Proof that evolution works

[QED]

Many people argue that evolution cannot work even in principle. This can be discounted by pointing to examples of practical applications based on the evolutionary principle. For example this report tells us that

Computer programs that "evolve" in ways that resemble natural selection can solve complex problems even their creators do not fully understand.

At the core of all evolutionary systems are genetic algorithms whether they be natural or synthetic, the same principles apply and are found to work as highly effective and autonomous design tools.

That nature uses more elemental ingredients is of no significance. The process is entirely scalable which is why we see evidence of evolution in a wide variety of different realms. Our own best tool for modelling evolution at the moment is the computer (our latest toy) but we may soon be able to work at the same atomic scale that nature uses and importantly, the transition would be a smooth one.

So given that we have taken a hint from nature and found her methods to be effective at producing such autonomous designs, why would we continue to doubt that this is the very system responsible for all the apparent natural "designs" we see around us? After all, would it not be an incredible coincidence if we had misinterpreted natures methods yet still ended up with a powerfully creative yet autonomous system purely by chance?

Additional material is widely available on the internet, for example:
The Genetic Algorithms Archive (a repository for information related to research in genetic algorithms and other forms of evolutionary computation.)
Genetic Algorithms and Artificial Life Resources
Wikipedias entry for genetic algorithms

[Gollum]

After all, would it not be an incredible coincidence if we had misinterpreted natures methods yet still ended up with a powerfully creative yet autonomous system purely by chance?

I agree with the primary thrust of your argument viz. genetic algorithms along with appropriate selection criteria provide a powerful "evolutionary" environment that appears to emulate the way that nature goes about natural evolution.

I'm a bit gun-shy about that last statement though. It sounds an awful lot like Michael Behe and IC. It has extremely low probability so someone must have designed it. You are of course driving at the conclusion that, because it is so improbable, the simplest way out is to conclude that our bits-and-bytes experiments with computers and nature's mechanisms are really one and the same ... just conducted using different computational timescales and platforms (i.e. Silicon Vs. DNA).

Fair enough but, just as we disparage the leap from complexity to design, I think we should also critically examine the leap from computing to nature just because they produce conceptually similar results.

[QED]

I'm a bit gun-shy about that last statement though. It sounds an awful lot like Michael Behe and IC. It has extremely low probability so someone must have designed it. You are of course driving at the conclusion that, because it is so improbable, the simplest way out is to conclude that our bits-and-bytes experiments with computers and nature's mechanisms are really one and the same ... just conducted using different computational timescales and platforms (i.e. Silicon Vs. DNA).

Fair enough but, just as we disparage the leap from complexity to design, I think we should also critically examine the leap from computing to nature just because they produce conceptually similar results.

Well, I did consider this point carefully and I want to be clear about this issue:

We are examining the principle of evolution here. This fact that we use computers to model the principle should be irrelevant. Such models could be continued to be refined, should the requisite technology become available, to the point where they operate at the very same level as nature. What threshold might there be at which this breaks down given that it has already been demonstrated to work at a much less refined level?

[juliod]

It has extremely low probability so someone must have designed it. You are of course driving at the conclusion that, because it is so improbable, the simplest way out is to conclude that our bits-and-bytes experiments with computers and nature's mechanisms are really one and the same

No. I think we should read QED's "by chance" as "by good fortune" rather than "by random events".

The concept developed like this. Computational math people are always looking for ways (algorithms) to find optimal solutions to complex equations or equation systems. There are many ways to do this, none of which I understand.

Eventually some people looked to evolutionary genetics for inspiration. They developed a fairly direct analogy of evolutionary priciples for what they called Genetic Algorithms. And it works.

Here's the relevant argument: By building up a functional system using bits and pieces in a one-to-one correspondance with the natural genetic system, they embodied good evidence that evolutionary principles work in practise and in a very different environment.

The argument is that genetic algorithms work because they mimic the genuine genetic system. If they had failed to mimic the genetic system then the algorithm would fail (except by chance, or "good fortune"). (I.e. It's possible that I could throw together some random code and generate an interesting an useful program, but it has never happened.)

If someone were to write a program than mimics Lemarkian genetics, or spontaneous generation, those systems would (assuredly) fail to be effective systems for optimization.

By the same token, you cannot use "creationist" means to optimize solutions. You cannot pray to god to create a solution ex nihilo, nor can you debug a program by "Casting" out "demons".

DanZ

[Gollum]

Such models could be continued to be refined, should the requisite technology become available, to the point where they operate at the very same level as nature.

There's a theorem in computing that says you cannot model a real-world physical process at the level of detail that the process uses any faster than it can happen in the physical world. That ('tis my understanding) has been proven using REAL MATH! ... don't know if the theoretical capabilities introduced by quantum computing makes a difference to that however.

To me the real test is that of any scientific hypothesis ... what kind of predictive power does it have? At the current stage of proceedings we are able to produce computational results and then find examples in nature that appear to be remarkably analogous to them. Presumably at the next stage or stages we will be able to predict that "given this starting point in nature and this set of selection pressures, we should see natural evolution following some path determined from the modeling exercise." AFAIK ... that hasn't happened yet.

[otseng]

Many people argue that evolution cannot work even in principle.

What exactly do you mean by "evolution"? If you mean simply "change", nobody is arguing that change cannot occur. If you mean the "theory of evolution", that is another matter. But, even in the ToE, there are some aspects of it that are undisputed, even among creationists. For example, microevolution is undisputed. One aspect of the ToE that is under dispute is common descent.

To show that genetic algorithms can change does not prove that common descent occurred. I might as well show how computer algorithms were created by programmers (which is by far how most algorithms come into being) and use that to prove that the universe and all life was also created.

[jwu]

I might as well show how computer algorithms were created by programmers (which is by far how most algorithms come into being) and use that to prove that the universe and all life was also created.

Could you explain this in detail?
I see how one could argue that since the algorithms were designed (necessarily in this case), evolution is designed - i.e. God "made" evolution. I don't see how one could get anywhere beyond theistic evolution based on this though.

[juliod]

To me the real test is that of any scientific hypothesis ... what kind of predictive power does it have? At the current stage of proceedings we are able to produce computational results and then find examples in nature that appear to be remarkably analogous to them.

No! Genetic algorithms are not a simulation of the evolutionary process. They are the use of evolutionary paradigms to produce something that we are otherwise unable to produce. (I.e. solutions to complex equations.)

I've used GA's in one application, namely, computational electrodynamics. Nasty stuff. I was looking at a system of equations where I had 15 variable, 14 of which were complex numbers. The equations were quite complicated, and included the function hyperbolic tangent. In other words, finding optimal solutions (minima, maxima, or whatever) is essentially impossible by normal means.

Solutions can be calculated, but I found that the time needed to compute solutions over the 15 dimensions at a useful resolution would have been 38 years.

But GAs can eat this up. You start with a population of 1000 random parameter sets, and let them evolve under a selective pressure. I did runs of 2000 generations and came up with a set of good solutions.

You could say that the GAs have predictive power because once they identify an optima, you can verify that it is an optima after all. But more important, they generate information that we don't have, and can't see any other method of acquiring.

So, GAs do not simulate biology. They use the priciples we established through biology to facilitate a computational application. And so, if the priciples weren't "true" the optimizing feats of GAs wouldn't happen.

DanZ

[juliod]

For example, microevolution is undisputed. One aspect of the ToE that is under dispute is common descent.

This is an obviosuly untenable position. Microevolution + reproductive isolation = macroevolution. There's no missing link there.

DanZ

[otseng]

I might as well show how computer algorithms were created by programmers (which is by far how most algorithms come into being) and use that to prove that the universe and all life was also created.

Could you explain this in detail?
I see how one could argue that since the algorithms were designed (necessarily in this case), evolution is designed - i.e. God "made" evolution. I don't see how one could get anywhere beyond theistic evolution based on this though.

I'm not trying to argue any point here except that the analogy is fallacious. Whether one shows that a computer algorithm evolved or was designed does not prove nor disprove the theory of evolution.

For example, microevolution is undisputed. One aspect of the ToE that is under dispute is common descent.

This is an obviosuly untenable position. Microevolution + reproductive isolation = macroevolution. There's no missing link there.

In the common descent thread, it should be noted that micro/macroevolution is not even used by Futuyma to argue for common descent. So, even the link between common descent and micro/macroevolution is questionable.