This should get some good discussion / denial!
652 thoughts on “Evolving Wind Turbine Blades”
Leave a Reply
You must be logged in to post a comment.
The Skeptical Zone
"I beseech you, in the bowels of Christ, think it possible that you may be mistaken."
Goodness you are one bad quoteminer. The comment *the full one* was about halting.
Let’s assume that a local maximum is arbitrarily defined as a solution, and the mutational model of the GA plus the ruggedness of fitness landscape doesn’t allow the GA to get unstuck of a local maximum. How do you know you’ve reached the local maximum?
If it’s search, who is setting what is searched for?
I spy the rabbit hole, but nonetheless, biological evolution is to search as rock is to door-stop. That something can be used as an X does not make it an X in all senses.
dazz,
When the output doesn’t change
Searches require sight and mind. Yet evolution is supposed to be blind and mindless.
Whoops
dazz,
The results didn’t change. That means no further changes were being done. Yours is a distinction without a difference.
AND it isn’t really done until it is put through a real world test. This is admittedly far from that
methinks it is like a weasel
methinks it is like a monkey
Oh look, two solutions, it can’t possibly be a search algorithm!
Do the math.
It’s hard for me to believe that no one has yet solved the problem of how to know you’re at a local maximum.
Mung,
How die has N sides, each with a different number. How many rolls must you make to ensure you’ve seen the highest number?
There is no guarantee you’ll ever see the highest number.
Very good.
How do you know you’ve reached the local maximum?
You crack me up. That algo was a search, because the results were known in advance. A GA has no clue what to look for, it just keeps taking crap shots and the fittests endure.
To turn a GA into a search, you need some external criteria to the algo to determine when to stop. But the algo is not searching for anything in particular
In the wind mill example, when each run converges into a different solution, it could be stuck in the edge of a plateau of the fitness function. If it kept running for long enough, it could potentially find a mutation that gets it out of there and into a new upwards hill.
One “limitation” of the implementation of this wind mill GA is that it doesn’t allow for drift. Imagine that there’s a hill in the middle of that plateau at which edge the algo gets stuck. Any mutation that hits the surrounding plateau is equally efficient as the current best, but since it doesn’t beat it, it’s discarded. If the algo allowed for equally (or even slightly deleterious) mutations to endure, it could drift through the plateau and eventually find the hill in the middle. The problem with this is that it would make the algo heavier computationally speaking: it would be more difficult to find local maximums (poorer short term performance) but would allow it to explore more of the fitness landscape in the long run. If you also add branching as happens in real life, that would be like running the algo multiple times, but instead of starting from some random, potentially low fitness, you start at some increasing level of fitness in every branch.
The result is a tendency of multiple branches for finding better and better solutions, spreading through more and more of the landscape, all by simple random variation plus selection. Just like evolution
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.11.7472&rep=rep1&type=pdf
“water searches for lower regions”
Indeed.
dazz,
Good post. Also, imagine if the wind patterns and strength were randomized…
Because with a real 3 dimensional, 3 blade wind turbine there will be many more variables which will affect the outcome.
Hill climbing is not a GA according to that. It changes one parameter at a time incrementally, not randomly, exploring the adjacent vicinity of the landscape. Random variation allows for larger “jumps”
If you’re gonna tell me that GA’s are a search for local maximums, because you can find them using a different algorithm then I’m done here
Exactly, randomized or even affected by the unpredictably ever evolving population. The result must be that “replaying the tape of life” would never produce the same result twice, even if starting in the same exact conditions.
dazz,
This is all new to Mung. But evolution is wrong.
See, you do have a sense of humor!
How do you know what variables does that thing model? It works with an input of the 3D shape of the blades, and then the heavy lifting is probably done in the wind dynamics. You’re definitely pulling this out of your rear end
What’s so funny about it?
You’ve been asserting over and over again that GA’s search for stuff but you keep us waiting for an answer as to what is it that it searches for.
Why be so cruel?
Now if a GA was meant to properly model evolution, it should also model mutational meltdown, which means that some branches of small populations would actually work their way downwards into extinction. Of course that would make no sense for engineering applications, but my point is that the closest we get to modeling evolution, the clearer it gets that the algo doesn’t search for anything, it’s just a process that happens to beat the crap out of design in many applications even if it doesn’t have a tiny fraction of the detail, time and resources that evolution has had throughout the history of life in the last 3 billion years or so.
Yeah, it’s a search, if you want it to be a search. Now what?
IDists have made it abundantly clear that evolution is a search, so it must be. And since it’s failed to find the crocoduck then it must be wrong.
Damn, they were right all along. Time to find the nearest church and pray forgiveness
You’re only forgiven if you promise never to think (non-trivially, anyhow) ever again.
Glen Davidson
Think ’bout wut? *slobbers*
It doesn’t actually matter whether it might be colloquially called a “search”. What matters is whether it meets the requirements of search as defined for “No Free Lunch” and similar mathematical results.
Is that where Dumbski claims CSI is “smuggled” into the fitness function?
EDIT: Oh, I see what you mean now
https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_optimization
Allan Miller,
This has a hard target, that is my whole point. It has ONE target, which has the most efficient rating in a wind tunnel. That’s the only target. Its no different than a computer selecting the longest piece of string. Its just a computer selecting it instead of a guy sitting in a lab all day testing shapes.
The entire problem is when you DON’T have a target, you know like evolution for example! The claim that it could also work with multiple variables is pure and utter bullshit. It couldn’t work with TWO variables, if the two variables were contradictory. You can’t accumulate successive better models if each time the trait selected for is different. if you want to make a good turbine, but one time good means wind efficient, and another time it means cheap, and another time it means looking like a bird feeder, each time you accumulate in one direction, you move away from the other direction. Each time you select for it looking like a bird feeder, you move AWAY from it being more efficient. To get back to efficient, you have to remove the traits which you selected for that made it look like a bird feeder. You go one direction, and then stop and go back in another. Nowheresville. You will get neither one that looks like a bird feeder, NOR an efficient one. And the more variables you allow, the worse it will get.
Will the weasel program work if we give it TWO targets? How about seven? How about 100?
I bet a 10 year old could see the problem, if he was given a five minute explanation. But you geniuses….Richard could pretend to read an abstract about it for a year and still babble, but, but its evolution, see.
no, I’m quite honest about what I’ve read. I didn’t read that whole paper you did a post on. Did you?
https://scholar.google.com/scholar?q=evolutionary+trade+off+examples&hl=en&as_sdt=0&as_vis=1&oi=scholart&sa=X&ved=0ahUKEwiCnLvHx5HKAhXLRiYKHRspAgcQgQMIGjAA
There’s no point looking if you already “know”, eh Phoodoo?
Evolving Soft Robots with Multiple Materials
Four different variables, the only “target” was faster walkers generate more offspring.
Look at all the variety that evolved.
Adapa,
This is some of Hod Lipson’s work!
http://www.forbes.com/pictures/lmm45emkh/7-hod-lipson-and-michael-schmidt-computer-scientists-cornell-university/
Richardthughes,
How long do you think it will take before chihuahuas become a new species Richard? I mean, we are speeding up the process to make it easier for them right. When?
phoodoo,
I don’t know, nor care, Phoodoo. I just try and learn about things before I talk about them. You might want to try it, to save future embarrassment.
Adapa,
Great Adapa, four methods to get at ONE target. Why only one target? How about four methods to get at six targets! The designs will get even worse won’t they? (Clue-the correct answer is yes).
How bad will the designs get when we give it 27 methods to reach 100 targets? Guess what you will come up with? Something akin to the unorganized, spastic disarray of Richard’s brain, that’s what.
phoodoo,
Well we can’t all be as bright or honest as you, Phoodoo? I don’t know why you’re here as you know everything already. I’m surprised you’ve not given us your thoughts as to Why Lizzie was off for a week?
Because life has only one target – survive to reproduce.
I did get a big laugh out of your silly claim GAs can have only one variable though.
Adapa,
Ah, but you see, we already have given the program the ability to reproduce right?
But now you want it to reproduce better right? Well, what does better mean? reproduce by being faster? Reproduce by being smarter? Reproduce by being bigger? Reproduce by being smaller? We need those variables right?
Now how do we get faster? Well, we could slither better, flap wings better, jump further, have bigger hind legs, lighter bones…We need variables for all the different ways in which we can be faster. And bigger, and stronger, and, and….
So without ONE definition of good, and just throwing them into a pot of good, you don’t build progressively better in any ONE way. You don’t get progressively faster if you are also building progressively fatter and progressively shorter and progressively warmer at the same time. The traits which make you fatter, hinder your ability to be faster. The traits that make you faster hinder your ability to make you fatter. Which does the computer choose for the next round? Ok, choose the fatter. Next round choose the fatter one that also has good fur. Next round choose the fat one with good fur, and throw him out of a tree and ask him to fly. Whoops!
I wonder why MIT does’t use this method to design spaceships. “Computer select me a cheap, fast big rocket to go to Mars, that uses propane, and also uses coal because its abundant, and that is very comfortable, and that can last a really long time, and that can be taken apart quickly, and can’t be burned up upon reentry. WTF, you selected a Winnebago with a fireplace? That can’t fly! Well, we had to make a few compromises. “
Animals in populations that evolve already have the ability to reproduce.
It simply means having a better chance to survive to reproduce than the neighbors you're competing against.
Only if you want your simulation to have more granularity.
The simulation I just showed you managed all those with just four variables in different combinations.
The only “good’ is to survive better than your neighbors. How that is achieved is unimportant.
You don’t understand how these GAs work even a little.
They could but I suspect the computing power required would be prohibitive.
Exactly! You finally get it. There can be many ways to survive better. So in order to show that you can build something smart by using any method you want to reproduce better, instead of just one (like walking better) you can’t only select for the ONE way! Congratulations Adapa, progress!
So if the program said, any way in which the windturbine got better we can select that. So looking like a birdfeeder would get selected just as often as having good aerodynamics. And being cheap to produce would also have as much priority for selection as good aerodynamics. Its like making a cake using a recipe for guacamole soup, and a recipe for fish balls, and one for daiquiris, and another for peanut butter, and another for haggis. The cake isn’t going to turn out very good, even though each recipe is great.
You are learning Adapa.
Actually I’ve never said that. You’re making things up again, and it shows the world your character.
Well, this is very exciting. We have two competing hypotheses:
1. Software engineers have mimicked a biological phenomena that isn’t real / doesn’t exist / can’t do diddly and somehow come up with a methodology that beats expert’s designs
or
2. Phoodoo, who’s posts thus fair have been a fact-free mix of conjecture, inaccuracy, misunderstanding and incorrect assertion, is right.
Place your bets now!
Sadly you never seem to get it.
Learning just how willfully ignorant a Creationist can be.
Its not a 3D shape, its 2D. And I cannot fathom what you are trying to say with the rest of that sentence.
As for the simulations representation of reality, we have this from the person responsible for the video:
That was 5 years ago. He may have improved his methods in the mean time. I would be interested to find out how his research is coming along because I do think it is a worthwhile project.
CharlieM,
He says in the main comment
“Note: I have since discovered I used the wrong viscosity for air on this experiment, so the results aren’t valid for use on the earth. (Maybe Jupiter.)”
So you keep mindlessly asserting with unearned confidence. Nobody gives a shit.
Prove it.
Oh for fucks sake, the measure of “better” for real organisms is reproductive success. It’s not “tall and short” or “furry and scaled” or “walking and burrowing”. It’s reproductive success full stop. Whatever compromise solution is better able to reproduce is the measure of success, regardless of how that is actually physiologically achieved.
Your claim is that when multiple selective pressures are in operation, nothing can evolve at all. You’ve blindly and staunchly asserted this multiple times. I don’t believe you are correct at all. I want you to put your money where your mouth is and PROVE YOUR ASSERTION.
I’m fine with calling it a search as long as we understand evolution isn’t consciously or deliberately searching for anything. That we use the word ‘search’ as a metaphor for how the evolutionary process results in adaptations in living organisms. If you’re married to the word search and just can’t let it go, then fine I can live with you using it.