652 thoughts on “Evolving Wind Turbine Blades”

  2. stcordova: never in the lab or field a mode change via small steps like a turbine going from vertical to horizontal

    So you are going to ignore the case of a logic circuit evolving to take advantage of the analog properties of the chip it is embodied on, something not anticipated by the GA design?

  3. dazz: Not true at all. All the program does is to produce random variations and apply selection based on fitness, using the physics simulation to get a performance figure. If the specification was in the program, it would converge to the same specification from any starting point, did you miss Prof. Felsenstein post?

    It depends on how “specified” you want your specified. From a blueprint all the way down to “good”, or “best performing”… even then of course there is no guarantee a GA hits the universal maxima.

  4. stcordova: Lynch argues in great detail in his Genome Architecture book that many features of Eukaryotic genomes must have emerged without the influence of selection as a matter of principle

    By fixation of neutral alleles maybe?

    stcordova: we clearly see bacteria have one mode of evolution (optimization of antibiotic resistance )

    What about motility? And does bacteria like bifidus, which is beneficial for the host organism, evolve towards antibiotic resistance too? It’s an honest question, I have no idea

    stcordova: Like the wind turbine example, we clearly see bacteria have one mode of evolution (optimization of antibiotic resistance ), but never in the lab or field a mode change via small steps like a turbine going from vertical to horizonta

    Well, If I got it right he defined the specification using 2D spines in the horizontal plane, it simply couldn’t go vertical without adding parameters for a 3rd dimension

  5. Richardthughes: It depends on how “specified” you want your specified. From a blueprint all the way down to “good”, or “best performing”… even then of course there is no guarantee a GA hits the universal maxima.

    I realize a simulation is going to have a finite set of possible outcomes, but it’s my understanding that for all intents and purposes it can be considered infinite in the sense that one could never evaluate every possible specification in a true search for the optimal one, and that’s why a GA is used instead.

    Seems to me that in a GA with a huge set of possible outcomes, there’s never a point where one can be sure it’s reached the optimal solution. If the GA is kept running a better solution may always emerge, particularly if the random variation allows to overcome getting stuck in local maximums (recombination?)

    For all I know, in nature there’s no such thing as an optimal organism

  6. This is just a slightly more sophisticated version of mistaking the map for the territory.

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

    I will not violate the site rules by suggesting that anyone is dishonest or arguing in bad faith.

    But this line of reasoning is moronic.

    It is possible to deceive oneself without lying to others.

  7. dazz: For all I know, in nature there’s no such thing as an optimal organism

    I’d agree, because the optimal is contextualized by environment and evolving competition. Top of the food chain a million years ago was very different, and I’m not arguing that being top is the optimal strategy!

  8. petrushka,

    Some would / have argued that any simulation is guided because people built it. By extension this also makes experiments all “evidence for design” because the experiments were designed. Gil at UD seemed to have troubles with this.

  9. Richardthughes: Top of the food chain a million years ago was very different, and I’m not arguing that being top is the optimal strategy!

    Bacteria are, and may always have been, at the top of the food chain.

  12. Frankie:

    I don’t deny microevolution. I don’t deny macroevolution. Obviously Allan Miller has personal issues

    Scroll up, sunshine. The statement to which “phoodoo and Frankie” was the answer was “this is a good example of microevolution. Who here denies this?”.

    You appear to deny it in this very thread, as you have denied in every thread a GA is ever mentioned.

    Do you in fact think it is a good example of microevolution?

  13. petrushka: So you are going to ignore the case of a logic circuit evolving to take advantage of the analog properties of the chip it is embodied on, something not anticipated by the GA design?

    The design would never be put into production. And I bet a human could design a better circuit. The circuit the GA was pitted against was not the best circuit available. It was just one patented circuit. And the GA produced what it did because it was designed to do so.

  15. Richardthughes :

    A man may fish with the worm that hath eat of a
    king, and cat of the fish that hath fed of that worm.

  17. Frankie:
    dazz,

    Random variations of what? How does it know what to select and actively search for?

    Of the length of the spines that determine the shape of the blades. Did you even bother to watch the vid?

    The GA simulates descent with modification (mating & mutation) and natural selection (evaluating the performance of each result), and is not directed in any way shape or form.

    It’s not a “search”, it just keeps selecting the fittest blades and keeps applying mating & random variation

  19. stcordova,

    If I were an evolutionist, I’d probably be leaning on the side of self-organization happening in modes mostly free of selection.

    I lean on the side of organs arising rapidly through selective tuning of an initially small collection of relatively undifferentiated cells. This would be expected to provide rapid appearance. Inevitably, all the stages at which a proto-organ is not detectable as such would be cryptic.

  20. dazz,

    Of the length of the spines that determine the shape of the blades

    That contradicts what you said earlier.

    The GA simulates descent with modification (mating & mutation) and natural selection

    Not natural selection. GAs are a search heuristic actively searching for a solution, Natural selection is not a search heuristic and it isn’t active nor searching

    ALL GAs are search heuristics. It is their very definition

  21. Frankie,

    That contradicts what you said earlier

    What?

    Not natural selection. GAs are a search heuristic actively searching for a solution, Natural selection is not a search heuristic and it isn’t active nor searching

    Yes, natural selection by evaluating the performance of the blade with a software simulation and picking the “fittest” while discarding the less “fit”

    ALL GAs are search heuristics. It is their very definition

    It’s not a search. If it was, it would always converge to the same solution no matter what starting point. It’s obviously not what happens.

  22. dazz,

    Natural selection doesn’t evaluate. And again, all GAs are search heuristics. That is just what they are. And no, being a search does not mean they will all converge on the same solution. You made that up

  24. What I should have said in my previous post was that this is a good simulation of an extremely limited case of micro-evolution and nothing more. Does anyone disagree with this assessment?

  25. To clear up some confusion:
    ID is not anti-evolution but it is anti-evolutionism. Evolutionists promote evolutionism

    Also we see the power of directed evolution, ID, whereas undirected evolution, Darwin, is impotent. Thus the distinction is easily seen.

    Carry on

  27. Frankie can’t see my comments…

    Richardthughes: But ID isn’t anti evolution.

    Frankie: To clear up some confusion:
    ID is not anti-evolution but it is anti-evolutionism. Evolutionists promote evolutionism

    It’s a miracle!

  29. CharlieM:
    What I should have said in my previous post was that this is a good simulation of an extremely limited case of micro-evolution and nothing more. Does anyone disagree with this assessment?

    It has no applicability to either the micro or macroevolutionary concepts, because there is no delineation of species in the simulation.

    What it is, is a demonstration of the power of random change subject to environmental selection.

  30. CharlieM:
    What I should have said in my previous post was that this is a good simulation of an extremely limited case of micro-evolution and nothing more. Does anyone disagree with this assessment?

    But a very powerful method for exploring large phase spaces, such as the space of all possible blade shapes, or the travelling salesman problem (which is a real world problem).

    What you cannot infer from a simple model is that biology has similar restrictions or limitations.

  31. Rumraket: It has no applicability to either the micro or macroevolutionary concepts, because there is no delineation of species in the simulation.

    What it is, is a demonstration of the power of random change subject to environmental selection.

    Random change to a very simple object in an environment which is a very poor representation of reality. What has been evolved in the way of novelty?

  32. CharlieM:
    What I should have said in my previous post was that this is a good simulation of an extremely limited case of micro-evolution and nothing more. Does anyone disagree with this assessment?

    Yes. I don’t think it’s a “simulation”. It’s an actual EXAMPLE of Darwinian processes optimising the configuration of something so that it performs a function.

    It’s a simulation in the sense that that the wind-tunnel is a simulation of real wind. But the process itself isn’t a simulation. It’s an exemplar.

  33. Joe Felsenstein,

    This is just such a pathetically naive answer its silly. All the program is doing is taking different variations of shapes and running them through a wind tunnel and deciding which is the most efficient. The program tells it EVERYTHING, except which shapes to try.

    Such a card trick should perhaps fool a five year old into thinking it was doing magic, but it shouldn’t fool a college professor for crying out loud.

    People here are claiming that it could still come up with a good design if you gave it more than one definition of what was good, and didn’t give any priority to any of the goods. That is poppycock. Light is good, strong is good, heavy is good, big is good, small is good, cheap is good, turning fast is good turning slow is good, looking like a bird feeder is good. Give the program these options of what is good, and you will end up with junk. You won’t end up with partially good, you will end up with worse than what you started with.

    This is an example of the power of an algorithm? Oh fucking brother. This is an example of running a million different shapes through a wind tunnel and seeing which has the highest rating. So what! Its a great example of a computers ability to test large quantities of shapes. You must be frickin amazed at three card molly players. Holy shit.

  34. CharlieM: in an environment which is a very poor representation of reality

    The environment here is simulated by the software that estimates the performance of the blades. That “virtual wind turbine” software he mentions at 2:38. Not sure if he coded that too or if that’s a 3rd party software, but how do you know that software is a “very poor representation of reality”?

  35. CharlieM: Random change to a very simple object in an environment which is a very poor representation of reality. What has been evolved in the way of novelty?

    It’s a toy example, but in reality, such programs solve problems that are not amenable to other methods.

  36. phoodoo:
    Joe Felsenstein,

    This is just such a pathetically naive answer its silly.All the program is doing is taking different variations of shapes and running them through a wind tunnel and deciding which is the most efficient.The program tells it EVERYTHING, except which shapes to try.

    Such a card trick should perhaps fool a five year old into thinking it was doing magic, but it shouldn’t fool a college professor for crying out loud.

    People here are claiming that it could still come up with a good design if you gave it more than one definition of what was good, and didn’t give any priority to any of the goods.That is poppycock.Light is good, strong is good, heavy is good, big is good, small is good, cheap is good, turning fast is good turning slow is good, looking like a bird feeder is good.Give the program these options of what is good, and you will end up with junk.You won’t end up with partially good, you will end up with worse than what you started with.

    This is an example of the power of an algorithm? Oh fucking brother. This is an example of running a million different shapes through a wind tunnel and seeing which has the highest rating.So what!Its a great example of a computers ability to test large quantities of shapes.You must be frickin amazed at three card molly players.Holy shit.

    I take it you don’t code or work with computers or math. Let’s say there 90 spines with lengths from 0-100. How many possible blades are there? If we exhaustively list all the permutations, how big a data set would that be. If we were go evaluate the list by reading 1 a second, how long would it take to get through it?

  38. CharlieM: Random change to a very simple object in an environment which is a very poor representation of reality.

    I think you need to come up with a few more rationalizations for why the result of this simulation is uninteresting to you.

    But thank you for conceding the simulation isn’t actually relevant to micro or macroevolutionary concepts.

    CharlieM: What has been evolved in the way of novelty?

    It seems the simulation has produced a veritable flood of novel approaches to try and make a religiously motivated downplaying rethoric, appear rational.

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

    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.

    Congratulations, you finally get it!

  40. petrushka: But a very powerful method for exploring large phase spaces, such as the space of all possible blade shapes, or the travelling salesman problem (which is a real world problem).

    A very powerful method for exploring searching large phase spaces.

    Fixed that for you.

    …or the travelling salesman problem (which is a real world problem)

    https://en.wikipedia.org/wiki/Travelling_salesman_problem

    Search for the word “search.”

  41. The phrase search is what’s causing the problem? It looks for a solution, not a specific solution.

  42. Richardthughes: Let’s say there 90 spines with lengths from 0-100. How many possible blades are there? If we exhaustively list all the permutations, how big a data set would that be. If we were go evaluate the list by reading 1 a second, how long would it take to get through it?

    It’s obviously a guided search.

  44. Richardthughes: The phrase search is what’s causing the problem? It looks for a solution, not a specific solution.

    And if you were in search of gold would you toss away every lump that didn’t meet a specific shape?

  45. Richardthughes: The phrase search is what’s causing the problem? It looks for a solution, not a specific solutio

    To the extent it is mimicking evolution, all tries are solutions. What it is doing is accumulating — iteratively — better solutions.

    In living populations, everything that reproduces is a solution. But changes accumulate, and occasionally a change is an advantage.

  47. Elizabeth,

    It’s an actual EXAMPLE of Darwinian processes optimising

    Except Darwinian processes do not optimize and they are not a search actively searching for a solution. IOW it isn’t an example of Darwinian processes at all.

  48. Mung: And if you were in search of gold would you toss away every lump that didn’t meet a specific shape?

    No, i’d take all the gold if I were unconstrained. Did you have a point?

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

    It should check Barry’s purse, OldMung. There’s something in there for sure!

Leave a Reply