Using “Evolutionary Algorithms” by Intelligent Design

CEH ^ | May 8, 2009

[GodGunsGuts]

Using “Evolutionary Algorithms” by Intelligent Design

May 8, 2009 — Evolution can’t be all bad if scientists can use it to optimize your car.  Science Daily said that scientists in Germany are “simulating evolution” to come up with ways to optimize difficult problems.  Using “Evolutionary Algorithms”, they can discover solutions for engineering problems like water resource management and the design of brakes, airbags and air conditioning systems in automobiles.  The simulated evolution program searches through a large number of random possibilities to make numerous successive slight improvements.

“The algorithms are called ‘evolutionary’ because the characteristics of evolution – mutation, recombination and selection – form the basis of their search for promising solutions,” the article claimed.  Solutions that show promise are mutated and further selected.

Conferences on Evolutionary Algorithms are held each year and the interest in them is spreading into other disciplines.  “The Evolutionary Algorithms are therefore a collective term for the various branches of research which have gradually developed: evolution strategies, evolutionary programming, genetic algorithms and genetic programming.”

Every once in awhile we need to give a refresher course about these reports, to show why the terminology is ludicrous.  This has nothing to do with evolution and everything to do with intelligent design.  Calling these

“evolutionary algorithms” is like calling Eugenie Scott a creationist.  Evolutionary Algorithm is an oxymoron – if it is evolutionary, it is not an algorithm, and if it is an algorithm, it is not evolutionary.  Why?  Because the essence of evolution, as Charles Darwin conceived it, has nothing to do with intelligent selection.  Evolution is mindless, purposeless, and without a goal.  These scientists, by contrast, have clear goals in mind.  They are consciously and purposefully selecting the products of randomness to get better designs – intelligent designs.  They may not know what the computer program will produce, but they sure well programmed the computer, and put in the criteria for success.  Employing randomness in a program does nothing to make it evolutionary.  The hallmark of intelligence is having a desired end and pulling it out of the soup of randomness.  This is something evolution cannot do – unless one is a pantheist or animist, attributing the properties of a Universal Soul to nature.  Undoubtedly, the NCSE would decry that.  They can barely tolerate theistic evolutionists – the well-meaning but misguided Christians who try to put God in the role of the engineer who uses evolutionary algorithms for his purposes (e.g., man).

Remember – if it has purpose in it, it is not evolution.  We must avoid equivocation.  To discuss evolution with clarity it is essential to understand the terms and not mix metaphors.  Charlie lept from artificial selection (intelligent design) to natural selection (materialism) only as a pedagogical aid.  He did not intend for natural selection to have a mind like the goal-directed farmer or breeder uses.  To think evolution, think mindless.  Notice that itself is a one-way algorithm.  You can think mindless, but the mindless cannot think.

For a definitive, in-depth treatment on why evolutionary algorithms cannot be mixed with evolution, see the book No Free Lunch in the Resource of the Week entry above.

[AndrewC]

you don’t know what the answer will be, just that there will be an answer.

I know what the answer will be, whatever the stategy you use. You lose. But you can have fun and free drinks while doing it.

[Ha Ha Thats Very Logical]

I know what the answer will be, whatever the stategy you use. You lose. But you can have fun and free drinks while doing it.

You don't lose every time. But yes, eventually your money will go extinct.

[PugetSoundSoldier]

By the logic you’re using there were not any daughters born to Eve nor most every other matriarch in the Bible - at least not until and unless explicitly mentioned. The most ridiculous assertations come from people who don’t truly study God’s Word and then mangle it further trying to interweave it w/ evolution.

No, what YOU are doing is adding information to the Bible; information NOT contained within it! If we are to only take a literal translation of the Bible as word-for-word accurate, then Adam and Eve had ZERO other children until after Cain married and had Enoch. Period.

PLEASE point out in the Scriptures where this is NOT the case. You are surmising and making assumptions about events which happened outside the Biblical texts, and that is the VERY same thing that GGG says is wrong in relation to the Creation story.

Bottom line: if you want to claim the Creation story as 100% accurate and essentially the scientific truth, then you have to explain where Enoch came from. Because according the Bible there was only one woman alive at the time, and that was Eve! She did not marry Cain, because he left Adam and Eve to go find his wife.

Essentially, if the Bible is historically and scientifically infallible then you have a BIG problem here.

Now, if the Bible is theologically and philosophically infallible, there's no issue. God created all - THAT is the message of the creation story! He started it, He is responsible! The actual method is irrelevant; what matters is that He did it.

Note that this allows for the fact that Adam and Eve may have been the first spiritually aware people; the first ones to understand and realize there is a God, and to have a relationship with Him. It does not discount the existence of others, like the wife of Cain...

So you deny the theory of evolution on the basis of "it's not scriptural". Great. So tell me where did Cain's wife come from? Because your claims about here are clearly just as unscriptural as you claim the theory of evolution to be.

[PugetSoundSoldier]

You don’t really see man as a ‘creator kind’ when he develops any simple program or something as complex as a genetic algorithm?

Yes, I do. And I don't see the theory of evolution saying how life was created, only how it changed over time (much like how a GA's population changes over time).

Computer logic and programming was created to simulate life - in it’s infancy in order to perform repetitious mind-numbing tasks involving math, sorting, filtering, and decision-making - but as our knowledge base grows who knows how close it may come to resemble God’s creation.

Yet all actions we make can be broken down into decision trees with weightings based upon our desires and intentions.

GAs don't work by simulating just an individual; they work by simulating populations. The averages over time are what get you the result.

Much like societies in general. We can say that societies that encourage freedom and private property rights become wealthier over time. "Bonus Points" the members of those societies. Societies that embrace homosexuality are typically in their later stages and close to decline, so lose a few points there.

GAs deal with populations of solutions, ebbing and flowing as conditions dictate until you end up with a fairly stable population that is relatively consistent in its environment and results.

The power of GAs does not reside in the creation of the population; the power lies in the way the population evolves. They are two separate events, much like creation and the theory of evolution.

[AndrewC]

In my previous replies I did not show the "genetic algorithm" which really troubles me since so many people use it as an example. Scientific American even mentioned it. In past threads long ago forgotten, I posted it. I found it again. Here it is in "full glory" from How "Chance" works in Evolution

THE COMPUTER PROGRAM IN APPENDIX E IN "UPON THE SHOULDERS OF GIANTS" BY 
RICHARD HARDISON

10 REM 1984 R. HARDISON
11 PRINT "RANDOMIZING ALPHABET"
12 PRINT "WRITE HAMLET, KEEPING"
13 PRINT "SUCCESSES."
14 PRINT :; REM N-COUNTER: # OF TRIALS
15 REM T=COUNTER:REUSE "TO BE"
16 PRINT "SUBROUTINE TO
17 PRINT "RANDOMIZE AND SELECT"
18 PRINT "LETTER"
30 N = 0
40 FOR G = 1 TO 10
50 T = 0
60 GOTO 80
70 X = INT (26 * RND (1)) + 1: RETURN
80 GOSUB 70
90 N = N + 1
100 IF X = 20 THEN PRINT "T": IF X = 20 THEN GOTO 120
110 GOTO 60
120 N = N + 1
130 GOSUB 70
140 IF X = 15 THEN PRINT "O": IF X = 15 THEN PRINT : IF X = 15 THEN GOTO 160
150 GOTO 120 
160 N = N + 1 
170 GOSUB 70 
180 IF X = 2 THEN PRINT "B": IF X = 2 THEN GOTO 200 
190 GOTO 160 
200 N = N + 1 
210 GOSUB 70 
220 IF X = 5 THEN PRINT "E": IF X = 5 THEN PRINT : IF X = 5 THEN GOTO 240 
230 GOTO 200 
240 T = T + 1
250 IF T = 2 THEN GOTO 460 
260 N = N + 1 
270 GOSUB 70 
280 IF X = 15 THEN PRINT "O": IF X = 15 THEN GOTO 300 
290 GOTO 260 
300 N = N + 1 
310 GOSUB 70 
320 IF X = 18 THEN PRINT "R": IF X = 18 THEN GOTO 340 
330 GOTO 300 
340 N = N + 1 
350 GOSUB 70 
360 IF X = 14 THEN PRINT "N": IF X = 14 THEN GOTO 380 
370 GOTO 340
380 N = N + 1 
390 GOSUB 70 
400 IF X = 15 THEN PRINT "O": IF X = 15 THEN GOTO 420 
410 GOTO 380 
420 N = N + 1 
430 GOSUB 70 
440 IF X = 20 THEN PRINT "T": IF X = 20 THEN PRINT : IF X = 20 THEN GOTO 60 
450 GOTO 420 
460 PRINT "N=";N;" KEYS PRESSED TO WRITE 'TO BE OR NOT TO BE'" 
470 PRINT "FOR";G;" RUN(S) OF PROGRAM"
480 PRINT 
490 NEXT G 
500 END
510 REM  IF THE PROGRAM WERE 
511 REM  WRITTEN TO INCLUDE 
512 REM  PUNCTUATION MARKS ETC. 
513 REM  THE PROGRAM WOULD 
514 REM  TAKE LONGER, BUT WOULD 
515 REM  STILL NOT BE PROHIBI- 
516 REM  TIVE 
517 PRINT
518 PRINT "WITH 3000 RUNS, THE MEAN" 
519 PRINT "# of trials=333" 
520 PRINT "THE MEAN TIME REQUIRED" 
521 PRINT "WAS .14 MINUTES TO PRINT" 
522 PRINT "TOBEORNOTTOBE" 
------------------------------- 
From this analysis of Darwin,
Hamlet, Dawkins, Hardison, coincidence, and selective evolution, we may
conclude that whether the reality of evolution is to be believed or not to be
believed, methinks it is like a weasel of truth nonetheless.

Michael Shermer

I modified it to fix the run on lines, but it still has syntax problems which I don't need to fix to show how silly the program is. Essentially the program waits around until it gets the correct letter then moves on to the next until the whole "phrase" "TOBEORNOTTOBE" is generated. The spaces are put in by the program. Whatta marroooonnn! -- Bugs Bunny. Then, as if it means anything, the average number of iterations over 3000 runs is given, viz. 333. Well, that algorithm waits on average 18 iterations to get a .5 probability of not getting the letter. And 18*13 is 234. Plus, if I threw away the "invalid" letters not matching each position and only considered the remaining letters(no replacement), at most only 26*13 iterations would be required. That is 338.

[PugetSoundSoldier]

LOL- Yep- nonsensicle- just make hte statement, and that’s the end of it- BOTH those articles broke down EXACTLT what was goign on- obviously you didn’t bother to even give htem a read- Your only point was about ‘stop executing’- but BOTH those articles addressed everyhtign you brought up-

OK, here you go...

1. The author claims that GAs are "inefficient awkward process", when in fact they are the MOST efficient means of solving and optimizing open-solution-set multi-variable solution space problems. Fundamentally the author discounts the entire reason that GAs are used for the problems they address, because he doesn't accept the well proven benefit they provide.

2. The author states "All too many evolutionary computationists fail to realize the purely formal nature of GA procedures. GAs are not dealing with physicodynamic cause-and-effect chains." and that shows his ignorance of how GAs work! GAs, in fact, model the same "physicodynamic" (usually called physiodyanmic) realm as genes in that genes consist of chromosomes you get from either your parents or a mutation. Where else do your chromosomes come from? Biology tells us there's only two sources, and that's what GAs use. So here the author tells us he doesn't understand how GAs even work.

3. The author states that the "overall process was entirely goaldirected (formal). Real evolution has no goal" both of which are false. The theory of evolution claims that evolutionary forces have a goal of higher survivability of the entity (rates higher in its fitness function, in the GA world). That the goal of evolution is to minimize your mortality (maximize your fitness score). Clearly the author doesn't understand how GAs are rated and scored after each generation, and how natural selection - the fittest tend to (but not always) survive better than the weakest - does in fact mimic the theory of evolution.

Would you like me to go on? Yes, it is nonsensical because the author shows - in a few paragraphs - he does not understand why GAs are used, where they are best used, how they operate, and how a computer scientist determines when to stop a run.

If someone claimed to be a Biblical expert who did not understand Judeo-Christian philosophy, did not read the Bible, and is ignorant of the Resurrection, would you call that person a Biblical expert, and call their ramblings on the Bible non-sensical?

Same thing here.

Teee heee heee- look everybody- Puget thinks he’s a smartie because he can ‘insult’ right alongside hte kiddies in the higher classes

Yeah, you're right, that was a cheap shot. I apologize, brother.

[Ira_Louvin]

Your reply shows a common misconception about science.

Science cannot “prove” anything. Using the scientific method everything is an “inference”

“Hypotheses and theories can never be proven true using the scientific method. Therefore, science advances only through disproof. This is a critical and often misunderstood point. To be scientific, theories can never be proven true, but all theories must be refutable. Therefore, all theories, and by extension all of science, are tentative.

As an example, let’s use a science fact that is known to most adults: the existence of electrons. We know that electrons exist, but here’s the rub: Science can never prove that electrons exist. Hypotheses about the existence of electrons have been supported after countless tests using the scientific method. In other words, they have not been refuted. Knowledge of the precise nature of electrons will always be undergoing refinement, but the weight of scientific evidence clearly supports the existence of electrons.

How about another example? This time we’ll use an example from plant biology and agriculture. A scientist states a hypothesis that adding nitrogen to the soil will result in increased grain production in corn (maize) plants. The scientist tests the hypothesis in a carefully controlled experiment. Her hypothesis is that nitrogen will increase grain production, and because the hypothesis must be subject to refutation, her alternative hypothesis is that nitrogen will not increase grain production. The experiment reveals that nitrogen does indeed increase grain production. Therefore, her initial hypothesis (also known as a null hypothesis) is supported. If the experiment had not resulted in increased grain production, the initial hypothesis would have been refuted and the alternative hypothesis would have been supported. The scientist can never prove that adding nitrogen to soil increases grain production, but if the hypothesis is supported time and time again, the weight-of-evidence convinces us that the relationship between nitrogen and increased grain production exists and is predictable.”

http://agbiosafety.unl.edu/science.shtml

Also you seem to have overlooked my question to you………

Please explain in detail step by step how evolution defies scientific examination

[PugetSoundSoldier]

Cain's wife is one those “other children, one of his sisters likely. Some of those sons and daughters of Gen. 5:4 born earlier.

Please point to the verse that states this; in fact, Genesis 5:4 states (KJV):

And the days of Adam after he had begotten Seth were eight hundred years: and he begat sons and daughters

Clearly the Bible lays out the chronology of the other sons and daughters as AFTER Adam had Seth, which happens AFTER Cain is married.

The Bible is unequivocal on this matter; there were NO other people on the earth at the time Cain slew Abel other than Adam, Eve, Cain, and Abel.

Exactly which one of Cain's female relatives became his wife isn't stated or how old he was when he took a wife and fled east. but since he did take a wife there were females, his relatives, alive at the time.

No, there weren't if you go by a strict reading of the Bible! See, you're allowing your own inferences to override the literal texts and words of the Bible. Biblically, the other relatives WERE NOT BORN until after Cain fathered Enoch. Period. Point to ANY other births prior to that, or to any verse stating that Adam had other children before (not after) Seth was born. You can't.

So you make a conclusion that it must have been true and this is an irrelevant detail to concern yourself with since it's the only thing that could have worked. Never mind the Bible doesn't mention it.

Yet you denigrate those who do the same for the Creation story. That maybe the importance of the Creation story is not in the methodology of the Creation as written but in the meaning of the Creation? That the Bible - just like with Cain's wife - does not record every event historically, but contains them for the fundamental philosophical and theological truths that God loves us!

[PugetSoundSoldier]

I've programmed computers since 1965.

Good; I've only done that since 1980. But I have also programmed several GAs and SAs, especially for fitting multi-variable solutions. For example, solving passive electrical filter network designs of high complexity (say, 30-40 components) where not only the fit of the impulse response is important but the cost, components count, weight, and volume of the filter must be considered as well.

Your experience in programming GAs is?

I do not think that it is a scientific principle that nature tries to evolve a population to solve a certain problem. I think that is a hallmark of intelligence.

Yes, it does. It's called survival of the fittest. That is what GAs look for - the fittest population to work with.

And as with life, for GAs there is no "best answer". There is a large - sometimes infinite - solution set, and you're looking for a good local fit. The GA continues modifying the members of the solution set - the population - based upon their fitness rating. Much like living in the Andes will weed out those who are physically incapable of such life.

The only difference between a GA and the theory of evolution is that the GA stops when the researcher decides they've reached a good enough answer; they'll happily keep chugging along creating generations of answers ranging from ludicrous to darn near perfect. And every generation you'll get that same range, but as time goes on you'll get more and more that fall towards the well-suited versus the ludicrous.

Much like, over time, the hearts and lungs of the native SA Andean indians has grown to allow for higher oxygen capacity when living in the mountains.

[PugetSoundSoldier]

Okay, when the program "displaying" the genetic algorithm terminates, describe the "solution". When does the genetic algorithm program terminate? Why does it terminate?

Define for me how much salt to add to 2 pounds of mashed potatoes. How much is the right amount? How do you know to stop adding salt? Is that amount of salt correct for GGG? For me? For everyone? Or only for you? Is that after or before addition of milk and butter, and what type of potatoes?

A GA is terminated (it doesn't end on its own) when the programmer/researcher decides he has an answer he can work with.

Imagine I give you a 30 element network to create a 20th order bandpass filter. You have unlimited choices of resistors, capacitors, and inductors to use. I want that filter to have a Chebychev response with a defined passband gain, passband ripple, and attenuation outside the passband.

Now, I also want you to make the filter as low cost as possible! However, since this is going on a research vessel, weight matters, as does size - we can't have a perfect fit filter that weighs more than 5 pounds, and occupies more than 200 cubic inches! And budgets being what they are, lower cost is always wanted.

In fact, if you can make it significantly smaller, we'd be willing to trade off some cost, or maybe some performance. If you can make it lighter, that's even better than small but not too much better. Light AND small together is really good - better than either one by themselves, or even together.

So how do you find the filter that best trades off functionality (the electrical passband/stopband performance) for cost, size, and weight? Each component you select has electrical, fiscal, and physical dimensions, and you have to solve for a solution that works.

There isn't just ONE solution here - given the complexity of the problem, there can be literally INFINITE valid solutions! But your problem is to figure out how to make one that YOU believe is pretty darn good.

THAT is how you use a GA. For solving problems where you have an idea of "how much salt is enough", but to exactly quantify it is essentially impossible and in fact irrelevant. You need a "good enough" answer that you - and your customer/boss - will be happy with.

[AndrewC]

The author states that the "overall process was entirely goaldirected (formal). Real evolution has no goal" both of which are false. The theory of evolution claims that evolutionary forces have a goal of higher survivability of the entity (rates higher in its fitness function, in the GA world). That the goal of evolution is to minimize your mortality (maximize your fitness score).

Sorry I can't let this pass. "Evolution" has no goal. It is inanimate. "Evolution" may have an outcome, a thunderstorm has that. If the goal of "evolution" was longevity, why the mayfly? Reproduction would be a better "goal", but it is not. And number of progeny would be an even better goal, but it is not.

That said, I posted a picture from the original paper under discussion in post 23. Do you notice anything peculiar about it?

[PugetSoundSoldier]

The point being is that an organism doesn't face one problem at a time.

Correct. And the problem address with a GA isn't singular either! Like that network I described - it has multiple problems (functionality, size, cost, weight) and all must be addressed SIMULTANEOUSLY to achieve a good answer. And we're willing to sacrifice in some areas to gain in others, but not too much.

GAs are great as solving "fuzzy, open" solution sets that exist in the real world. Where you can have dozens, thousands, or infinite numbers of solutions but where you just want one that works really well for your own needs.

Again, like mashed potatoes. How much salt is enough? You can't answer that unless you know what I like, how much I want to eat, what else I am having with the potatoes (gravy or sauces), how long the potatoes will linger on my plate (and thus intensify the taste of salt)...

What if the better the potatoes, the slower I eat them as to savor them? How much salt to add at that point? Or knowing that if the beef is dry, I will use more gravy? Or if it's a cold night the food will chill faster and that will decrease the flavor intensity? Am I going to choose a red wine, beer, or simply water with my meal, knowing that alcohol will decrease my sensitivity to salt? Am I going to eat a 6 PM or 10 PM, where physical exhaustion is an issue in terms of salt intake and perception?

Suddenly the amount of salt is dependent not just on the potatoes, but my actions (speed of consumption, gravy, other items consumed), completely external variables (temperature and time of day), and that some of those actions are driven by other items (the dryness of the beef or what I choose to drink with my meal)!

See, it becomes a rather complex problem VERY quickly. THIS is the type of problem you try to solve with a GA.

[PugetSoundSoldier]

Yes, and in a programmed genetic algorithm, that "solution" is used in each iteration to tailor the results of the iteration.

Except "that solution" is actually a set of usually thousands, if not millions of solutions some of which are really good fits and some of which are terrible. There ISN'T a single solution with a GA; there is a solution space (a set of solutions) that have varying degrees of fitness based upon how you weight.

Typically at the end of a GA run (at least my runs, with populations typically maintained at 50,000) you end up with a population where the top 1,000 are all EXTREMELY good and solid answers! The next 10,000 are also really good, but not as good.

Then I go and choose one of the top 1,000, and often that choice is based upon other external variables (like with the filter, I may limit my solution choice to what components I have laying around right now, so as to minimize the number that I have to order).

GAs don't just "give you an answer"; they give you a large set of potential answers, and the fitness of those answers - and the size of the set - both grow over time.

[tacticalogic]

Hmmmm BIG difference between the two issues Tactic- big difference- The mathematics don’t figure in biologically impossible and unrealistic factors like the GA’s do- they simply figure out the naturally probable without comming to some highly improbable conclusions based on assumptions like hte GA’s do as well

As far as I'm concerned, they're both digital models of an analog system, and neither one can even lay claim to knowing what all the variables are that I can see.

Step away from the glittering generalities.

[AndrewC]

Define for me how much salt to add to 2 pounds of mashed potatoes. How much is the right amount? How do you know to stop adding salt? Is that amount of salt correct for GGG?

That is called cooking. It is definitely an intelligent process.

A GA is terminated (it doesn't end on its own) when the programmer/researcher decides he has an answer he can work with.

And how does he decide that? Flip a coin?

Imagine I give you a 30 element network to create a 20th order bandpass filter. You have unlimited choices of resistors, capacitors, and inductors to use. I want that filter to have a Chebychev response with a defined passband gain, passband ripple, and attenuation outside the passband. ...

...

But your problem is to figure out how to make one that YOU believe is pretty darn good.

Type I or type II?

You start here.

P.S. A bandpass filter is a filter that "attenuates" outside the passband.

P.P.S. There's a heck of a lot more information needed than what you provided. Power supplies, loads, temperature conditions, stability over time and or temp... etc.

[WVKayaker]

Kilroy? Is that Kilroy?

[AndrewC]

There ISN'T a single solution with a GA; there is a solution space (a set of solutions) that have varying degrees of fitness based upon how you weight.

And nowhere in the process of generating those "solutions" is there any hint of the weighing? I really doubt it.

[AndrewC]

I’m keeping mum.

[PugetSoundSoldier]

There is weighting of the results! I've stated that. And the weighting is typically non-linear because the results have different values.

GAs are used where there aren't hard-defined answers like you want. Where you can't exactly quantize every single potential input and output. Where the answer is more like the judge's conclusion on pornography: "I'll know it when I see it".

I'd suggest this as an excellent primer on GAs, and also the Wikipedia page on GAs as good background.

Like so much with life - where the answers/solutions are varied and shaded and fuzzy - GAs excel at operating in those situations.

[AndrewC]

Your experience in programming GAs is?

About the same as Richard Hardison who was cited in Scientific American.

Yes, it does. It's called survival of the fittest. That is what GAs look for - the fittest population to work with.

No it doesn't, according to those who speak to evolution.

Darwin first used Spencer's phrase "survival of the fittest" as a synonym for "natural selection" in the fifth edition of On the Origin of Species, published in 1869.[1][2] It is a metaphor, not a scientific description,[3] and is both incomplete and misleading. Survival is only one component of selection, and for example where a number of males survive to reproductive age, but only a few ever mate, the difference in reproductive success stems mainly from ability to attract mates rather than ability to survive. In an evolutionary sense, fitness is the average reproductive output of a class of genetic variants in a gene pool, and should not be confused with physically fit meaning biggest, fastest or strongest, which does not necessarily lead to reproductive success.[4] It is not generally used by modern biologists, who use the phrase "natural selection" almost exclusively.