox genes are defined as having - a DNA sequence known as the homeobox their location in gene clusters on the genome an expression pattern along the anterior-posterior (head to tail) axis that corresponds to the relative location of their genes within the Hox gene cluster . . . . The products of Hox genes are known as Hox proteins. Hox proteins are transcription factors, as they are capable of binding to specific nucleotide sequences on the DNA called enhancers where they either activate or repress genes. The same Hox protein can act as a repressor at one gene and an activator at another. For example, in flies (Drosophila melanogaster) the protein product of the Hox gene Antennapedia activates genes that specify the structures of the 2nd thoracic segment, which contains a leg and a wing, and represses genes involved in eye and antenna formation.[3] Thus, legs and wings, but not eyes and antennae, will form wherever the Antennapedia protein is located. The ability of Hox proteins to bind DNA is conferred by a part of the protein referred to as the homeodomain.

In short, HOX genes are very complex and functionally specific genes, that loop back into DNA functionality chicken and egg style. The best empirically warranted explanation for such a pull up by bootstrap system is that it is designed in to initiate from a start point, carry out a process and terminate at the appropriate point. GEM of TKI kairosfocus

PS: Joseph is precisely correct the issue is the origin of HOX genes not HOX genes can do magic. kairosfocus

Onlookers: The above is simply put up for record; predictably it will be brushed aside again as it has been for months. There is a refusal to recognise that he origin of complex and specific functionality based on coded information is a problem that has to be explained within the resources of the solar system or the observed cosmos on chance + necessity if you are going to resort to a non-design explanation. There is instead an assumption that once one has function it can vary step by step in unlimited ways. But you see embryonic development from a zygote or equivalent [what's the plant equivalent?], is an algorithmic process, and it varies considerably across even homologous-ly structured creatures. Body plan variations have to be explained on mutations expressed very early in development, and such mutations will have to be co-ordinated to achieve integrated function. Notoriously, such early mutations are lethal, precisely because throwing a monkey wrench into a tightly integrated system is vastly more likely to disrupt than to enhance. I think it is time for me to cite Meyer in that famous, hated 2004 PBSW article again (and I see talk Origins is hoping to make Expelled drop down the memory hole . . . ), so you can see what is being ducked and/or distracted from: _________ >> The Cambrian explosion represents a remarkable jump in the specified complexity or "complex specified information" (CSI) of the biological world. For over three billions years, the biological realm included little more than bacteria and algae (Brocks et al. 1999). Then, beginning about 570-565 million years ago (mya), the first complex multicellular organisms appeared in the rock strata, including sponges, cnidarians, and the peculiar Ediacaran biota (Grotzinger et al. 1995). Forty million years later, the Cambrian explosion occurred (Bowring et al. 1993) . . . One way to estimate the amount of new CSI that appeared with the Cambrian animals is to count the number of new cell types that emerged with them (Valentine 1995:91-93) . . . the more complex animals that appeared in the Cambrian (e.g., arthropods) would have required fifty or more cell types . . . New cell types require many new and specialized proteins. New proteins, in turn, require new genetic information. Thus an increase in the number of cell types implies (at a minimum) a considerable increase in the amount of specified genetic information. Molecular biologists have recently estimated that a minimally complex single-celled organism would require between 318 and 562 kilobase pairs of DNA to produce the proteins necessary to maintain life (Koonin 2000). More complex single cells might require upward of a million base pairs. Yet to build the proteins necessary to sustain a complex arthropod such as a trilobite would require orders of magnitude more coding instructions. The genome size of a modern arthropod, the fruitfly Drosophila melanogaster, is approximately 180 million base pairs (Gerhart & Kirschner 1997:121, Adams et al. 2000). Transitions from a single cell to colonies of cells to complex animals represent significant (and, in principle, measurable) increases in CSI . . . . In order to explain the origin of the Cambrian animals, one must account not only for new proteins and cell types, but also for the origin of new body plans . . . Mutations in genes that are expressed late in the development of an organism will not affect the body plan. Mutations expressed early in development, however, could conceivably produce significant morphological change (Arthur 1997:21) . . . [but] processes of development are tightly integrated spatially and temporally such that changes early in development will require a host of other coordinated changes in separate but functionally interrelated developmental processes downstream. For this reason, mutations will be much more likely to be deadly if they disrupt a functionally deeply-embedded structure such as a spinal column than if they affect more isolated anatomical features such as fingers (Kauffman 1995:200) . . . McDonald notes that genes that are observed to vary within natural populations do not lead to major adaptive changes, while genes that could cause major changes--the very stuff of macroevolution--apparently do not vary. In other words, mutations of the kind that macroevolution doesn't need (namely, viable genetic mutations in DNA expressed late in development) do occur, but those that it does need (namely, beneficial body plan mutations expressed early in development) apparently don't occur.6 >> _________ There is no sound answer to this challenge, never mind the sound and fury that you will hear, and the glittering computer simulations presented as though they are an answer. And, going back to Avida, the presence of NAND gates in software form and in the underlying hardware, are exactly what tells me that the whole is in the context of an island of intelligently designed function. Don't let the distractions make you miss this key point. BTW, trying to pull parts out of a grab bag and putting them together based on trial and error is a way to trigger explosions in the electronics lab. GEM of TKI kairosfocus

Joseph:

Elizabeth, The use of “evolution” is an equivocation because people use it against ID as if ID is anti-evolution.

Well, that's why I asked how you are defining the word. Then there will be no more danger of equivocation.

And if Darwinian processes cannot explain the first living organism than it cannot say anything about evolution as the two are directly linked-> organisms are designed then it is reasonable to infer they were designed to evolve- ie evolved by design. It is that simple.

Well, Darwinian processes can't explain the first living organism, if living organisms are defined as entities that are capable of evolving by Darwinian means. Darwin didn't even attempt to explain that. It's still a mystery, although there are strong clues. And this is why I think Darwin gets an unfair press here - people seem to think that Darwinian evolution is "tumbling", in the face of the rise of ID, and yet what they really seem to think is tumbling is the idea of abiogenesis. Which Darwin didn't even attempt to explain. Elizabeth Liddle

Joseph:

Elizabeth, Again you cannot use that which needs explaining in the first place to do the explaining. That means you cannot just say “Hox genes” and expect me just accept that.

Well, I didn't. I referred you to a paper that showed you how natural selection acting on variation (i.e. evolutionary processes) could produce new body parts. If you want an explanation of how Darwinian processes could produce Hox genes, then I don't know, off hand. I'll see if any research has been done. But the interesting thing is that they are fairly simple genes. The interesting thing about them are the genes that switch them on and off during development, resulting in very different body plans from genes that are highly conserved across species. Elizabeth Liddle

Elizabeth, Again you cannot use that which needs explaining in the first place to do the explaining. That means you cannot just say "Hox genes" and expect me just accept that. Joseph

Elizabeth, The use of "evolution" is an equivocation because people use it against ID as if ID is anti-evolution. And if Darwinian processes cannot explain the first living organism than it cannot say anything about evolution as the two are directly linked-> organisms are designed then it is reasonable to infer they were designed to evolve- ie evolved by design. It is that simple. Joseph

Joseph:

Yes, I read it. Nothing on blind, undirected processes and still nothing on how Darwinian processes produced HOX genes in the first place

Well, that wasn't what you asked for. and nothing on how Darwinian processes produced cellular differentiation. Nor that. I can probably find you a paper on that, if you are interested.

IOW you don’t have anything but hopes and prayers- although I doubt Darwin will answer them.

And research, fortunately. More effective for most practical purposes than prayer :) Although I don't knock prayer. Elizabeth Liddle

F/N: It seems that I have to repeat, again and again and again . . . , that material causes break down into chance and/or mechanical necessity. In that context, necessity -- as Newton pointed out 300 years ago and more -- does not account for high contingency, but for natural regularities. That is a frequent target of scientific investigation: identify regularities and deduce covering laws then embed into theoretical models. And yes, a theory is an explanatory model. High contingency -- wide range of possible outcomes under materially similar initial conditions -- is accounted for by chance and/or choice. If choice is ruled out ex hypothesi, then that leaves only chance. Random character generation is then what we are left with to account for digital, coded information. A white noise source feeding a random walk in a configuration space is a good model for that. In effect, monkeys at keyboards or the equivalent. Such a random walk has no constrained initial point, nor any oracle that can magnetise non-functional configs, if it is to be reasonable. That is where Weasel went wrong. Only when function does emerge can one then talk about rewarding differential success via hill climbing. the problem is the complexity involved in getting to function specifies config spaces that are easily vastly beyond the credible reach of our solar system or the cosmos. More sophisticated GAs and their kin go wrong when they present a system that is moving in a carefully constructed and constrained zone of function and hill climbs as though this does not beg the question of how one can get to a zone of function to begin with. First, navigate to your islands by a random walk . . . Nor does assuming that once one has the equivalent of a unicellular organism -- a huge quantum of function relative to cosmos scale search resources -- one then can smoothly move to higher complexity answer to the body plan origination challenge. So, we see the bait and switch from yet another angle: at best a model of micro-evo enough to plausibly explain varieties of dogs and kin or red deer and American Elk or maybe varieties of trout and salmon, is being extrapolated beyond all reasonable limits as though it explains origin of vertebrates, arthropods, worms, plants, fungi etc. kairosfocus

Joseph:

And enough of the equivocation- “evolutionary” processes/ mechanisms is an equivocation as ID is not anti-evolution.

There is no "equivocation" here, Joseph. However, I am pleased to hear that "ID is not anti-evolution". However, to prevent any further misunderstanding, could you explain what you mean by that?

What is needed is a simulation of blind, undirected processes as that i what ID argues against-> blind, undirected processes having sole dominion over evolution.

Well, Avida incorporates "blind undirected processes" in that the mutations are random, and the only criteria that decides whether or not an organism "breeds" is whether it performs some function. That is what happens in Nature - random mutations confer some function (protection from antibiotic toxins, for a bacterium, for instance) and if it helps it breed, then that mutation propagates. Exactly the same is true in biology. And if you are not "anti-evolution" you will have no problem with that. It is easily demonstrable. However, what Darwinian evolution does NOT explain, is how whatever the simplist possible self-replicator in the tree emerged in the first place. Darwin did not attempt to explain that. What is also still the subject of much ongoing research is the mechanisms of variance - how "random" are they? Are they, in fact, biased towards success? Again, Darwin did not tackle this, because he had no idea what the causes of variance actually were. We have more idea, but we still know very little about what the mechanisms of variance actually are. Elizabeth Liddle

Yes, I read it. Nothing on blind, undirected processes and still nothing on how Darwinian processes produced HOX genes in the first place and nothing on how Darwinian processes produced cellular differentiation. IOW you don't have anything but hopes and prayers- although I doubt Darwin will answer them. Joseph

Joseph, it isn't "bluster". Did you read the article about treehoppers? Elizabeth Liddle

And enough of the equivocation- "evolutionary" processes/ mechanisms is an equivocation as ID is not anti-evolution. What is needed is a simulation of blind, undirected processes as that i what ID argues against-> blind, undirected processes having sole dominion over evolution. Joseph

Elizabeth, For all of your bluster (re HOX genes) there still isn't any evidence that Darwinian processes can construct new, useful and functional multi-part systems. And there isn't any evidence Darwinian processes produced HOX genes nor is there any evidence taht Darwinian processes can change body plans. Joseph

Dr Bot: If you read in context you would see that I am specifically focussing on the equivalent of origin of body plans, and separately address origin of a self-replicating entity in a PS; which responds to a remark by Dr Liddle. The notion that in effect a hello world can be transformed into a system controller step by small step on one branch and an operating system on another, plus a video editor on a third, etc, on small random changes backed by trial and error success, is on its face highly dubious, and lacks empirical support. I repeat, the GAs and the like all show VARIATION WITHIN A HIGHLY SPECIFIED AND INTELLIGENTLY ARRIVED AT ISLAND OF FUNCTION. At most, on a generous analogy, micro-evo. Perhaps the most misleading Darwinian icon of all is the smoothly varying tree of life, once the complex information systems embedded in such life forms were identified. Your valid model of body plan origin evolution is _______? Your empirical evidence of spontaneous, un-designed origin of body plans by chance variation and culling through differential reproductive success on a tree of life type model starting with the Cambrian layers is ___________? (By contrast we can present he book publishing industry and the ICT industry as examples of intelligent design as the empirically warranted source of FSCI.) GEM of TKI kairosfocus

But the switcheroo now happens: presto, variation within an island of function on some hill-climbing algorithm or other, is now presented as an explanation of how one gets TO an island of function in an extremely large config space.

kindly explain where either myself or Elizabeth attempt to use evolutionary processes to explain the origin of self replicating systems. We are talking about evolutionary processes which require replication with variation where the variety affects replication rates. In the case of the things I have discussed, I would argue that these are instances of evolution, but not models of biological evolution.

In that context, Dr Bot et al, the challenge that even the NAND gate that is the underlying functional element involved, becomes a highly significant demonstration of the reality of already being on an island of function by design when you begin.

Yes, we start with building blocks, in the cases I've outlined we are evolving circuits from functional logic gates, although it has been tried with transistors, capacitors, resistors etc. (real components, not simulated) And with lithography mask patterns for transistor fabrication (which are tested in an industrial simulator used for verifying integrated circuit designs)

That is why I have so often raised the issue of using monkey at keyboard equivalents to write the program modules, then once functional modules exist we can then talk about moving around on islands of function and hill-climbing to one’s heart’s content. Such a model would definitively show that one can arrive at islands of function by chance variation and trial and error; on success.

But you just said this:

The pivotal issue is: exactly what one is modelling, and what does this signify? If we were doing a valid model that correctly simulated something and accurately said what it simulated with what limitations

Your proposed model is of random noise, it accurately simulates random character generation, but I suspect it would have a skewed distribution. If you want to establish if chemistry - the laws of physics - can get to an island of function then you need a model that represents these laws, interactions and contingencies. DrBot

PS: Origin of cell based life with metabolism integrated with an information based von Neumann self-replicator is of course simply the first case of body plan origin to be explained. While it is credibly at lest 100x the 1,000 bit threshold for FSCI, it is in turn actually far simpler in information origin terms than origin of body plans for complex organisms with specialised organs, which requires credibly 10 - 100+ MILLION bits of information dozens of times over for the dozens of main body plans. The hiving off of OOL as thought that is the special case that has not been solved, but hey, here's our wonderful branching tree of life model that allows us to see smoothly varying populations giving rise to body plans is yet another bait and switch in the teeth of the observed fossil evidence that body plans originate suddenly, and show stasis on basic form. kairosfocus

Indium: FYI, on odds, I was probably working with models (electronic ckt design h-parameter models are models, for just one instance [e.g. tell me how the AC ground and bypass capacitor lo freq rolloff are integrated to model a full amplifier . . . ] . . . and the classic kinetic theory of gases is a model, even, the Newtonian point-particle is a model [recall his reduction of a spherical planet to a particle equivalent; what is the density of a point-mass?], etc) when you were still in nappies. So, kindly, do not poison the atmosphere by presuming or suggesting ignorance or stupidity etc -- a la Dawkins and ilk. The pivotal issue is: exactly what one is modelling, and what does this signify? If we were doing a valid model that correctly simulated something and accurately said what it simulated with what limitations:

(all models are strictly false but exploit the peculiarity of the implication operator that a strictly false antecedent can under limited conditions entail correct consequences; hence the need for validation and rationalisation . . . ),

. . . that would be one thing; but what is happening here is that the "model" is being claimed to be a case of the real thing. What is actually being modelled is that: within an island of function, there can be successful adaptation of a complex functional information based entity on intelligent direction and arguably chance variation and differential success; i.e. at most microevolution. That is not in dispute, not even by young earth creationists. But the switcheroo now happens: presto, variation within an island of function on some hill-climbing algorithm or other, is now presented as an explanation of how one gets TO an island of function in an extremely large config space. [And, just to underscore the point, this is a cut down version of the phase space model used in statistical thermophysics and even in mathematics.] In that context, Dr Bot et al, the challenge that even the NAND gate that is the underlying functional element involved, becomes a highly significant demonstration of the reality of already being on an island of function by design when you begin. If you look at a hardware NAND, it is practically impossible to arrive at an island of function by chance on the gamut of the solar system's resources. When you move to a software NAND, which depends on the underlying hardware NAND or near-equivalents in several ways to make it go [as in how do you actually construct registers and the arithmetic and logic unit that is used to execute instructions], the same injection of huge dollops of design is there in the background. So, the validity of the claimed models, as touching the issue of arrival at islands of complex and specific function, is in immediate and serious doubt. That is why I have so often raised the issue of using monkey at keyboard equivalents to write the program modules, then once functional modules exist we can then talk about moving around on islands of function and hill-climbing to one's heart's content. Such a model would definitively show that one can arrive at islands of function by chance variation and trial and error; on success. But the problem is of course that 73 ASCII characters worth of functional information is beyond the resources of our solar system and 143 the observed cosmos. At the same time, 63 or even 125 bytes of information carrying capacity are impossibly small to implement any serious software process. Pointers that make something else do the heavy lifting DO NOT COUNT. (And FYI, that is what the so-called genomes in GAs do; real genes store specific control info for assembling proteins, and to regulate that assembly. Yet another point of relevantly inapt analogy.) In short, we have inappropriate extrapolation of a model beyond its generously plausible limits [micro-evo] to something it is profoundly and patently dis-analogous to, the claimed spontaneous origin of body plans by chance variation and differential reproductive success [body plan level macro evo]. (And if you want to play the card that macro is micro accumulated, that needs to be shown not asserted, on pain of bait and switch. Beyond that, the attempt to say that only skeptics speak in these terms so the terms are invalid is either devastatingly ignorant of the facts or outright irresponsible and/or willfully dishonest, cf the UD weak argument correctives.) Sorry, in fact Weasel was inappropriately used to persuade the general public and a lot of the scientific community that computer simulations were a valid way to model and present the unobserved macro-evolution. this was deceptivce, as Dawkins himself admitted and/or directly implied. Subsequent efforts have been poisoned by that underlying assumption or belief. The models, consistently show instead what intelligent design of a program, setting up on an island of function based on knowledge and skill, and exploiting the capacity of hill-climbing as a way to make explicit the peaks of designated functions on spaces surveyed is. The whole exercise is wrong-headed, and is further poisoned by an initial action that was plainly wrong hearted too. That is why we need to look back at Gil's point in the original post, and soberly reflect on it. GEM of TKI kairosfocus

Well, in a sense Avida is a model, but in another sense it's a perfectly real, potentially useful code-writing tool. And other GAs really are - they aren't written as a "model of" evolution, they are "modeled on" evolution, precisely because evolutionary processes (namely, replication with random variation that affects the probability of reproduction) is a superb way of generation novel and ingenious solutions to real world problems! I entirely take the point that this doesn't explain how self-replicators that replicate with variation that affects the probability of reproduction got started, in life, in the first place, but it seems to me Darwin is getting all the misplaced stick here. Darwinian evolution clearly works, as Avida shows. It can generate useful, functional information. The as yet unsolved problem is how the first minimal self-replicator got there in the first place. Elizabeth Liddle

Somehow I get the impression that neither Mung nor kf understand the concept of a "model". Indium

KF@68 We are discussing how evolutionary processes can generate novel solutions. All the examples I have given re logic gates illustrate how evolutionary mechanisms can generate novel designs.

Those who imagine Avida is an example of evolution: have you ever designed and built a NAND gate circuit? Do you know what is involved in getting a transistor made and working? In making sure you do not “let the smoke out” of same? Do you understand what is involved in making a flip flop, or how that is involved in counters, registers, and other processing units such as are involved in a computer? Do you know what is involved in setting up a simulation of same and getting it to run on a specific hardware machine?

You appear to be arguing that that the people who design and use these simulations of logic circuits, or the people who use GA's to evolve live electronics do not know anything about creating simulations or working with live electronics. This makes no sense - they are the ones actually writing the simulations and using real electronics, how could they do this if they had no understanding of simulations and electronics, how could they get results - real results.

If you do not, please listen to those of us who have done these things, and things far more complex than that.

We have done these things, we have used evolutionary processes to design real electronics, create better antenna, even to search for better transistors by evolving micro-lithography mask patterns. Don't tell us that something that has been done can't be done :)

They may show how one may, on terms of an existing design, one may move around within an island of fucntion by using hill climbing algorithms or the like, but that is a very different matter.

Yes, that is the topic we are discussing - EVOLUTION - the effect of selection on a population of imperfect replicators - which are, by definition, functional as replicators.

There is a massive bait and switch game going on

Yes indeed, we are discussing evolution, now you have moved the goalposts to biogenesis. We are discussing how a population of imperfect replicators can evolve, not the origin of those replicators. This is what the term evolution applies to when talking about biology and GA's - replication with error leading to variable replication rates in a population - it doesn't refer to the creation of self replicators, or the physical world. Equivocation: A fallacy based on the use of the same term in different senses. Avida is an example of evolution. DrBot

mung@37

DrBot:

Generally speaking any search has to have a target or it isn’t a search …

Thank you, thank you, thank you. I can’t believe how often I have to argue over this very simple fact (MathGrrl comes to mind).

I said:

Generally speaking any search has to have a target or it isn’t a search, but ‘targeted’ means something more specific – pre-specified and unchanging

Which is where your argument with mathgrrl comes from - she was using the correct terminology in the context of discussions about search algorithms. If you say 'I am searching' then that implies you are searching for something - it would make no sense to say 'I am searching for nothing' unless you meant 'I am not searching' - so in this general sense any search, including random search is looking for something, and we can call that a target. The phrase 'targeted search' is more specific and refers to having an explicit, pre-specified goal, rather than a set of dynamic criteria to match. Think about it this way - if all searches have to, by definition, have some form of 'target' in the loosest sense then arguing that they are therefore all 'targeted searches' is pointless semantics - the word 'targeted' is redundant because they are all searches and by definition there have to be targets of some kind. The inclusion of the word 'targeted' is there to indicate reference to a subset of search evaluation criteria, namely those that are explicitly defined and fixed in advance. DrBot

Mung:

So if it meets a target or goal, specified in advance, programmed in advance, it gets rewarded.

If it performs some function, specified in advance, it gets rewarded. It's important to be clear about the difference between a genotype that is rewarded for performing a function (cf being able to jump), and being rewarded having a specific genotype (as in WEASEL). The genotypic configuration that are rewarded are not specified in Avida - indeed they are not even known, and may not even be a finite set. In other words, the is target defined by the problem to be solved, not the solution that will solve it, of which there are a large unknown number.

Why do some logic functions get rewarded more than others, i.e., why do some logic functions have more “computational merit” than others?

Because that's how life is. Some functions are only marginally advantageous, some more so. Elizabeth Liddle

From the Wikipedia link cited above:

Intelligent design is the proposition that "certain features of the universe and of living things are best explained by an intelligent cause, not an undirected process such as natural selection." It is neo-creationism, a form of creationism restated in non-religious terms. It is also a contemporary adaptation of the traditional teleological argument for the existence of God, but one that deliberately avoids specifying the nature or identity of the intelligent designer. Its leading proponents—all of whom are associated with the Discovery Institute, a politically conservative think tank—believe the designer to be the Christian God.

Does this include the agnostic David Berlinski? And Moonie Jonathan Wells? Are there others? NZer

F/N: Collins English Dictionary: equivocation [??kw?v??ke???n] n 1. the act or an instance of equivocating 2. (Philosophy / Logic) Logic a fallacy based on the use of the same term in different senses, esp as the middle term of a syllogism, as the badger lives in the bank, and the bank is in the High Street, so the badger lives in the High Street Collins English Dictionary – Complete and Unabridged © HarperCollins Publishers 1991, 1994, 1998, 2000, 2003 --> equivocation is a fallacy, not always a willful deception --> However, insistent refusal to correct error in light of what one knows or should know can convert error into willful deception, so we must beware fallacy [?fæl?s?] n pl -cies 1. an incorrect or misleading notion or opinion based on inaccurate facts or invalid reasoning 2. unsound or invalid reasoning 3. the tendency to mislead 4. (Philosophy / Logic) Logic an error in reasoning that renders an argument logically invalid [from Latin fall?cia, from fallax deceitful, from fallere to deceive] Collins English Dictionary – Complete and Unabridged © HarperCollins Publishers 1991, 1994, 1998, 2000, 2003 --> Such fallacies, plainly, may involve simple error, not just willful deception --> Pardon, too: I seem to be especially typo prone these past couple of days. kairosfocus

From the Nature PDF cited above: Our experiments demonstrate the validity of the hypothesis, ?rst articulated by Darwin and supported today by comparative and experimental evidence, that complex features generally evolve by modifying existing structures and functions. Some readers might suggest that we ‘stacked the deck’ by studying the evolution of a complex feature that could be built on simpler functions that were also useful. However, that is precisely what evolutionary theory requires, and indeed, our experiments showed that the complex feature never evolved when simpler functions were not rewarded. Our experiments also show that many different genomic solutions produce the same complex function. Following any particular path is extremely unlikely, but the complex function evolved with a high probability, implying a very large number of potential paths. Although the complex feature ?rst appeared as the immediate result of only one or two mutations, its function invariably depended on many instructions that had previously evolved to perform other functions, such that their removal would eliminate the new feature. NZer

This has been a fascinating discussion, and so much more refreshing than the nastiness seen on other forums. Keep at it folks -- keep creating more light than heat!!! Personally I would like to see some documented examples of what GAs have been able to accomplish so that we can evaluate the best that they can produce. NZer

Mung (and Dr Liddle): Besides which [as was already pointed out], a physical NAND gate or even a computer simulation running on a machine in turn dependent on such instantiations or the like, is already a context so deeply embedded with complex designs that the whole exercise is one in suspension of knowledge in the interests of a fantasy. In short Avida is fatally compromised from teh outset. It is no more a true representation of chance variation and blind trial and error giving rise to increasing complex, specific functional order than is Weasel. Which, as you pointed out, was a sly rhetorical fraud. (The very name is a hint of that . . . ) I think Dr Liddle as a great many others, is being taken in by a clever fraud that has so pervaded our intellectual climate that many of us cannot see the obvious as just what it is. There is an equivocation in the system, I suspect rooted in the fraud of a few, but now embedded in the way of thinking of the many. And when engineers, applied scientists and software engineers familiar with what is being glossed over try to point out what is wrong, we are often dismissed out of hand. Those who imagine Avida is an example of evolution: have you ever designed and built a NAND gate circuit? Do you know what is involved in getting a transistor made and working? In making sure you do not "let the smoke out" of same? Do you understand what is involved in making a flip flop, or how that is involved in counters, registers, and other processing units such as are involved in a computer? Do you know what is involved in setting up a simulation of same and getting it to run on a specific hardware machine? If you do not, please listen to those of us who have done these things, and things far more complex than that. Long befoe you get tothe sort of demonstration toy program like Avida that is being showcased as an example of evolution by chance variation and natural selection, a cumulatively huge amount of design and testing was involved to get to an island of function. Avida et al are actually examples of how real world limited variation can be programmed into complex designed systems, by their designers. We must not allow evolutionary materialism promoting rhetoric to distract us from that. On the evidence to date, computer simulations inherently can provide no conclusive evidence of the power of chance variations and natural selections to originate complex, specific function. They may show how one may, on terms of an existing design, one may move around within an island of fucntion by using hill climbing algorithms or the like, but that is a very different matter. There is a massive bait and switch game going on, and if we play by the rules of those who set it up to begin with, we will be taken in. And, Mung, you are right to highlight how Dawkins gave the game away with some weasel words he knew would not be heard amidst the beguiling spectacle of seeming evolution in action:

It . . . begins by choosing a random sequence of 28 letters ... it duplicates it repeatedly, but with a certain chance of random error – 'mutation' – in the copying. The computer examines the mutant nonsense phrases [i.e. they are non-functional!], the 'progeny' of the original phrase, and chooses the one which, however slightly, most resembles the target phrase, METHINKS IT IS LIKE A WEASEL . . . . What matters is the difference between the time taken by cumulative selection [on non-functioning phrases rewarded by a designed so-called fitness function that ignores that inconvenient fact], and the time which the same computer, working flat out at the same rate, would take to reach the target phrase if it were forced to use the other procedure of single-step selection: about a million million million million million years. This is more than a million million million times as long as the universe has so far existed . . . . Although the monkey/Shakespeare model is useful for explaining the distinction between single-step selection and cumulative selection, it is misleading in important ways. [So, why was it used at all?] One of these is that, in each generation of selective 'breeding', the mutant 'progeny' phrases were judged according to the criterion of resemblance to a distant ideal target, the phrase METHINKS IT IS LIKE A WEASEL. [so Dawkins KNEW that this was not a rewarding of increments in function . . . ] Life isn't like that. Evolution has no long-term goal. There is no long-distance target, no final perfection to serve as a criterion for selection, although human vanity cherishes the absurd notion that our species is the final goal of evolution. In real life, the criterion for selection is always short-term, either simple survival or, more generally, reproductive success. [i.e. there must first be function before one can hill-climb, the very point of the objection that the real challenge is to get to the islands of function] [TBW, Ch 3]

Deception by distraction, AKA bait and switch. QED GEM of TKI kairosfocus

From the outset, Darwin realized that “organs of extreme perfection and complication”, such as the eye, posed a difficulty for his theory. Such features are much too complex to appear de novo, and he reasoned that they must evolve by incremental transitions through many intermediate states, sometimes undergoing changes in function.

Next time someone complains that Darwin is being taken out of context refer them to this paper. They fail to mention that Darwin didn't really think "organs of extreme perfection and complication, such as the eye" posed any difficulty for his theory. By the way, I think that complexity here is a bit of a red herring. The issue isn't just complexity, it's how everything works together to accomplish a purpose. Now it's wonderful how Avida can evolve an EQU function. But what goal does that serve? Mung

Elizabeth Liddle:

Avida is indeed sets up an “evolutionary” process, where “evolutionary” is being used in its original Darwinian sense of being a process by which a population of organisms reproduce with random variation, and those that are most successful in surviving the environment replicate most often.

I'd like to thank Elizabeth for providing a link to the paper. http://myxo.css.msu.edu/papers/nature2003/Nature03_Complex.pdf ...and those that are most successful in surviving the environment replicate most often.

...digital organisms compete for energy and, depending on the environment, can obtain energy by performing logic functions... ...Each digital organism obtained ‘energy’ in the form of SIPs at a relative rate (standardized by the total demand of all organisms in the population) equal to the product of its genome length and computational merit, where the latter is the product of rewards for logic functions performed.

See Table 1 Page 140. So if it meets a target or goal, specified in advance, programmed in advance, it gets rewarded. Why do some logic functions get rewarded more than others, i.e., why do some logic functions have more "computational merit" than others? Mung

And I suggest that the reason it looks as though I am saying something that to you doesn’t make sense isn’t because I am equivocating, but because you are “steeped in a way of thinking” that prevents you from seeing the simple truth of what I am saying

Well, there's a slight flaw in your reasoning. :) If that were so, I wouldn't be pointing out the different ways of looking at what you were saying, would I? So I see what you're saying just fine, thank you. I've rarely (though not never) charged you with making no sense. Rather, what I say is, what you are saying can be taken in more than one sense. Then you don't say no, it can only be taken in one sense. Rather, you typically tell me I misunderstood the sense in which you meant it. I disagree of course, though I may not bother to argue about it. I likely remind you that you are the one telling us you are striving to be clear and precise and unambiguous. Mung

Mung:

You are certainly not immune from the charge of taking out GA’s and using them as proof of what evolution can do.

No, indeed. I think they are an excellent example of what evolutionary processes can do. And I suggest that the reason it looks as though I am saying something that to you doesn't make sense isn't because I am equivocating, but because you are "steeped in a way of thinking" that prevents you from seeing the simple truth of what I am saying :) Anyway, I appreciate at least the courtesy you have extended me of granting that it is possible that I am honest ;) Kf: thank you also for your clarification: however, I must still rebut the charge. I do not equivocate, if equivocate is defined as Mung does above. I'm off to bed now. Sleep well guys :) Elizabeth Liddle

equivocate – to use ambiguous or unclear expressions, usually to avoid commitment or in order to mislead; prevaricate or hedge Perhaps you're the exception that proves the rule Lizzie. :) Let me be clear that I'm not saying you have those motives. I frankly believe you're steeped in a certain way of thought and that comes out in your posts as unreflective statements in which you use terms in a particular way because that's how you are used to using them. I'm trying to show you how that use can make it appear like you are saying something other than what you intended to say. For example, when you point out the wonderful things humans can create using "selection" without in any way trying to contrast the sorts of results that intentional/intelligent selection can bring about with those which non-intelligent and un-intentional "selection" can bring about, someone could well think you've made a valid point about what non-intelligent and un-intentional "selection" can do. But if they did so, it would be because of the equivocation involved, whether intentional on your part or not. At least Dawkins, when describing the Weasel program, told us, but evolution doesn't work that way. Which makes us wonder what the point was of the whole exercise. To me, it was exactly so that people would think that's how evolution works. The proof is certainly in the pudding, as they say. You are certainly not immune from the charge of taking out GA's and using them as proof of what evolution can do. Mung

PS: Avida is exactly NOT a metabolising, reproducing organism, that then evolves on differential reproductive success in the context of competition for scarce resources. The term evolution may have persuasive value in this context, but it is being used in two very different senses without due notice of the difference being made. kairosfocus

Dr Liddle: Pardon, I am not accusing, I am observing. Evolution is being used in drastically different senses in a context where that difference in sense is material. GEM of TKI kairosfocus

PS: Do you know how much functional specificity is involved in building a real NAND gate (as opposed to a simulation, which BTW requires the real deal to build the computer circuits)? 1: The saturable active element, e.g. a bipolar junction transistor [BJT], which requires a thin lightly doped base, sandwiched between emitter and collector with specialised contacts to give the electrical inputs and outputs. 3: this needs to be characterised as to its behaviour [especially parameter BETA or ALPHA, which are related], so we know when it moves to saturation zone. 4: A bias supply, from a DC power supply or a battery, itself a complex and functionally specific object. 5: A circuit with input and load resistors, of specific values, and the right polarity. 6: Here, we need two BJTs in series for the easiest design, each controlled by an input circuit. 7: Controlled inputs, to deliver the controlled outputs. If I were to see a NAND gate, I would immediately infer to design as its best explanation. Actually, just the transistor or a resistor alone would be enough. kairosfocus

kf: I have to say, I'm getting a little fed up of being constantly accused of "equivocation", especially if it is defined as Mung just defined it. I may be unclear; I may be mistaken; I may be confused; but I do not ever, ever, deliberately write to evade or obfuscate. Avida is indeed sets up an "evolutionary" process, where "evolutionary" is being used in its original Darwinian sense of being a process by which a population of organisms reproduce with random variation, and those that are most successful in surviving the environment replicate most often. The environment is simply an analog of an real life environment in which survival is a contingent of the ability to extract energy from it. My planned simulation is not, however, an evolutionary algorith, or, at least, it won't start that way. I hope one will emerge, and evolution will proceed from there. Elizabeth Liddle

Dr Liddle: The use of the term "evolutionary" in this context is an equivocation. GA's and the like can make hill-climbing tours of islands of function they were set up on, but the result is in no wise even relevant to the issue of creating 500+ bits worth of functionally specific complex information out of chance variations filtered by trial and error; as opposed to an intelligently designed algorithm. In your pending simulation, please make sure you do not tumble into that error. GEM of TKI kairosfocus

F/N: The EDM dissection of Avida is here. As background for the onlooker, NAND and NOR gates, in combination, can express any logic function, including memory [by incorporating digital, cross-coupled feedback, i.e the Flip Flop], and minimisation to express a function most economically is -- or has been -- a fairly common digital exercise. Avida seeks to implement various logic functions using a 'sea" of interconnected NAND gates. This is of course quite remote from the real-world task of a metabolising, von Neumann self-replicator based living cell. The term "organism," is little more than a sales gimmick, as is the claim that we are looking at "evolution." (Unfortunately, this seems to have been got away with over in Dover PA in 2005.) Sample clip:

Chips containing a sea of gates [6], [11], [13], all nand, are therefore capable of universal logic. This property allows evolutionary development of logic functions using the nand gate as the single logic component [1], [12], [25], [26], [32]. Avida [16], illustrated in Figure 1, performs logic synthesis using only the nand gate. The motivation behind Avida is not engineering design, but is rather to [16] “... show how complex [biological] functions can originate by random mutation and natural selection.” Avida generates an output equal to a logic combination of the inputs, X and Y. The logic operation under consideration that requires the most nand gates (five) is the XNOR. This is the ultimate goal of the search . . . . Avida uses a small alphabet of instructions (see Table I) to import the inputs, perform manipulation, and export the output. The instruction tape runs in a continuous loop. The number of entries can change during the search process . . . . Like choosing letters from the alphabet to form a sequence of words that will pass a spell check, the goal of Avida is to search for a sequence of instructions that has meaning with respect to the logic functions in Table II. The question is - how difficult is it to generate the logic functions in Table II [a table of 9 combinational logic functions] using the N = 26 instructions in Table I in a sequence of instructions?

While EDM give a more detailed analysis, the irrelevance, simplified and strawman nature of the task in view are readily apparent to the onlooker. kairosfocus

kf 54: yes, I know. I was addressing a different point - whether evolutionary processes can result in the writing of code. They can. It is, demonstrably, possible to have a computer program that writes good code, all by itself - in other words the code that is written by the computer program evolves - it is not input by the designer of the program. I quite agree that that does not address the problem of where the underlying code (the code in which the whole thing is written) came from, but that was not what I was claiming to provide in this thread. That is the subject of a project I hope to get started on in the next couple of weeks. Elizabeth Liddle

Avida is an example of an evolving population of algorithms that perform a function, using code. That code gets written, by Darwinian mechanisms, during the course of a run. Obviously the code in which the Environment is written doesn't evolve, nor does the code in which the organisms are written, but the code that the actual organisms contain is written, if not "by itself" by the process of random variation and replication with natural selection. Each digital organism, at the end, includes code that was not there at the beginning, and is there at the end, and performs a complex function. Who wrote it? I say evolutionary processes wrote it, and as those evolutionary processes were coded in the first place (by the AVIDA writers) then what they did was write code that wrote its own code - not the code that it was written in, duh, but the code that is created during the course of the run. BTW, I'm hoping for an response from you on the tennis thread, before the night is out :) Elizabeth Liddle

Dr Liddle: I am sorry, but the context of the concerns expressed should be quite clear after these many weeks. GA's and the like START within islands of very intelligently created function, and go through hill climbing exercises per design, basically uncovering implicit information already built in. The real issue is to get to islands of function by chance variation and selection on emergent function. Avida, the program most centrally discussed in the linked paper, is a case in point of the concern. GEM of TKI kairosfocus

The circuits they evolved were also cleaner and simpler than those designed by humans – I believe no one had ever managed to design a circuit like this where the fault checking was integral to the logic function rather than a separate circuit (that couldn’t detect faults in its self)

Actually you've described how they just did manage such a thing. Via the genetic algorithm they designed. Humans get the credit. Genetic algorithms give answers that we couldn't have come up with without their aid. But then so do algorithms that calculate the millionth digit of pi or report the millionth prime. There is nothing special about GA's in and of themselves. They are a tool that helps us realize the designs we are after, and couldn't possibly exist without our intelligent designing of them. But then so is a hammer, when it comes to building a house. We don't give hammers undeserved credit, so why GA's? Matteo

Elizabeth Liddle:

I was simply asked to provide an example of a computer code that writes itself, by means of random variation and selection. Avida does that,

Avida does not modify the code that it's written in. The Avida code modifies the digital organisms. It [Avida] does not modify it's own code! Avida is not an example of a computer code that writes itself, much less one that does so using random variation. Mung

And you'll have to forgive the irritable tone. I do get a bit fed up when people repeatedly accuse me of deliberate evasion and dishonesty. I make mistakes, but never deliberately, and I always correct myself as soon as I discover an error. I hope you will eventually come to realise that. Elizabeth Liddle

Mung, you may find my writing opaque, but I am not attempting to avoid commitment or to mislead, prevaricate or hedge. I'm finding you very hard to communicate with, but I guess I will keep trying. No, of course, EA didn't evolve Java. Nor does an organism that evolves camouflage also evolve DNA at the same time. It should be clear, with a little imagination, which level I am talking about. Obviously, I am talking about the level of code at which the digital organism performs the logic function, not the level of code in which the digital organism is written. And it is the code at that level that is randomly varied. Organisms do not evolve by randomly altering the bases from which DNA is constructed. Variation occurs at a higher level than that. Ditto with Avida - the variation occurs at the level of the virtual genome. If you read the linked paper, you would have understood what I meant. Actually, it seems pretty clear to me anyway. Elizabeth Liddle

Elizabeth Liddle:

That’s exactly what evolutionary algorithms do – evolve their own code.

You could do a better job of being less ambiguous. What do you mean by your claim that EA's evolve their own code? So take ev, for example. It's written in Java. It evolves the Java code that it's written in? It evolves the fitness function? It evolves the code that codes for the fitness function? And this is the way EA's always work, by loading their code into memory and randomly changing the memory locations and hoping that the EA still functions? No, I'm afraid not. So what on earth do you mean when you say that EA's evolve their own code? You don't really mean they evolve their own code, the code they are written in, do you? equivocate - to use ambiguous or unclear expressions, usually to avoid commitment or in order to mislead; prevaricate or hedge Mung

kf: it is not "bait and switch". I was simply asked to provide an example of a computer code that writes itself, by means of random variation and selection. Avida does that, and it does it on "differential functionality". I can't tell you how many bits, but I wasn't making any claims about how many bits. I guess I could find out. Elizabeth Liddle

Chris Doyle @ 7: "It’s like you know you can’t justify what you’re saying, but feel the need to say it anyway." Are you asking me to justify your claim that I can't justify my claims? paragwinn

Dr Lidle: Pardon me but these are as is directly stated, designed, digital "organisms" being simulated on a computer. This is yet another case of bait and switch. That may be effective sales tactics, but it is not serious evidence. In particular, we are not here seeing chance variation plus selection on differential functionality, writing complex information that fulfills real and specific function beyond 500 - 1,000 bits. GEM of TKI kairosfocus

kf:

Dr Liddle: You may believe that, but the problem is that here are so many serious issues of in effect digital code writing itself out of lucky noise filtered on function, that we would like to see an observational data point before taking such seriously. As Gil in effect pointed out in the OP. GEM of TKI

Well, the observational data are here: http://myxo.css.msu.edu/papers/nature2003/Nature03_Complex.pdf The case study figures are particularly informative. Elizabeth Liddle

That's a fair general point, CY (it is true that the word "evolution" is used in a great many senses) but in the OP Gil linked to, he was using it in a sense in which it is never (at least in my experience) used, namely, as a synonym for "development". And if he really was using it in the sense of "change", and thus covering "development", then his claim is completely irrelevant to the explanatory power of Darwinian theory. Darwinian theory isn't even about the way organisms change (although it accounts for it) but the way that populations do. Elizabeth Liddle

Lizzie, "In Gil’s case, if he thinks that his former thread about believing in evolution had anything to do with believing in evolution, then I think he’s missed something" I don't think Gil needs any defenders; he does well enough on his own. I really think it depends on what is meant by evolution, and that meaning is often vague among Darwinian evolutionists. At one moment it means microevolution and at the next moment it means macroevolution, and yet at another moment it means both. At least one of those examples of evolution I gather Gil accepts. In fact, I know that he does; so I don't think that charge is exactly fair. You have to understand what both creationists and ID people get accused of often by Darwinists, and it gets quite nasty. First of all, when a person mentions that they are a creationist (I'm not, but I think they get the brunt of the abuse) the insinuation goes like this: "Ok, well you don't believe in evolution then, which is quite surprising, since it's so obvious; things change." That's where it starts, and this obviously refers to that which no-one really disputes: microevolution. So when the creationist clarifies that he/she does in-fact believe in microevolution, that's where the abuse begins to spin out of control; because in using the term "microevolution," they have demonstrated their ignorance of evolution, because the term is not used anymore; so it is claimed. Suffice to say that creationists and some ID folk still use the terms because they do not accept the other part of evolution: extrapolating that small changes mean large changes. In other words, they believe that the evidence for microevolution is strong, while the evidence for macroevolution is forced from the fact of microevolution and a few other factors. So it really becomes a double standard when the Darwinian evolutionist is allowed to use microevolution (small changes) in such a way to point to macroevolution (large changes), while when the creationist or ID supporter uses microevolution to distinguish the two, they are accused of being ignorant of evolution. They are not ignorant of evolution; they just understand the distinctions a little better. That the Darwinists don't use the terms any more is misleading. SOME may not use the terms, but they use the concepts all rolled into one sort of like...I don't know, a bible as a weapon, so to speak? So I would say that Gil truly does believe in evolution, and he's not being disingenuous in stating so. CannuckianYankee

Dr Liddle: You may believe that, but the problem is that here are so many serious issues of in effect digital code writing itself out of lucky noise filtered on function, that we would like to see an observational data point before taking such seriously. As Gil in effect pointed out in the OP. GEM of TKI kairosfocus

Mung:

Elizabeth Liddle:

Avida doesn’t specify how the function is to be performed, and indeed, in any given Avida run, the digital organisms that perform the function can differ greatly in how they do it. In other words, there is no target solution, merely a range of problems to be solved.

Lizzie, you’re not doing yourself any favors. Read the sections on Avida in SitC. Chapter 13 p. 286 Chapter 13 fn46 p. 533 For the moment I’m going to avoid comment on what you’ve written and give you a chance to read up. You might want to do so before continuing to post more material on Avida. Cheers

I appreciate it Mung, but I'm quite happy for you to post your own critique here, now. I have read the cited pages and footnotes of Meyer's book, and nothing he says there renders my statement false. I didn't claim that Avida was a simulation of biology, nor did I claim that it was anything like as complex. I simply said it was an example of "code writing itself". It does. Yes, it does so within a highly designed environment, in which the replication machinery is already present (analogously to starting biological evolution with a functional living cell already in existence), but each digital organisms at the beginning of each run performs no function EXCEPT self-replication, and at the end we have a population of organisms that perform complex logic functions. That code that is contained within those organisms was programmed by nothing more than Darwinian processes. There is no target solution, only a target function, and as I said, every solution AFAIK is different. Elizabeth Liddle

kf:

Do genetic algorithms write themselves?

The environment doesn't, but the digital organisms, in effect do. Or rather, they are written by "Chance" and the rules of "Necessity" within that environment. Exactly as Darwinian theory postulates organisms are "written" in Nature. Elizabeth Liddle

Mung:

Lizzie: I’m still stuck at my computer, unfortunately. I’m trying to debug some code, You should try writing your code test first ;) Are you using Java? ANT, JUnit. Haha, even one written by a guy named Darwin: http://www.darwinsys.com/java/testfirstjava.pdf Google Test First Development, Test Driven Development and Test Driven Design. http://www.extremeprogramming......first.html http://c2.com/cgi/wiki?TestDrivenDevelopment http://www.agiledata.org/essays/tdd.html I can recommend some Java books if you want.

Thanks, but it's not Java, it's MatLab, and it's a stimulus presentation set, so I need to check the timings, given the refresh rate of the computer I'm using. So I need to run it in real time. Also there are some other problems in getting the triggers to talk to the data acquistion computer. Elizabeth Liddle

Elizabeth Liddle:

Avida doesn’t specify how the function is to be performed, and indeed, in any given Avida run, the digital organisms that perform the function can differ greatly in how they do it. In other words, there is no target solution, merely a range of problems to be solved.

Lizzie, you're not doing yourself any favors. Read the sections on Avida in SitC. Chapter 13 p. 286 Chapter 13 fn46 p. 533 For the moment I'm going to avoid comment on what you've written and give you a chance to read up. You might want to do so before continuing to post more material on Avida. Cheers Mung

DrBot:

Generally speaking any search has to have a target or it isn’t a search ...

Thank you, thank you, thank you. I can't believe how often I have to argue over this very simple fact (MathGrrl comes to mind). Mung

Lizzie:

I’m still stuck at my computer, unfortunately. I’m trying to debug some code,

You should try writing your code test first ;) Are you using Java? ANT, JUnit. Haha, even one written by a guy named Darwin: http://www.darwinsys.com/java/testfirstjava.pdf Google Test First Development, Test Driven Development and Test Driven Design. http://www.extremeprogramming.org/rules/testfirst.html http://c2.com/cgi/wiki?TestDrivenDevelopment http://www.agiledata.org/essays/tdd.html I can recommend some Java books if you want. Mung

Elizabeth Liddle @12:

Well, yes, they do. Or can. That’s exactly what evolutionary algorithms do – evolve their own code.

Like Avida? Mung

How much does stupid cost? Because I've been getting it for free for years now. ;) Mung

Do genetic algorithms write themselves? kairosfocus

Do your parallel processing algorithms write themselves, if so, how?

No. Why do you ask? its not like I ever claimed that they were (and btw, they are not 'mine' - I was referring to research done by others) Of course you could use a genetic algorithm to design the code, but biology doesn't use algorithms, it uses neural circuits, so you could try evolving the processing circuitry. If you were attacking a low level vision task like optic flow field sensing then the basic mechanism is quite simple - a photo receptor and a few transistors repeated n times across a surface - quite easy to see how something like this can evolve when you understand the processes involved. DrBot

And, of course the ornithopter software being discussed was sure not written by monkeys at keyboards, fileted by trial and error and fed into hill climber algors.

No, but genetic algorithm could have been employed as part of the process. DrBot

KF:

I can imagine someone going into a board-room and proposing to write software by putting monkeys at keyboards, filtered through trial and error,t hen when one works, it is fed into a hill-climber algorithm. No-one in his right mind would walk into a boardroom with such a proposal.

No, that would be crazy, what you described would never work, but there are a number of companies founded on the idea of using genetic algorithms to generate designs, be they code or neural networks, or electronic circuits. They get venture capital funding because they can demonstrate that the techniques they use deliver results. DrBot

Joseph: Joseph:

Elizabeth:

This program evolves, through Darwinian mechanisms, result in a program that is quite different, and has different functionality, from the original human-written original.

It didn’t say that.

It may not have done, but that's what it does.

Populations of digital organisms often evolved the ability to perform complex logic functions requiring the coordinated execution of many genomic instructions.

http://myxo.css.msu.edu/papers/nature2003/Nature03_Complex.pdf The initial seeded digital organisms perform no logic functions. Their only function is self-replication. What evolves are digital organisms that incorporate code that performs complex functions.

And I would bet Avida uses a targeted search.

As in WEASEL? No, it doesn't. The fitness function simply requires that the organism performs a function. The more complex the function, the more "energy" it receives (analogous to an organism evolving functions that enable it to catch more nutritious, if trickier, prey). Avida doesn't specify how the function is to be performed, and indeed, in any given Avida run, the digital organisms that perform the function can differ greatly in how they do it. In other words, there is no target solution, merely a range of problems to be solved. Elizabeth Liddle

People I know who have done this have often remarked how “clean” and compact the resulting code often is, compared with the code produced by human programmers.

I knew someone who used a GA to evolve logic circuits that included integral fault checking systems - so the circuit would flag a fault if any part of the circuit developed a fault (including the fault checking part) (one of their papers is here.) The circuits they evolved were also cleaner and simpler than those designed by humans - I believe no one had ever managed to design a circuit like this where the fault checking was integral to the logic function rather than a separate circuit (that couldn't detect faults in its self) joseph @ 26 - by targetted do you mean simply chasing and trying to match a set of criteria, or searching for something that has been specified in advance? Generally speaking any search has to have a target or it isn't a search, but 'targetted' means something more specific - pre-specified and unchanging (I believe - but I could be wrong) DrBot

Joseph:

Elizabeth: There is plenty of evidence that Darwinian process can result in new, useful and functional multi-part systems.

That is false.

Well, I don't think it is false! We know the kinds of phenotypic effects that incremental variants of regulatory genes do to the expression of hox genes, and why this results in potentially large changes to, for example, the numbers of limbs, the length of a limb, the number of bones in a limb, the number of digits, etc. And I gave you a reference to a paper in Nature that describes the evolution of a whole new body part (the helmet of tree hoppers).

BTW Darwinian mechanisms cannot account for HOX genes. And you cannot use what needs an explanation in the first place to do the explaining.

Why can't Darwinian mechanisms account for Hox genes? Elizabeth Liddle

Elizabeth:

This program evolves, through Darwinian mechanisms, result in a program that is quite different, and has different functionality, from the original human-written original.

It didn't say that. And I would bet Avida uses a targeted search. Joseph

Elizabeth:

There is plenty of evidence that Darwinian process can result in new, useful and functional multi-part systems.

That is false. BTW Darwinian mechanisms cannot account for HOX genes. And you cannot use what needs an explanation in the first place to do the explaining. Joseph

Joseph @ #19, from your quotation:

The virtual environment is initially seeded with a human-designed program that self-replicates. This program and its descendents are then subjected to random mutations of various possible types which change instructions within their memory; resulting in unfavorable, neutral, and favorable program mutations. Mutations are qualified in a strictly Darwinian sense; any mutation which results in an increased ability to reproduce in the given environment is considered favorable. While it is clear that the vast majority of mutations will be unfavorable—typically causing the creature to fail to reproduce entirely—or else neutral, those few that are favorable will cause organisms to reproduce more effectively and thus thrive in the environment.

In other words, the environment is seeded with a very simple, human-written program. This program evolves, through Darwinian mechanisms, result in a program that is quite different, and has different functionality, from the original human-written original. Apart from demonstrating biological evolution, this kind of environment is genuinely useful. We can start out with extremely basic code, and let something far more complex and clever evolve. People I know who have done this have often remarked how "clean" and compact the resulting code often is, compared with the code produced by human programmers. Elizabeth Liddle

Joseph:

And Liz- Darwinian processes do have limits- there isn’t any evidence that they can construct new, useful and functional multi-part systems. Why doesn’t that count against it?

Well, it would be if there wasn't any evidence but there is! There is plenty of evidence that Darwinian process can result in new, useful and functional multi-part systems. That evidence may not amount to any one complete specification of any one complete system (I don't know about that) but there is plenty of evidence of the kind of mechanisms that would enable it to do so, most importantly being the discovery of hox genes. And there was a very interesting recent paper here: http://www.nature.com/nature/journal/v473/n7345/full/nature09977.html?WT.ec_id=NATURE-20110505 Elizabeth Liddle

Well, CY, when a smart person believes something that I think is stupid, I want to know why :) In other words, it gives me pause to consider that I might have missed something. In Gil's case, if he thinks that his former thread about believing in evolution had anything to do with believing in evolution, then I think he's missed something :) And conceivably, given that he is a smart person, that might explain why he thinks that evolution is stupid. In my own case, several people have pointed me to Meyer's book. So I'm reading it. I want to know whether I've missed something. If I have, I want to know what it is. Equally, if Meyer has, I want to know what it is. Elizabeth Liddle

Lizzie, People believe stupid things all the time. Believing in stupid things does not necessarily make the believer stupid. I don't think you believe anybody here is stupid, but I gather that you believe that ID is one of those stupid things. CannuckianYankee

And Liz- Darwinian processes do have limits- there isn't any evidence that they can construct new, useful and functional multi-part systems. Why doesn't that count against it? Joseph

Elizabeth:

It doesn’t evolve the code that runs the code, of course, because it’s a simulation, but what evolves is code – algorithms that take input and produces output.

Avida:

The avida system creates an artificial (virtual) environment inside of a computer. The system implements a 2D grid of virtual processors which execute a limited assembly language; programs are stored as sequential strings of instructions in the system memory. Every program (typically termed cell, organism, string or creature) is associated with a processor, or grid point. Therefore, the maximum population of organisms is given by the dimensions of the grid, N× M, and not by the size of the total genome space of the population, as in tierra. For purposes of Artificial Life research, the assembly language used must support self-reproduction; the assembly language instructions available are described in the Virtual CPU section. The virtual environment is initially seeded with a human-designed program that self-replicates. This program and its descendents are then subjected to random mutations of various possible types which change instructions within their memory; resulting in unfavorable, neutral, and favorable program mutations. Mutations are qualified in a strictly Darwinian sense; any mutation which results in an increased ability to reproduce in the given environment is considered favorable. While it is clear that the vast majority of mutations will be unfavorable---typically causing the creature to fail to reproduce entirely---or else neutral, those few that are favorable will cause organisms to reproduce more effectively and thus thrive in the environment. Over time, organisms which are better suited to the environment are generated that are derived from the initial (ancestor) creature. All that remains is the specification of an environment such that tasks not otherwise intrinsically useful to self-reproduction are assimilated. A method of altering the time slice, or amount of time apportioned to each processor, is described in the Time Slicing section. While avida is clearly a genetic algorithm (GA) variation (to which nearly all evolutionary systems with a genetic coding can be reduced), the presence of a computationally (Turing) complete genetic basis differentiates it from traditional genetic algorithms. In addition, selection in avida more closely resembles natural selection than most GA mechanisms; this is a result of the implicit (and dynamic) co-evolutionary fitness landscape automatically created by the reproductive requirement. This co-evolutionary pressure classifies avida as an auto-adaptive system, as opposed to standard genetic algorithms (or adaptive) systems, in which the creatures have no interaction with each other. Finally, avida is an evolutionary system that is easy to study quantitatively yet maintains the hallmark complexity of living systems.

Joseph

Hi Gil! Pleased to meet you finally!

Elizabeth Liddle: You’ve said a number of times in your posts that believing in evolution is stupid.

Actually, I’ve said that I do believe in evolution. Remember?

But that wasn't evolution, Gil! That was development! Evolution is something that happens to populations, not individuals, although of course it affects the way individuals develop.

What I don’t believe in is the near infinite creative power ascribed to the proposed Darwinian mechanism.

But nobody I know ascribes "infinite creative power" to any Darwinian mechanism. Indeed some of the most powerful evidence for Darwinian processes, as opposed to other kinds of processes, is that Darwinian process have limits, and those limits are exactly the limits we see in the nested hierarchies of living things - the almost total restriction of novelty to a single lineage, regardless of how useful that novelty might be in another. I'd say it's the very limitations of Darwinian processes (lack of foresight; inability to transfer solutions from one lineage to another) that make its most powerful differential predictions compared to what you might expect from an Intentional Designer (as exemplified by human designs). If you think what I have written above is "stupid", please explain why :) I may not be brilliant, but I'm not normally considered stupid. Nor am I anti-theist. I was a theist for half a century, and in some ways I still miss it. So I have no dog in that fight. Elizabeth Liddle

Hiya Lizzie, I hope you took up the narrowboat option instead of staying in and battling out here all day!

Hi Chris! Hope the weather stayed fine for the fete! I'm still stuck at my computer, unfortunately. I'm trying to debug some code, but have to run it in real time, (it's code for presenting stimuli during an experiment). I hope we'll get up the Soar for an evening barbecue on the riverbank :)

Meyer deals with Avida in SITC. Once you get to it, I hope you’ll agree that it doesn’t actually offer any evidence for neo-darwinian evolution whatsoever. On the contrary, it’s more like evidence for Intelligent Design.

OK, I'm reading approximately one chapter per night, so I'll get there eventually :) Elizabeth Liddle

paragwinn: When are you going to find something new to post about? My interest in ID is the product of my background in multiple engineering disciplines, so I often emphasize the argument from engineering. Keep in mind that UD has a continuous supply of new visitors who might not be familiar with such argumentation. DrBot: I would disagree with Gil when he says that this is stuff we can only dream about. Actually, I agree that we will probably one day develop the technology required to produce an autonomous, self-contained ornithopter with the capability of landing on a swaying tree branch in gusty wind, but this represents a phenomenally sophisticated and difficult engineering task. Engineers do indeed dream about such things as technology advances through intelligent design. But this technology has already been developed! And it wasn't developed through random mutation and natural selection. Elizabeth Liddle: You’ve said a number of times in your posts that believing in evolution is stupid. Actually, I've said that I do believe in evolution. Remember? What I don't believe in is the near infinite creative power ascribed to the proposed Darwinian mechanism. GilDodgen

Hiya Lizzie, I hope you took up the narrowboat option instead of staying in and battling out here all day! Meyer deals with Avida in SITC. Once you get to it, I hope you'll agree that it doesn't actually offer any evidence for neo-darwinian evolution whatsoever. On the contrary, it's more like evidence for Intelligent Design. Chris Doyle

Yes, Avida. It doesn't evolve the code that runs the code, of course, because it's a simulation, but what evolves is code - algorithms that take input and produces output. That code is not provided by the Avida designers, but evolves within the Avida environment. Elizabeth Liddle

Elizabth:

That’s exactly what evolutionary algorithms do – evolve their own code.

Any examples of that? Joseph

Gil:

It’s really easy to figure out that the Darwinian mechanism of random mutation and natural selection cannot possibly do that with which it is credited. Life is fundamentally based on information and information processing — a software computer program and its associated, highly functionally integrated execution hardware. Computer programs don’t write themselves,

Well, yes, they do. Or can. That's exactly what evolutionary algorithms do - evolve their own code.

and they especially don’t write themselves when random errors are thrown into the code.

But they do - that's how they work. If you don't set the thing up to give "random errors" i.e. you set it up so that replication happens with 100% fidelity, your code won't evolve.

The fact that biological computer programs can survive random errors with remarkable robustness is evidence of tremendously sophisticated fault-tolerance engineering. The same goes for the hardware machinery of life.

Well, it is evidence that only robust code is able to evolve, which would rule out a lot of human-written code. In other words, evolution can only occur in a self-replication system in which a large subset replication variants are viable. You've said a number of times in your posts that believing in evolution is stupid. Obviously I disagree, but does it never occur to you that if a lot of apparently intelligent people seem to think it makes sense, that it is not as "obviously" stupid as you seem to think? Remember that DNA "code" is a single vector. Your own code is not. A code that consists of a single vector, in which any three consecutive items potentially codes for a viable output, is clearly much more robust than a code in which the slightest misplaced bracket or mangled bit of syntax will cause the code to fail. Elizabeth Liddle

Dr Bot: Do your parallel processing algorithms write themselves, if so, how? GEM of TKI kairosfocus

Gil: You nailed it. Well done. I can imagine someone going into a board-room and proposing to write software by putting monkeys at keyboards, filtered through trial and error,t hen when one works, it is fed into a hill-climber algorithm. No-one in his right mind would walk into a boardroom with such a proposal. Unless he or she wanted to be fired. And, of course the ornithopter software being discussed was sure not written by monkeys at keyboards, fileted by trial and error and fed into hill climber algors. No ifs, ands or buts about it. What have we been collectively smoking? Whatever it is, it makes the strongest sensi I have seen pale beside it. I think we have been in a cave full of laced incense and fed on clever stories that the bewitching mind-drugs have made seem real. Time to wake up, folks. (Detractors: cf the problem of filter psychology [cf here and here (this last on emotional blocks)], for messages apt to be filtered out, they do need to be said, in slightly different ways with diverse examples, over and over and over again to begin to get through.) GEM of TKI kairosfocus

Gil, The second law is the "common sense law of physics", see my video beginning at the 17:45 mark. But people like Daniel Styer (see video beginning at 3:45, also here ) have perverted it beyond recognition, into something very counterintuitive. Common sense is not always good science, but in this case it is, and the mathematics (video beginning at 8:45 mark) actually supports the common sense interpretation, not the perverted one. Granville Sewell

The task of designing GN&C algorithms and the associated hardware that would permit an ornithopter to land on a swaying tree branch in gusting wind is so far ahead of our most sophisticated human technology that we can only dream about such a thing. Yet, birds do this with ease.

Some of the latest research is quite impressive. The control software for that kind of precise flying is good enough to perform those kinds of maneuvers, the problems that remain are more about the computation required for visually guided behavior like this. Here is a video of a flying robot performing some very precise stunts. The flight stability, and general control are all done on board, but the visual navigation is done off board because it is so computationally demanding. It's worth noting though that by computationally demanding I don't mean it is beyond our capabilities, it is actually more to do with the nature of the computers we use. Vision is an inherently parallel task but we tend to do it with sequential machines. I believe some research groups are working on vision sensors that perform some of the vision processing in situ - each pixel is a simple processing unit that communicates with near neighbors so the 'computer' is actually an array of thousands of identical processing units working at a relatively low speed. In other words you have thousands of simple computers working at a few Khz instead of a single complex core that has to work at many Ghz. The fly eye is a good example that some robotics researchers are drawing inspiration from. I would disagree with Gil when he says that this is stuff we can only dream about. DrBot

You make lots of unsubstantiated claims, paragwinn. When challenged to support them, you disappear. It's like you know you can't justify what you're saying, but feel the need to say it anyway. Is that what it is? Chris Doyle

Chris Doyle: you mean the claims of repetition and lack of new information? paragwinn

Great post GilDodgen, I loved the point about designing something to land on a branch. We take it for granted, but we shouldn't. Paragwinn: when are you going to start substantiating your claims? Chris Doyle

B1 paragwinn

GilDodgen, When are you going to find something new to post about? Quite literally, the repetition is providing no new information. paragwinn

Better link: God Is God http://www.youtube.com/watch?v=l_GQsVQikXE bornagain77

But alas Gil, the alternative is 'unthinkable'; :) GOD is GOD - Steven Curtis Chapman http://www.youtube.com/watch?v=oCrBczelMwM&NR=1 bornagain77