652 thoughts on “Evolving Wind Turbine Blades”

  1. petrushka: Now when someone in the ID movement comes up with an actual sequence of mutational changes in actual evolutionary history that could not happen via known mutational processes, I’ll listen.

    “If it could be demonstrated that any complex organ existed, which could not possibly have been formed by numerous, successive, slight modifications, my theory would absolutely break down.”

    ― Charles Darwin, The Origin of Species

    Some things never change.

  2. Mung: It’s not searching for anything, which is why it doesn’t find anything. Right?

    How do you know when it’s found something?

  3. Rumraket,

    Its entirely NOT how evolution supposedly works. Geez, you guys are sure fooled easily. They are all picked for the exact same reason. Its like if you had a program whose job is was to pick the longest piece of string. Then you gave it a bunch of random pieces of string, and let it pick the longest one. You are right, it would pick the longest one! Yeah!! Look at the genius of the algorithm. It was able to find the longest piece. How did it know, it wasn’t told which was longest!! It figured out the solution all by itself.

    What the program is NOT doing is picking for MORE than one reason-you know, like evolution does. If the only criteria for being the best tiger was one need, being the fastest for example, then yes, I guess over time you would end up with the fastest tigers. But if the criteria is multi-facted, you are not choosing for anything. Sometimes fast is good, sometimes slow is good. Sometimes dumb is good, sometimes smart is good. If one round gets chosen for being smart because that is good, but the next round it gets chosen for being dumb, because that is also good, you get nowhere! Smart dumb tigers? Slow fast tigers? Big leg-small leg tigers? Cold-blooded-warmblooded tigers? Fast dumb little leg tigers with cold blood, and big ears and three lungs and no spleens and top teeth which are for tearing meat, and bottom beaks for making it more streamline in flight.

    There are many strategies for surviving. If there was only one, it would be easy to chose.

  4. Richardthughes: Did you have a point?

    Did you have a point to make in the OP? I couldn’t find one.

    This should get some good discussion / denial!

    That’s quite the stance you’ve taken there Richardthughes! Why not tell us what you really think?

  5. phoodoo: It was able to find the longest piece.

    Actually, It didn’t. Did you watch the video? Non convergent runs. And how many permutations are there? I don’t think you understand the problem.

  6. Mung,

    We’ll the program seemed to evolve a better solution than an Intelligently designed one. 😉

    Oh, buy my wristband.

  7. Joe Felsenstein: Those who say that genetic algorithms are not relevant to discussions of evolution say repeatedly that this is because the final goal must be stored in the algorithm.

    It sure helps if you want to demonstrate the power of cumulative selection!

  9. Frankie:

    Eliminating is not the same as selecting

    It leads to exactly the same result, so that is irrelevant. Positively choosing a fraction of a population to breed is exactly the same, in effect, as eliminating ‘the other’ fraction from the pool instead.

  10. Mung: “If it could be demonstrated that any complex organ existed, which could not possibly have been formed by numerous, successive, slight modifications, my theory would absolutely break down.”

    ― Charles Darwin, The Origin of Species
    Some things never change.

    So do you have an example of such an entity?

  11. phoodoo: There are many strategies for surviving. If there was only one, it would be easy to chos

    Of course there are, that’s precisely why evolution is not a search. The thing is that, in these GA’s, even with simple fitness functions the result doesn’t converge to one unique solution. If the simulation allowed for branching, simulating speciation, it would explore multiple parallel routes. That would be computationally heavy of course

  12. phoodoo: What the program is NOT doing is picking for MORE than one reason-you know, like evolution does. If the only criteria for being the best tiger was one need, being the fastest for example, then yes, I guess over time you would end up with the fastest tigers. But if the criteria is multi-facted, you are not choosing for anything. Sometimes fast is good, sometimes slow is good. Sometimes dumb is good, sometimes smart is good. If one round gets chosen for being smart because that is good, but the next round it gets chosen for being dumb, because that is also good, you get nowhere! Smart dumb tigers? Slow fast tigers? Big leg-small leg tigers? Cold-blooded-warmblooded tigers? Fast dumb little leg tigers with cold blood, and big ears and three lungs and no spleens and top teeth which are for tearing meat, and bottom beaks for making it more streamline in flight.

    There are many strategies for surviving. If there was only one, it would be easy to chose.

    I agree ,it is so confusing not having a black and white dichotomy. If only scientists could construct something to control all variables but one, oh well ,that is just being ridiculous.

  13. petrushka,

    The limitations of a model are projected onto the original object or system.

    Also, the fallacy of over-extension (which I made up). Saying essentially, if a more efficient result cannot be achieved in a specific direction in a specific model, it cannot be achieved period.

  14. Allan Miller,

    It leads to exactly the same result, so that is irrelevant

    No, it doesn’t.

    Positively choosing a fraction of a population to breed is exactly the same, in effect, as eliminating ‘the other’ fraction from the pool instead.

    But that isn’t how it works in the real world. In the real world only a small fraction gets eliminated as opposed to a small fraction being selected. Mayr goes over the distinction in “What Evolution Is”:

    Do selection and elimination differ in their evolutionary consequences? This question never seems to have been raised in the evolutionary literature. A process of selection would have a concrete objective, the determination of the “best” or “fittest” phenotype. Only a relatively few individuals in a given generation would qualify and survive the selection procedure. That small sample would be only to be able to preserve only a small amount of the whole variance of the parent population. Such survival selection would be highly restrained.

    By contrast, mere elimination of the less fit might permit the survival of a rather large number of individuals because they have no obvious deficiencies in fitness. Such a large sample would provide, for instance, the needed material for the exercise of sexual selection. This also explains why survival is so uneven from season to season. The percentage of the less fit would depend on the severity of each year’s environmental conditions.

    Yeah I know that you don’t care what the experts have to say. Too bad their word means something and your objection doesn’t

  15. Richardthughes: We’ll the program seemed to evolve a better solution than an Intelligently designed one.

    It would appear then, that the program was searching for a solution to a problem, and the problem to be solved was provided by the designer of the program.

    Do you deny this rather obvious fact?

  16. dazz: Of course there are, that’s precisely why evolution is not a search. The thing is that, in these GA’s, even with simple fitness functions the result doesn’t converge to one unique solution. If the simulation allowed for branching, simulating speciation, it would explore multiple parallel routes. That would be computationally heavy of course

    Right! Undirected/ unguided evolution is not a search. However GAs are search heuristics actively searching for a solution to the problem they were designed to solve.

  17. As a side note – killing active information. I’m sure it’s possible for a GA to create a solution bigger than the GA’s code, given enough computational resources. What does this mean for CSI?

  18. We’ll the program seemed to evolve a better solution than an Intelligently designed one.

    The program intelligently designed a solution. The program just takes the place of humans actively doing the work.

  19. Mung: It would appear then, that the program was searching for a solution to a problem, and the problem to be solved was provided by the designer of the program.

    Do you deny this rather obvious fact?

    This is the nature of simulations, Mung. Supposing though we call it a “sandbox” and let it just do its thing based on the same parameters. Then what?

  20. phoodoo: Its entirely NOT how evolution supposedly works. Geez, you guys are sure fooled easily. They are all picked for the exact same reason. Its like if you had a program whose job is was to pick the longest piece of string.

    First you say:

    phoodoo: This is an example of running a million different shapes through a wind tunnel and seeing which has the highest rating. So what!

    I respond:

    : So that’s how evolution works. Millions of different, randomly varying organismal features are run through the “wind tunnel” of the real world, and the best reproducing ones go on for new rounds of millions of random variations.

    And now you start blathering about multiple traits being selected for simultaneously. I know that real evolution actually selects multiple traits simultaneously, which is why I talked about the real world. Nevertheless, the simulation still mirriors real evolution because it has the fundamental aspects: Random variation and selection. It doesn’t matter that only one aspect is selected for, it’s still generated randomly and then selected.

    Did you know there are simulations that simulate multiple selective pressures simultaneously? They still work. Evolution still works even when you add multiple selection criteria simultaneously. This often leads to niche formation. It’s why Darwin’s Finches have split into niches with different beak adaptations.

  21. dazz: The thing is that, in these GA’s, even with simple fitness functions the result doesn’t converge to one unique solution.

    So?

  23. Richardthughes: Supposing though we call it a “sandbox” and let it just do its thing based on the same parameters. Then what?

    Then it would be more like evolution.

  24. Mung: So?

    So it’s not a search. If you think it is, how do you know when you’ve found the “target?”

  25. phoodoo,

    This is an example of running a million different shapes through a wind tunnel and seeing which has the highest rating.

    No, that would be more like random search. This is not random search.

  27. dazz: So it’s not a search. If you think it is, how do you know when you’ve found the “target?”

    ALL GAs are search heuristics. If it isn’t a search then it isn’t a GA

  28. Mung,

    Some things never change.

    Because Darwin’s implicit challenge has yet to be met. Although I don’t think it could be met even in principle, in fact, regardless of the truth or otherwise of evolution.

  30. Frankie: ALL GAs are search heuristics. If it isn’t a search then it isn’t a GA

    You and Mung keep asserting stuff but provide no substance. Answer this, if it’s a search, what is it searching for and how do you know you’ve found it?

  32. Allan Miller: Although I don’t think it could be met even in principle

    I think it could be met in principle.

    We use Darwin’s principle in DNA forensics to determine guilt or innocence. There are controversies about how many data points it takes to demonstrate identity or relatedness, but we make life and death decisions based on the likelihood that people are the same or related. It is a narrow example of using the criterion of small changes. When this is extended, we have evidence for common descent.

  33. dazz,

    Look up the definition of “genetic algorithm”. It will contain the words “search heuristic”.

    In this case it is searching for efficient wind turbine blades and it knows it found it when the specifications are met

  34. Allan Miller:
    Mung,

    Because Darwin’s implicit challenge has yet to be met. Although I don’t think it could be met even in principle, in fact, regardless of the truth or otherwise of evolution.

    Darwin’s challenge has been met- irreducible complexity blows it away. And all you have are promissory notes to answer IC

  35. Mung,

    It sure helps if you want to demonstrate the power of cumulative selection!

    You could just as easily easily demonstrate the power of cumulative selection in an algorithm lacking a hard-coded target. Such as … ooh … this one! Or, for that matter, a modified WEASEL with a runtime-supplied or random target.

  37. phoodoo,

    Yes, it’s not got all of biology in it, dammit! Because if it did – if it had billion-base genomes and translation and folding and development and realistic ecologies and reproductive isolation and stuff … if it had all that, then I, phoodoo, declare that it would not work! Prove it? No, why would I need to prove it? It’s obvious!

  39. Allan Miller:
    Mung,

    You could just as easily easily demonstrate the power of cumulative selection in an algorithm lacking a hard-coded target. Such as … ooh … this one! Or, for that matter, a modified WEASEL with a runtime-supplied or random target.

    No, the weasel algo was a search. This is not, and you can’t respond to how do you know you’ve found the target (or targets)

  40. Frankie: In this case it is searching for efficient wind turbine blades and it knows it found it when the specifications are me

    Bzzzzzzt. False. From the YouTube transcript: “4:32I let the simulation run for quite a long time”

  42. Frankie: it knows it found it when the specifications are met

    Yes, but the algo could be run forever and you would never be sure there’s a better solution. Obviously to be of any use for engeneers one needs to stop the process at some point, but that’s arbitrary and not part of the algo itself. In nature it never stops, there’s no way to get to an “optimal” organism because evolution never stops

  45. petrushka,

    I think it could be met in principle.

    The problem is in eliminating all incremental paths. I think combinatorial explosion soon takes over. Particularly when multi-base mutational events are included. They are essentially irreversible and unrepeatable, but if you could see how the change was done, you could easily find you’d violated no limitation on probability in first doing it.

    One would need a proper view of the probabilistic landscape of all regressive pathways before one could declare a current structure incrementally ‘impossible’ from any precursor state. It’s close to being expected to prove a negative.

    It’s at about this point in the conversation we usually start talking about the flagellum.

  46. Allan Miller,

    No, the problem is that you have no way to test your claims. You can’t show that any incremental path can do it and science only allows so much luck. Unfortunately all you have is luck to find the proper path if one even exists

  47. I don’t mean you can eliminate all possible paths. I mean that given X differences and Y time, one could infer the probability that X occurred via small incremental change.

    I think this is already a hot topic of research.

    To bad the ID community isn’t doing the research.

  48. I always get a kick out of watching people try to deny that GAs have any relevance to evolution. The early designers of the same took as faithful a representation of the natural process as practical, generating variation, copying and an environmental reason for variable copy numbers, and discovered that it made a powerful tool – for both search and optimisation applications.

    This thing that doesn’t work, that’s tautological, materialistic garbage, that allows contradictory attributes to be ‘the most fit’ depending on circumstances … turns out it works, algorithmically. What an enormous stroke of luck!

Leave a Reply