Author
|
Topic: The Illusion of Design (Read 1245 times) |
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
  
|
 |
The Illusion of Design
« on: 2005-11-24 15:35:21 » |
 |
source: eSkeptic
The Illusion of Design
by Richard Dawkins
The world is divided into things that look as though somebody designed them (wings and wagon-wheels, hearts and televisions), and things that just happened through the unintended workings of physics (mountains and rivers, sand dunes, and solar systems). Mount Rushmore belonged firmly in the second category until the sculptor Gutzon Borglum carved it into the first. Charles Darwin moved in the other direction. He discovered a way in which the unaided laws of physics — the laws according to which things “just happen” — could, in the fullness of geologic time, come to mimic deliberate design. The illusion of design is so successful that to this day most Americans (including, significantly, many influential and rich Americans) stubbornly refuse to believe it is an illusion. To such people, if a heart (or an eye or a bacterial flagellum) looks designed, that’s proof enough that it is designed.
No wonder Thomas Henry Huxley, “Darwin’s bulldog,” was moved to chide himself on reading the Origin of Species: “How extremely stupid not to have thought of that.” And Huxley was the least stupid of men. The breathtaking power and reach of Darwin’s idea — extensively documented in the field, as Jonathan Weiner reports in “Evolution in Action” — is matched by its audacious simplicity. You can write it out in a phrase: nonrandom survival of randomly varying hereditary instructions for building embryos. Yet, given the opportunities afforded by deep time, this simple little algorithm generates prodigies of complexity, elegance, and diversity of apparent design. True design, the kind we see in a knapped flint, a jet plane, or a personal computer, turns out to be a manifestation of an entity — the human brain — that itself was never designed, but is an evolved product of Darwin’s mill.
Paradoxically, the extreme simplicity of what the philosopher Daniel C. Dennett called Darwin’s dangerous idea may be its greatest barrier to acceptance. People have a hard time believing that so simple a mechanism could deliver such powerful results.
The arguments of creationists, including those creationists who cloak their pretensions under the politically devious phrase “intelligent-design theory,” repeatedly return to the same big fallacy. Such-and-such looks designed. Therefore it was designed. To pursue my paradox, there is a sense in which the skepticism that often greets Darwin’s idea is a measure of its greatness.
Paraphrasing the twentieth-century population geneticist Ronald A. Fisher, natural selection is a mechanism for generating improbability on an enormous scale. Improbable is pretty much a synonym for unbelievable. Any theory that explains the highly improbable is asking to be disbelieved by those who don’t understand it.
Yet the highly improbable does exist in the real world, and it must be explained. Adaptive improbability — complexity — is precisely the problem that any theory of life must solve and that natural selection, uniquely as far as science knows, does solve. In truth, it is intelligent design that is the biggest victim of the argument from improbability. Any entity capable of deliberately designing a living creature, to say nothing of a universe, would have to be hugely complex in its own right.
If, as the maverick astronomer Fred Hoyle mistakenly thought, the spontaneous origin of life is as improbable as a hurricane blowing through a junkyard and having the luck to assemble a Boeing 747, then a divine designer is the ultimate Boeing 747. The designer’s spontaneous origin ex nihilo would have to be even more improbable than the most complex of his alleged creations. Unless, of course, he relied on natural selection to do his work for him! And in that case, one might pardonably wonder (though this is not the place to pursue the question), does he need to exist at all?
The achievement of nonrandom natural selection is to tame chance. By smearing out the luck, breaking down the improbability into a large number of small steps — each one somewhat improbable but not ridiculously so — natural selection ratchets up the improbability.
As the generations unfold, ratcheting takes the cumulative improbability up to levels that — in the absence of the ratcheting — would exceed all sensible credence.
Many people don’t understand such nonrandom cumulative ratcheting. They think natural selection is a theory of chance, so no wonder they don’t believe it! The battle that we biologists face, in our struggle to convince the public and their elected representatives that evolution is a fact, amounts to the battle to convey to them the power of Darwin’s ratchet — the blind watchmaker — to propel lineages up the gentle slopes of Mount Improbable.
The misapplied argument from improbability is not the only one deployed by creationists. They are quite fond of gaps, both literal gaps in the fossil record and gaps in their understanding of what Darwinism is all about. In both cases the (lack of) logic in the argument is the same. They allege a gap or deficiency in the Darwinian account. Then, without even inquiring whether intelligent design suffers from the same deficiency, they award victory to the rival “theory” by default. Such reasoning is no way to do science. But science is precisely not what creation “scientists,” despite the ambitions of their intelligent-design bullyboys, are doing.
In the case of fossils, as Donald R. Prothero documents in “The Fossils Say Yes” [see the print issue of Natural History in which this article first appeared], today’s biologists are more fortunate than Darwin was in having access to beautiful series of transitional stages: almost cinematic records of evolutionary changes in action. Not all transitions are so attested, of course — hence the vaunted gaps. Some small animals just don’t fossilize; their phyla are known only from modern specimens: their history is one big gap. The equivalent gaps for any creationist or intelligent-design theory would be the absence of a cinematic record of God’s every move on the morning that he created, for example, the bacterial flagellar motor. Not only is there no such divine videotape: there is a complete absence of evidence of any kind for intelligent design.
Absence of evidence for is not positive evidence against, of course. Positive evidence against evolution could easily be found — if it exists. Fisher’s contemporary and rival J.B.S. Haldane was asked by a Popperian zealot what would falsify evolution. Haldane quipped, “Fossil rabbits in the Precambrian.” No such fossil has ever been found, of course, despite numerous searches for anachronistic species.
There are other barriers to accepting the truth of Darwinism. Many people cannot bear to think that they are cousins not just of chimpanzees and monkeys, but of tapeworms, spiders, and bacteria. The unpalatability of a proposition, however, has no bearing on its truth. I personally find the idea of cousinship to all living species positively agreeable, but neither my warmth toward it, nor the cringing of a creationist, has the slightest bearing on its truth.
The same could be said of political or moral objections to Darwinism. “Tell children they are nothing more than animals and they will behave like animals.” I do not for a moment accept that the conclusion follows from the premise. But even if it did, once again, a disagreeable consequence cannot undermine the truth of a premise. Some have said that Hitler founded his political philosophy on Darwinism. This is nonsense: doctrines of racial superiority in no way follow from natural selection, properly understood. Nevertheless, a good case can be made that a society run on Darwinian lines would be a very disagreeable society in which to live. But, yet again, the unpleasantness of a proposition has no bearing on its truth.
Huxley, George C. Williams, and other evolutionists have opposed Darwinism as a political and moral doctrine just as passionately as they have advocated its scientific truth. I count myself in that company. Science needs to understand natural selection as a force in nature, the better to oppose it as a normative force in politics. Darwin himself expressed dismay at the callousness of natural selection: “What a book a Devil’s Chaplain might write on the clumsy, wasteful, blundering low & horridly cruel works of nature!”
In spite of the success and admiration that he earned, and despite his large and loving family, Darwin’s life was not an especially happy one. Troubled about genetic deterioration in general and the possible effects of inbreeding closer to home, as James Moore documents in “Good Breeding” [see November issue of Natural History magazine], and tormented by illness and bereavement, as Richard Milner’s interview with the psychiatrist Ralph Colp Jr. shows in “Darwin’s Shrink,” Darwin’s achievements seem all the more. He even found the time to excel as an experimenter, particularly with plants. David Kohn’s and Sheila Ann Dean’s essays (“The Miraculous Season” and “Bee Lines and Worm Burrows” [See November issue of Natural History Magazine]) lead me to think that, even without his major theoretical achievements, Darwin would have won lasting recognition as an experimenter, albeit an experimenter with the style of a gentlemanly amateur, which might not find favor with modern journal referees.
As for his major theoretical achievements, of course, the details of our understanding have moved on since Darwin’s time. That was particularly the case during the synthesis of Darwinism with Mendelian digital genetics. And beyond the synthesis, as Douglas J. Futuyma explains in “On Darwin’s Shoulders,” [see November issue of Natural History Magazine] and Sean B. Carroll details further for the exciting new field of “evo-devo” in “The Origins of Form,” Darwinism proves to be a flourishing population of theories, itself undergoing rapid evolutionary change.
In any developing science there are disagreements. But scientists — and here is what separates real scientists from the pseudoscientists of the school of intelligent design — always know what evidence it would take to change their minds. One thing all real scientists agree upon is the fact of evolution itself. It is a fact that we are cousins of gorillas, kangaroos, starfish, and bacteria. Evolution is as much a fact as the heat of the sun. It is not a theory, and for pity’s sake, let’s stop confusing the philosophically naive by calling it so. Evolution is a fact.
|
|
|
|
|
Blunderov
Archon
Gender: 
Posts: 3160
Reputation: 8.91
Rate Blunderov
"We think in generalities, we live in details"
|
|
Re:The Illusion of Design
« Reply #1 on: 2005-11-29 02:01:39 » |
|
[Blunderov] Found this somewhere in the bitbucket:
"Those who reject the notion that biological systems exhibit intelligent design are guilty of a serious logical inconsistency. To illustrate this, consider the Sphinx. Even though we don’t know for certain which society created the Sphinx, everyone who looks at it realizes instantly that someone designed it. It is certainly not the result of random process involving erosion by wind and water, etc. Even though the Sphinx could be nothing more than a great accident, its inherent design is immediately obvious to everyone. However, compared to that first living organism capable of reproduction, the design of the Sphinx is relatively simple. The simplest organism capable of reproduction (by means of autonomous self-replication) must have been many times more complex than the Sphinx (with all our technology, we have yet to create such an organism in the lab). Therefore, if the design of the Sphinx is instantly obvious to everyone, then it is logical that the first living organism must also exhibit intelligent design."
The classic IDiot argument. ("Therefore, if the design of the Sphinx is instantly obvious to everyone, then it is logical that the first living organism must also exhibit intelligent design" is almost endearingly naive. There are the beginnings of reason here, but I digress.)
I have to wonder why it is that they (IDers) lay such stress upon actual life. Why is it THIS that they assert to be the phase transition between complexity and 'irreducible complexity' (as they call it)?
I have a largish quartz crystal knocking around here somewhere (unless the offspring have appropriated it for some fiendish purpose). It is quite, quite different from every single rock I have ever seen lying around in the veld. If I didn't know better, I would swear that someone had made it. It gives every appearance of having been deliberately designed by virtue of an 'unnatural' symmetry and also a complete variance from every other rock in the vicinity.
Also, very many celestial objects are circular in, err, design. In a random universe one might quite reasonably expect all object to be randomly shaped but not so! Something, or someone, must (!) have seen fit to make sure that all the points at the edges are just as far away from the centre as any of the others.
A creationist might accuse me of being disingenuous here. Everybody knows, he might say, that the force of gravity causes these objects to be round. I might retort that the force of gravity must therefore be intelligent.
Just look at all the sphinxes it made!
|
|
|
|
|
gaiaguerrilla
Initiate
Gender:
Posts: 58
Reputation: 5.84
Rate gaiaguerrilla
|
|
Re:The Illusion of Design
« Reply #2 on: 2005-12-29 05:28:00 » |
|
In my experience, those that reject the concept of evolution also reject the calculator lacking a percentile button. It's not so sad that people disbelieve darwin's claim; what's really sad is where they place their beliefs in social and familial bias. They think and do like mom and dad.
Darwin, like Gallileo, was detested because people exaggerated the assumption that they suggested our existence is not as "special" as we once thought. Imagine that we came to our capacity without being ordained. Imagine that living creatures can will their own bodies into change (where even science cannot fully explain), even if the change is minimal each generation.
I began to reason evolution as viable by several considerations: (1) Molecules are perfect mechanical parts. They turn, shift, bond, separate. They're the ultimate lego toy. (2) Such machines mixing in all sorts of ways over time after time after time can accomplish incredible things. (3) The odds against the accidental creation of life on our planet is not only 5 billion years of the planet. It's 5 billion years for every planet in the universe that originally held the potential to reach the same situation. If bubble theory holds true, 5 billion years for every planet in every other universe of similar creation to reach the same level. And with any of those beings capable of thought similar to us that may exist out there, they may have once thought "Imagine- the odds that it was us whom came to this level?" (4) When numbers begin to exceed human imagination (which by astronomy they certainly do), all concepts from those numbers also begin to exceed human imagination, including chance. I may trust my intuition, but I don't always trust my imagination.
Imagine- if someone told me that a calculator can work well without using the percentile button, but I could never test the results. Would I still find a way to believe them?
|
I've become the official property of a 10-year-old girl.
|
|
|
Gods Imaginary ...
Acolyte
Gender:
Posts: 15
Reputation: 5.00
Rate Gods Imaginary ...
Know Thyself
|
|
Re:The Illusion of Design
« Reply #3 on: 2006-03-28 21:57:37 » |
|
The most irritating point of the hurricane and sphinx arguments is that they assume evolution to be random. Any computer scientist will tell you that an evolutionary, or better yet, memetic algorithm is an advanced heuristic. Advanced heuristics will almost always outperform a random process, a random process has no measure of fitness for it's own solutions and no faculty to choose from among these solutions based on said fitness. Ironically the two approaches are equivilant in problems such as guessing the secret name of god mentioned in a paper below.
IDers show a complete lack of understanding for the replicating and natural selection mechanics of evolution. How Ironic that a memetic process drives them to pass this ignorance on to their children through their desire to influence the education system. If these IDers could see how an evolutionary algorithm allows a blind agent to consistently find one of the highest peaks in a seemingly infinite multi-dimensional landscape in reasonable time, perhaps they would find evolution a more compelling explanation for the natural world. If the blind agent analogy seems out of context, please read "On Evolution, Search, Optimization, Genetic Algorithms and Martial Arts Towards Memetic Algorithms" citeseer have copy, but the link is dynamic so I could not include it.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:The Illusion of Design
« Reply #4 on: 2006-03-31 11:59:26 » |
|
How would you describe the difference between genetic and memetic algorithms in the context of computer science?
|
|
|
|
|
Gods Imaginary ...
Acolyte
Gender:
Posts: 15
Reputation: 5.00
Rate Gods Imaginary ...
Know Thyself
|
|
Re:The Illusion of Design
« Reply #5 on: 2006-03-31 23:36:14 » |
|
That paper I mentioned explains it very well, though it is a quite a heavy read. Here is a brief look at how it would work:
Assume you have a maths equation with three unknown variables, x, y and z, such as z = 2x + y. An interesting way to visualise this 'solution space' is to imagine a 2D grid representing all values for x and y, where x gives the column and y gives the row. The grid is going to be very fine of course given that x an y are continuous values, so it is best to imagine the junction of an x and y value as a point, rather than a cell. If you then went to each point, took the x and y values represented by that point, and then put them into the afore mentioned equation, you will get a value for z returned. You can then use this value to define a position along a third axis, perpendicular to the grid. If you were to do this for every point on the grid you would end up with a landscape in 3 dimension, the tallest point of which would give the x and y values that maximise the equation.
In creating this landscape, you would be required to evaluate every x and y value possible, taking an extremely long time. This is equivalent to a random search for a solution. A smarter way to deal with the problem is to imagine yourself as an agent standing on this 3 dimensional landscape. As you haven't evaluated every x and y value you can't 'see' the landscape, thus you would be surrounded by a thick fog. You could however take a single step, measure your height (you happen to have an altimeter) and if your height is better than your previous position you could restart your approach from your current position, looking for the next best step to take.
The previous approach is called a local search for obvious reasons, and it allows you to find the highest peak in your local area. However, some of the more complex landscapes defined by more complex equations can contain plateau and multiple peaks, some higher than others. Using the previous approach, or algorithm, will cause you to run into trouble. On a plateau no step will yield a better position. The algorithm may also lead you to climb a peak near you, though this may not be the highest peak in the entire landscape. This problem of finding a good solution, but not the best solution is called local maxima. This local search algorithm finds an approximation quickly, but it is not guaranteed to find the optimal solution, thus it is called an heuristic.
Another approach to this problem is an evolutionary heuristic, rather than a local search. Here each solution can be represented as a binary string, so eight ones or zeros could represent the x value, and another eight could represent the y value. You now have a 16 bit string representing your position in terms of latitude and longitude. Rather than a single agent exploring the solution space, you increase the number to have a population of agents, each starting at a random spot in the solution space. Each of these agents have a 16 bit binary string to represent their position, this string is like their DNA. Without going into the mathematical details, an evolutionary algorithm will basically wipe out agents with poor altimeter readings, and bread the agents with better ones. Breading involves using standard evolutionary methods such as cross-over and mutation to create an offspring, in this case the two parents would each contribute their 16 bit DNA yielding a child with a 16 bit DNA string/location. This method is actually very effective in exploring the landscape and the majority of the time finds the highest peak. This is because such evolutionary algorithms are excellent at exploring.
One problem with evolutionary algorithms is that they may find the highest peak in a solution space, but they may not find the very tip of that peak due to the course nature of an evolutionary algorithm. However, what would happen if once a reasonable solution was found using an evolutionary algorithm, you applied local search? The first would get you to the best peak in the landscape, the second will find you the best position on that peak.
Each generation can therefore have an initial position through an evolutionary search, then improve this position with a local search. The final positions height can then be measured to give the solution/agent a fitness, which will then be used for the next round of breading and natural selection. This approach is called a memetic algorithm and it is very effective. The parallel with the real world is clear; the evolutionary step is like the creation of a person, the local search is like the learning that the person does throughout their life. Both the person’s genes and knowledge are passed on to their offspring.
For the mathematicians reading this who could use calculus to solve the initial problem, remember that this is a simplified example. Many computer science problems are all but proven impossible to solve through complexity theory. They may involve thousands of variables, which would create a landscape with a thousand dimensions. This is harder to comprehend but the same approach explained here can be used to explore it.
I have seen a memetic algorithm find better approximation and in less time than any other approach available. My supervisor back at uni used this approach to find approximations to some of the hardest outstanding problems in the field of bioinformatics.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:The Illusion of Design
« Reply #6 on: 2006-04-03 12:47:08 » |
|
OK, I can see how genetic algorithm + local hill climbing can yield good results but I'm not sure why you say that genetics algorithms have a 'course nature". Eventually most of population will end up gathered around the best solution found so far (with a few exploring the landscape for even better solutions), so most of the population will continue exploring the peak for the highest point on the peak.
I guess it depends (as always) on the exact nature of the fitness landscape. If the peak has several small peaks on top then a memetic algorithm may be stuck in a local optimum again while a genetic algorithm finds the real peak. I'm guessing this was not the case with the problems your supervisor was attacking.
|
|
|
|
|
Gods Imaginary ...
Acolyte
Gender:
Posts: 15
Reputation: 5.00
Rate Gods Imaginary ...
Know Thyself
|
|
Re:The Illusion of Design
« Reply #7 on: 2006-04-04 02:25:46 » |
|
I'm no longer at the uni, but this site details their current ambitions and is quite informative.
Newcastle Bioinformatics Initiative
To answer your question, the GA's are not making gradual increments to converge on the solution, they use rather abrupt manipulations like mutation and crossover that have a much higher chance of distorting the 'DNA'. For this reason they explore the space well, but they can 'bounce' around the final set of good approximations, abit like simulated annealing with the heat turned up to high.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:The Illusion of Design
« Reply #8 on: 2006-04-04 12:55:23 » |
|
Crossover doesn't have to be abrupt, it depends on the function that translates genotypes into phenotypes. I can easily imagine two solutions on either side of peak being combined with crossover to yield a solution on the peak.
Anyway, thanks very much for your explanations.
|
|
|
|
|
Gods Imaginary ...
Acolyte
Gender:
Posts: 15
Reputation: 5.00
Rate Gods Imaginary ...
Know Thyself
|
|
Re:The Illusion of Design
« Reply #9 on: 2006-04-11 18:44:32 » |
|
No problem,
with the memetic algorithms: keep in mind that the local search is used to improve the solution for every generation. In the begining you would want course crossover and mutation to explore the space well, the local search will then tidy the position found before the next round of breading. Also, some memetic algorithms only use the improved fitness gained from local search for the breading selection process, they don't actually use it to change the agents position. Finally, like you said, it can all depend on the crossover/mutation manipulators used. Memetic algorithms can use more generic operators to attack a wider base of problems, as the local search can smooth out any irregularities introduced by a such an approach.
The argument about memetic Vs. purely genetic is a valid one, and there are many papers that take both sides in the debate. My bias towards memetics is the fact that in some clustering problems that I have seen they outperform the best genetic approaches (see web site mentioned earlier for more details). Obviously this is not a universal test for complete dominance of one approach over the other, but it shows that in some cases memetics yields better solutions.
|
|
|
|
|
|
Logged
