Tuesday, May 22, 2007

The Blind Watchmaker, chapter 3.

This one has been talked about much due to the "me thinks it is like a weasel" example. Dawkins begins by setting up a straw man argument: Evolution is implausible, therefore it is impossible. This is trivially defeated because too many implausible things have become reality. He then sets out to imply the opposite: Evolution must be true because it is plausible. Everything from snake oil salesmen to Star Trek suddenly comes to mind.

The core of his case is the "me thinks it is like a weasel" example where he shows how a genetic algorithm using random letters can be rearranged in just a few generations. Creationists have principally objected that his algorithm chooses the best of each generation and this greatly exaggerates the convergence rate. Of course, if we have a large enough population together with some of the other properties of the problem, this kind of convergence is guaranteed, so this isn't the problem.

The real problem with the example is that it is a single objective, monotonic, constant slope function that is being optimized in a highly restricted design space. The constant slope guarantees that convergence will proceed at maximum speed while the other problem restrictions guarantee that the exponential blowups of probabilities won't be apparent in this case. In the class room, constant slope optimization is a rarity and a key reason that calculus is part of the curriculum.

Even this, however, is trivial compared to the single objective monotonicity of the problem. Mount McKinley is lower than Mount Everest, but it isn't all down hill from Everest to McKinley. A downhill only algorithm will leave you at the bottom of the ocean. Similarly, as we go from one viable software design to another viable software design, the number of non-viable intermediate states is frequently greater than astronomical. For evolution to be plausible, there must be a continuous, monotonic, viable path of fitness that connects from the simplest organism to the most complex. If even one step is missing, then the entire theory collapses. The experience of engineering design is that such a path sometimes exists between two nearly identical designs, but never in general. This is one of the elephants that Dawkins is trying to sweep under the rug.

The other elephant is that before Dawkins could do any of his computational experiments with a genetic algorithm, he needed to have a fully functioning, Intelligently Designed computer. Next, he needed to write a few dozen lines of software with each character being critical to the function of the whole. There are no points for "almost works" unlike the "me thinks it is like a weasel" example. The biologist knows that a certain minimum amount of design complexity is mandatory to represent a genome, preserve it, grow its host, and replicate it. He also knows that this minimum level of complexity is beyond his comprehension and a perfect solution is needed before Darwinian evolution can do anything at all. The Miller-Urey experiment in a way acknowledges the existence of this problem. What isn't acknowledged is that evolution is entirely silent on this issue. "Where did the first genetic replication machine come from?" remains just as intractable as "Where did God come from?".

11 comments:

WAR_ON_ERROR said...

"The real problem with the example is that it is a single objective, monotonic, constant slope function that is being optimized in a highly restricted design space. The constant slope guarantees that convergence will proceed at maximum speed while the other problem restrictions guarantee that the exponential blowups of probabilities won't be apparent in this case. In the class room, constant slope optimization is a rarity and a key reason that calculus is part of the curriculum."

But isn't the problem here that evolution is not the classroom? Surely Dawkins' example is simplistic, but it is only meant to convey the basic idea. The real idea is completely unfathomable. When we take it into the reality of speciation in the natural world, replicators are trying virtually everything out from things not optimized at all to things as maximally optimized as they can be. When we ask students, "How much money would you have at the end of the month if I gave you a penny on the first day and doubled your money every day?" Their guesses will not be at all accurate. Maybe they'll say $50, but in reality it is millions (if I recall correctly). We are not capable of processing the actual implications of the real version. We just don't deal with those kinds of numbers in our normal lives.

Granted, this brings up the issue of "If we can't process these numbers, how can evolutionists be so sure evolution works?" But notice, though, that this is now an *agnostic* case, and not a *negative* case that creationists need to take down evolution. And even if we don't have a computer simulation that can prove uphill climbs in nature's design space are possible, this leaves the door open for other evidence (like the age of the earth, and the relative positions of more and more complicated organisms in more recent geological stratas) to strongly suggest it does actually work out.

Ben

Looney said...

Ben, perhaps you can look at it a different way:

In order for Dawkins to construct a scenario for how evolution might work, he has subconsciously introduced Intelligent Design features into his evolution algorithm.

Per the first three chapters of The Christian Delusion, we all do things subconsciously to rationalize, and this is exactly what Dawkins is doing. My argument is that because Intelligent Design is so much a core part of the human psyche, it is impossible for intellectuals to conceptualize evolution without unwittingly mixing it up with intelligent design.

WAR_ON_ERROR said...

"In order for Dawkins to construct a scenario for how evolution might work, he has subconsciously introduced Intelligent Design features into his evolution algorithm."

Is he *introducing* ID, or is he sequestering the part of the vast spectrum of evolution (which just means change) which produces results compatible with ID?

Another way to look at it is thus:

We could have a football field filled with toddlers who can't aim guns worth anything. Then we have another football field where there is just one sharp shooter.

Dawkins is focusing on that one toddler who just so happens to be aiming in the same direction as the sharp shooter.

I would be afraid of being shot in either situation.

Ben

WAR_ON_ERROR said...

Oh, and I wanted to say I agree with you on your last point on TCD and biases. There are probably many evolutionists who are trying to battle their cognitive predispositions to see intentionality in the biological world. And there probably is a lot of cognitive dissonance and rationalizing going on. That's one of the many reasons I don't think it is productive to put such an authoritarian foot down on ID and creationism even though I believe they are still mistaken positions. It is a very natural and powerful mistaken position that requires a great deal of care to explain and accept.

Looney said...

"Is he *introducing* ID, or is he sequestering the part of the vast spectrum of evolution (which just means change) which produces results compatible with ID?"

Given that evolution has an infinite number of definitions, we can certainly argue that he is sequestering a part of the "vast spectrum of evolution". Something is always a subset of everything, but it is impossible to make any deductions from this.

From my experience as an engineering professional, the selection of the design variables and design space are crucial to determining if convergence is obtainable.

WAR_ON_ERROR said...

I thought you were criticizing Dawkins' weasel example because it was too specified. And I was trying to point out that he's only talking about the part of evolution where it happens to be getting it right. 99% of all species ever on this planet are extinct, so I hear. And supposedly life spent 3/4's of its 4 billion years not doing a whole lot other than producing things in the ballpark of single celled organisms. And even now most species are not as advanced in complexity as mammals and reptiles. There's a whole lot of evolutionary "failing" going on and the only reason we might think evolution succeeds in principle is by trying a whole lot of everything. And I don't think we can really fathom how much trial and error that really is, as I've been trying to show. So I don't think your experience as an engineer is exactly applicable.

It is still logically possible that perhaps the process is not able to flip the bill, but I don't see a strong reason to believe so. Only subjective difficulty in verifying it to the utmost degree.

Ben

Looney said...

In design optimization, we routinely hit places where the algorithm won't converge. It can run an infinite number of iterations with no further improvement, but an unspoken postulate of evolution is that this never happens. As we more variables and relax the constraints of the space, the chances of this happening increase exponentially.

The way we get around the algorithm locking is to change the design variables, design space or starting point(s) until we get something that converges. This how we play god in the intelligent design profession. When an evolutionist postulates a scenario based on a restricted design space with no potential lockups, he has inadvertently introduced god into the scenario through a back door.

The other point where Dawkins plays god is by introducing the evolution machine. My son reports that the molecular biologists have made no progress on this subject since the Miller-Urey experiment. Atheists still must embrace spontaneous generation or a god for the first bug that can do evolution.

WAR_ON_ERROR said...

And yet that doesn't prove that the historical evolutionary biological "algorithm" that played out on earth didn't "converge" since not all systems are the same. As I said, at best, all you can claim is "maybe not." I can't vouch for any evolutionists who may be advocating that it "had to happen" or that the process could never fail, but the open possibility in context of all the other evidence for the history of life on earth would seem to strongly indicate (at the very least) that it did.

Any time scientists are tinkering with evolution software to get it to work doesn't necessarily introduce ID. It just means they don't have the time to sit around and wait for a program that literally tries everything. They are still able to show the basic principles of natural selection at work. Surely you have to admit that there are many examples where the end result is a mystery to the evolutionary software designer. They don't know what the final product is because they didn't design it. And the computer program they set up can't be said to know what it is doing ahead of time either as though it is an intelligent agent. So how can there be design without an designer? It's at least a partial proof of concept that defies strict creationist categorization.

The example you give indicates what happens to one specific set of parameters when you open up the local task to more things. But what about an extremely large set of many different parameters? How can you be sure that one of those won't happen to correspond to what evolutionary software designers have conveniently sequestered?

Similarly with abiogenesis, there are literally trillions of galaxies. What molecular combination hasn't been tried out on all the worlds out there? Can you really say with certainty that with such unfathomable levels of trial and error, that the probability of something like abiogenesis happening nears exponential impossibility? I don't have that kind of confidence.

I can appreciate some skepticism on this front, but in contrast to all the absolutely implausible things we have to accept to get Biblical Christianity to work, it seems greatly unfair to assume abiogenesis and evolution cannot work out for the reasons given. Especially when the proof of concept one would have to have in order to completely verify these things (in the ways we are currently discussing) would be extremely hard to come by even if totally true.

Ben

Looney said...

I will just note again that all of the genetic algorithm experiments involve Ph.d's in a room with an intelligently designed computer doing something trivial.

Another item which you might give some thought to is the fact that the theory of evolution was never defined in Dawkins' book. My molecular biology text doesn't define it either. What some people people have proposed for evolution usually looks like "the change of alleles in a population over time". This effectively includes all the laws of physics, chemistry and biology. A consequence of this is that if someone claims they understand the theory of evolution, then they are claiming that they are a god-like supernatural intelligence!

In my experience, all scientific theories show up as a fixed equation involving measurable quantities like gravity: force, mass and distance. Einstein's equation for gravity has been fixed for a century, while Newton's for several centuries. Because evolution has no form, we can attribute god-like powers to it, but it is impossible to prove things one way or another. Or to put it another way, a statement can be shown to be true or false, but a non-statement cannot be proven to the true or false!

WAR_ON_ERROR said...

"I will just note again that all of the genetic algorithm experiments involve Ph.d's in a room with an intelligently designed computer doing something trivial."

Trivial? As I've said, neither the PhD's nor the intelligently designed computer know what the specific constitution is going to be at the end of the line. How did the algorithm solve the engineering problem? And the end result if someone else came along would indicate that it was directly "designed." You want to brush that under the rug, but I don't think that's legitimate. How else could they ever show that the principles of replication, mutation, and competition for resources can actually make something by accident? Just because an intelligent scientist drops a rock, that doesn't prove the theory of intelligent falling correct.

"What some people people have proposed for evolution usually looks like "the change of alleles in a population over time". This effectively includes all the laws of physics, chemistry and biology."

I don't really know what you are getting at.

Ben

Looney said...

Ben,

"I don't really know what you are getting at."

To put it simply, the scientific theory of evolution doesn't exist. The GA experiments which purport to show that evolution works are in fact so tainted by ID that they are probably 99.99% ID and only 0.01% evolution. When biologist claim they have explained something with evolution, you dig in and find that they are actually using ID paradigms, as Dawkins does throughout this book.

Or to put it another way, we can say, "Edison invented the light bulb" to produce an ID notion. The corresponding evolutionary statement is "The light bulb was invented". The evolution paradigm is simply a mirage of rhetoric, while the only paradigm that actually exists is the ID one.