So You Think You Are a Darwinian?

Royal Institute of Philosophy ^ | 1994 | D. C. Stove

[Junior]

Then either you do not understand the concept of an argument from consequences, the theory of evolution, or both. Odds are, in light of your constantly making stupid or inane statements (remember the circle?) it is both.

[Doctor Stochastic]

Noam Chomsky and Roger Penrose...

Make that Roger Penrose...

[NukeMan]

Just for the record, it answers the question of where the "information" comes from. It does not need to come from precognition of cause and effect. mutation and selection are sufficient.

Wouldn't the ultimate source of the information be the fitness function the programmer added? The genetic algorithm is not going to converge on "Have a Nice Day" if the original fitness function says "Methinks it is like a weasel" is the optimal one....

[NukeMan]

In my experience, "Darwinian" or "Darwinist" as a term, exists only in the vocabulary of anti-evolution Creationists.

FWIW, the author of this piece (David Stove) is an atheist. Dunno if that affects your perception of this article or not!

[PatrickHenry]

"How come, when g3k is around, Con X Poser isn't"

Maybe even they find one another intollerable?

[js1138]

Wouldn't the ultimate source of the information be the fitness function the programmer added?

How is that different than saying selection is the source of the information -- other than semantics?

The key thing being addressed here is the claim made by IDers that the laws of probability prevent anything complex from arising through a random process.

The patent application for a new electronic circuit completely blows this away. This is not imitating something already done, like a phrase from Shakespeare. This is creating something new and useful, something that has never been seen on earth before, and something beyond the understanding of the team that programmed the computer.

This simple demonstration proves that mutation and selection combined can produce new information. Interestingly, the circuit doesn't work if you change or remove any of the components, so it meets a key criterion for being "irreducibly complex".

[js1138]

For it to work as evolution supposedly does it would have to be able to write an original Shakesperean play.

The program I described in the article designed a useful electronic circuit that has never appeared on earth before -- hence the patent application. And this is just a baby step. So it blows away two of your "impossibilities" -- originality of product and complexity arising through a random process of mutation and selection. the interesting thing about the new design is that it does NOT produce the "why didn't I think of that" response. It is really new and very surprising.

[js1138]

Make that Roger Penrose...

It's interesting that great geniuses cannot always filter out the gibberish they generate, but the community of thinkers manages to select the new and useful. Reminds one of analogous processes.

[and the horse you rode in on]

Anyone know the difference between a lib attacking character as their only and immediate defense, and an evoloonist?
me neither

Is that supposed to mean something?

Be an answer to anything?

What in the world makes you convolute liberalism and Darwinism? Conservative Scientists, the people who make the world run for the rest of you, are almost exclusively Darwinists, just like the Liberal Scientists who whine and teach. It is a matter of brains, not politics.

[NukeMan]

How is that different than saying selection is the source of the information -- other than semantics?

Well, in genetic programming or genetic algorithms, there exists a final cause - a goal, a template - to compare members of the population for 'fitness'. Evolution has no goal at all, if I understand the theory correctly. If the programmer put in different criteria (i.e. a different goal) a different circuit would have been formed. The programmer is inserting information (albeit in a diffuse and indirect way). Evolution is not supposed to have a goal, only differential survival.

[js1138]

The programmer is inserting information (albeit in a diffuse and indirect way). Evolution is not supposed to have a goal, only differential survival.

These are good thoughts and need to be addressed.

To say that evolution has no goal or destination is not entirely true. Differential survival is not a trivial or random thing. It is an absolute judgement on "designs". If you introduce penicillin intro the environment of bacteria, it will produce a focused outcome, perhaps several solutions, but all meeting the selection criterion.

Genetic programming, as tortoise asserts, may not revolutionize computer design and programming, but it has proved that new and unexpected designs can emerge by "random" mutation guided only by the fitness function. That completely blows away a central tennant of ID, which states that such outcomes are forbidden by the laws of probability.

Does it prove biological evolution? Of course not. But it supports it.

[js1138]

I might add that the entire article takes about six pages of the current issue of Scientific American, and lists many other "inventions" (most of which are not absolutely new). Some early twentieth century circuit inventions were replicated on a desktop computer. The more complex circuits were replicated on an array of 1000 Pentium II 350s -- a poor man's supercomputer. The number of calculations required is astronomical, but not at all impossible.

The really interesting thing is that these results can easily be tested and replicated by skeptics of the process. IDers are welcome to dig into the process and try to prove there is a midget lurking in the box.

[NukeMan]

Well, to the extent it truly models nature, it supports evolution. But GAs and GPs are simplified models of what some biologists suspect happens. They have to be simplified, else the problem would be intractable.

I'll agree that GA/GP can produce bizarre, unforseen, fascinating solutions. I started tinkering with them after Koza's first book came out (highly recommended). Again though, the imposition of a fitness function is a HUGE 'cheat' on the whole process. You don't have to take my word for it - I am merely Nukeman, semicompetent befuddled layman. Here is John Maynard Smith Evolutionary Genetics(1998)(page 307) "Perhaps it is the fact that the program has a representation of the optimum message and determines the 'fitness' of actual messages by comparing them to the optimum. No analogous process occurs during natural selection."

Programmers have simplified the process in other ways in order to solve the problem: selection pressure is high, and selection is immediate and certain. The programs can combine subtrees of instructions (if you are using a parse-tree representation a la Koza) without any restrictions. Pleiotropy and polygeny are usually missing. There are other simplifications. The circuit designers (or what-have-you) are usually trying to design a circuit, not solve our metaphysical curiosity.

The ID'ers would claim that the process is not at all like nature, and they score points with that assertion (at least with me). In the end I suspect that ID'ers and traditional biologists will simply argue endlessly over what the true probabilities are, what is a reasonable representation of nature and what is a cheat, etc. In short, nothing will change!

[js1138]

The ID'ers would claim that the process is not at all like nature, and they score points with that assertion (at least with me). In the end I suspect that ID'ers and traditional biologists will simply argue endlessly over what the true probabilities are, what is a reasonable representation of nature and what is a cheat, etc. In short, nothing will change!

I can't argue with your last sentence. But genetic programming does prove that you cannot simply assert something is impossible due to improbabiliy.

As for the assertion that a selection function is a cheat, I would point out that only the final behavior is predetermined, not the structure of the solution. We simply don't know enough about biochemistry to assert there aren't equivalent "cheats" built into the structure of nature.

According to ID, the process should not work at all.

I'm willing to stake my personal credibility on the assertion that these are baby steps, and that within 10 or 20 years, computers will routinely find design solutions worthy of patents. In fact, I would be surprised if they are not doing most routine programming.

[Ethan Clive Osgoode]

genetic programming does prove that you cannot simply assert something is impossible due to improbabiliy.

You don't need genetic programming to tell you that. All you need enough common sense to notice that "improbable" does not mean the same thing as "impossible". Fairytales may be improbable, but not necessarily impossible.

[js1138]

In this case, "impossible" simply means having such a low probability that it won't happen in a specified period of time. IDers claim to have calculated the probability of certain things happening. The success of this programming method proves that useful and previously unknown things can arise through mutation and selection.

That does not "prove" it happens in biology, but it demonstrates the weakness of the "improbability" argument.

[balrog666]

...but it demonstrates the weakness of the "improbability" argument.

Weakness bump!

[js1138]

Waiting for the Blue LaGoon bump.

[tortoise]

Make that Roger Penrose...

I was being generous.

[Ethan Clive Osgoode]

It is improbable that I throw 50 sixes in a row on an unloaded die, but not impossible of course. Everyone knows that. Well, everyone but Darwinians, who need massive computing power to figure it out. I'm not at all surprised that Darwinians would wave computer simulations around as impressive demonstrations of something that every reasonable person already knows: that improbability and impossiblity are two different things.