Tom Schnieder’s ‘the And-multiplication Error’ Article Refuted

[Guest George R]

More comments from Tom Schneider:

 

http://www.lecb.ncifcrf.gov/~toms/paper/ev/blog-ev.html

 

~~ Paul

1633[/snapback]

 

As a courtesy I can add that the very long list of notes on the cited page are chronolological. You will see that notes from April 21 2005 & on are relevant to this forum as they directly reference this thread.

 

I shamelessly copy them here to save the thread from unintelligiblity as readers are focred to follow a line of debate splattered across other web sites.

 

2005 Apr 21: Evolution Fairytale Forum -> Tom Schnieder's [sic] 'the And-multiplication Error' Article Refuted http://www.evolutionfairytale.com/forum/in...p?showtopic=165 Gosh they pile up the errors quickly ... One or two notes: Not responding to all objections does not mean that reasonable answers do not exist! It means I have more important things to do than to respond to everything. Learn the basic information theory before making comments. The ev program does not evolve a genetic code. Read the paper. If you don't like the form of selection, make it different (e.g. probabalistic). Try it out yourself! The goal of ev was to see if Rsequence evolved to Rfrequency. The answer was yes. How about dealing with that? Does Ev show information increase based entirely on replicatin, mutation and selection? Of course! Look at the data!

 

2005 May 1: In the Evolution Fairytale Forum http://www.evolutionfairytale.com/forum/in...p?showtopic=165 at Apr 29 2005, 04:35 PM, Fred Williams complains that the "program is not real-world, not even close. New information was not created naturalistically." It is not clear what he means by 'naturalistically'. If you read the Ev paper carefully you will note that the model parallels the natural situation. There are places on the genome where having a binding site would be advantageous whether or not there is a site there. Genomes are of a certain size. These two factors determine Rfrequency in natural genomes and the situation is the same in the ev simulation. So Fred Williams needs to be explicit about what his problems are. I suggest that he do some honest work and write his own program. I think he will find when he does this that the Ev program is about the minimum coding that one can do that still matches the natural world. Sure, you could do molecular modelling in three dimensions of molecules diffusing through space and binding. Unfortunately you won't have enough computer power to do this simulation. So replace the molecular modelling with a model of a sequence recognizer - a weight matrix. These are well recognized in the molecular biology literature, if you read even a few papers! The sequence does not need to be three dimensional, it can be a string in the computer. That does not alter the computation of the Shannon information at all, so it is still a good model.

 

2005 May 1: In the Evolution Fairytale Forum http://www.evolutionfairytale.com/forum/in...p?showtopic=165 at Apr 28 2005, 09:07 AM Fred Williams complains that codes do not evolve. I suggest reading: Shannon1949 and ccmm. All specific molecular interactions in biology are part of codes. See the glossary definition: code.

[Guest 92g]

Fred Williams complains that codes do not evolve. I suggest reading: Shannon1949 and ccmm. All specific molecular interactions in biology are part of codes. See the glossary definition: code

I don't think he understands the complaint. I wont pretend to speak for Fred, but for me, the question at the fundamental level does not center around the information stored by means of the genetic code, and I think that what Mr. Schneider is attempting to manipulate with his program, but the origin of the genetic code itself.

 

Terry

[Guest George R]

I think he will find when he does this that the Ev program is about the minimum coding that one can do that still matches the natural world.

This is a telling comment from any designer of a model.

 

If I tell you that my model of the real world contains the minimum number of steps and variables to produce an intended result, I am saying that it is irreducibly complex.

 

If true, the irony is that ev is self-refuting: It will not achieve the results if parts are removed. So if it does generate information, but also demonstrates that this achievement can be the product of a irreducibly complex algorithm.

 

It may indeed be ideally the product of an irrreducible complex algorithm once the cost of reproducing "life" that contains any reproductive algiorithm as part of its own replication.

 

Before all the hand-waving starts about redundant computer steps and components that would in fact make the mechanism "better", bear in mind that (1) there is a real "cost" of buying theoretical add-ons in algorithms and in nature, and (2) the edxistence of a spare limb will not necessarily prepare or place that limb to take on the task of a specialized limb eg a broken right leg... nor does an extra add instruction nmake for a "takover" of a missing one in an algorithm (3) that a hit on one of the irreducible parts would terminate replication.

 

Granting FTSA that ev does perform the task that the author set out for it to achieve, the maintenance of the algorithm by a similar process as part of replication would seem not to be possible, and yet a looser "spare parts" algorithm would come at a higher cost and not necessarily achive any positive results.

[Guest 92g]

This is a telling comment from any designer of a model.

 

If I tell you that my model of the real world contains the minimum number of steps and variables to produce an intended result, I am saying that it is irreducibly complex.

 

If true, the irony is that ev is self-refuting: It will not achieve the results if parts are removed. So if it does generate information, but also demonstrates that this achievement can be the product of a irreducibly complex algorithm.

 

Excellent points...

 

Terry

[Modulous 0]

I don't think he understands the complaint.  I wont pretend to speak for Fred, but for me, the question at the fundamental level does not center around the information stored by means of the genetic code, and I think that what Mr. Schneider is attempting to manipulate with his program, but the origin of the genetic code itself.

 

Terry

1732[/snapback]

 

That's fine, really. The program isn't there to demonstrate the origins of the genetic code, I think we can agree here. It was orginally brought up by Paul C when he said "Information can arise from random sequences of DNA, as Tom Schneider has shown with Ev.", but then mistakenly said "Tom Schneider's Ev program creates a code from random DNA.". Obviously this is an error. Creating a code from a random code doesn't really make any sense, right?

 

But we are essentially agreed, the origin of codes isn't the issue with the Ev program. That side of the debate seems to have stemmed from an overlooked mistake.

 

This is a telling comment from any designer of a model.

 

If I tell you that my model of the real world contains the minimum number of steps and variables to produce an intended result, I am saying that it is irreducibly complex.

 

If true, the irony is that ev is self-refuting: It will not achieve the results if parts are removed. So if it does generate information, but also demonstrates that this achievement can be the product of a irreducibly complex algorithm.

 

It may indeed be ideally the product of an irrreducible complex algorithm once the cost of reproducing "life" that contains any reproductive algiorithm as part of its own replication.

1733[/snapback]

Ironic, given that the program actually evolves irreducibly complex structures isn't it? I think we can agree that Schneider hasn't programmed artificial intelligence Unfortunately, your argument rests on the idea that irreducibly complex things must have been designed and you neglected to actually demonstrate this to be true. If you want to start up a thread about ID and irreducibly complex structures, I'm more than happy to join in.

[Guest Paul C. Anagnostopoulos]

If I tell you that my model of the real world contains the minimum number of steps and variables to produce an intended result, I am saying that it is irreducibly complex.

So you're saying that Behe's idea was that every object and process has a fundamental core of components that, when reduced, will render the object or process "broken"? That's not what he said, but let's assume it.

 

Such a statement is vacuous. It renders every object and process irreducibly complex at some point. Consider the process of snowflake formation. If we remove water vapor, temperature, gravity, or various other components, we can't obtain snowflakes. Are they irreducibly complex?

 

I thought the idea was that things are potentially irreducibly complex only if they are contingent.

 

~~ Paul

[Guest Paul C. Anagnostopoulos]

That's fine, really. The program isn't there to demonstrate the origins of the genetic code, I think we can agree here. It was orginally brought up by Paul C when he said "Information can arise from random sequences of DNA, as Tom Schneider has shown with Ev.", but then mistakenly said "Tom Schneider's Ev program creates a code from random DNA.". Obviously this is an error. Creating a code from a random code doesn't really make any sense, right?

Schneider thinks that Ev creates a code, and I agree. It is not the genetic code; no one ever said that. It is a simple code that distinguishes binding sites from all other sites. If someone disagrees, then he needs to try harder to define what a code is, ensuring that the definition doesn't rule out everything but the extant genetic code.

 

Unfortunately, your argument rests on the idea that irreducibly complex things must have been designed and you neglected to actually demonstrate this to be true.

And the entire idea is only interesting if the definition of irreducible complexity isn't a tautology.

[Modulous 0]

Schneider thinks that Ev creates a code, and I agree. It is not the genetic code; no one ever said that. It is a simple code that distinguishes binding sites from all other sites. If someone disagrees, then he needs to try harder to define what a code is, ensuring that the definition doesn't rule out everything but the extant genetic code.

1746[/snapback]

Then maybe I missed a major part of his paper. Where does he say anything about the code being created by the program?

[Guest Paul C. Anagnostopoulos]

He says nothing about it in the paper. He said it to me personally, and has now mentioned it in his blog.

 

If we could just define code carefully, we could decide for ourselves whether Ev creates a code.

 

~~ Paul

[Guest 92g]

He says nothing about it in the paper. He said it to me personally, and has now mentioned it in his blog.

From Schneider's quotes above:

 

The ev program does not evolve a genetic code. Read the paper

Terry

[Guest Paul C. Anagnostopoulos]

Correct, it does not evolve something like the extant genetic code. It does evolve a code, however. There is more than one possible code!

 

Let me see if I can get him to clarify, or state that I am wrong.

 

This is an awful lot of argument over a topic we could clarify if we would just define code.

 

~~ Paul

[Fred Williams 0]

Before I critique the latest comments, I want to qualify a few things, comments I have made in the past but probably not yet on this forum. Dr. Schneider’s work with sequence logos is both interesting and IMO a good approach at identifying binding sites. I think this aspect of his work is good science. I also find his description of Shannon information as a “reduction in uncertainty” to be correct, it is the same way I interpreted Shannon’s paper. Both evolutionists and creationists have stumbled on this, often thinking uncertainty is information, which it is not, nor did Shannon intend it to be taken that way. Where we quickly diverge though stems from our worldviews, Schneider’s being that all life arose naturalistically. Whenever you start from a flawed assumption, you will ultimately produce bad science, and I believe Schneider’s Ev program has been overwhelmingly proven to be bad science and I think it would serve him best to abandon it and remain focused on his sequence logo work.

 

Now back to the latest comments put forth …

 

He says nothing about it in the paper. He said it to me personally, and has now mentioned it in his blog.

 

If we could just define code carefully, we could decide for ourselves whether Ev creates a code.

I think there is a good reason Tom told you this personally and did not mention it in his paper. Even through a biased evolutionary journal, this claim would have been tough for any reviewer to swallow because it is blatantly incorrect.

 

As I pointed out in this post back in April, specifically item #1:

 

“Easily the most common tactic used by evolutionists is to equivocate on what a code is. ”

 

On several occasions on this board we have provided a definition for a code. I know Terry has, and speaking for myself I have offered this straightforward definition to work with:

 

Code:

 

A formal communication system consisting of symbols, syntax, and semantics. Examples: English language, Morse code, Fortran, ‘C’, Swahili, genetic code.

 

This is what we mean when we say “code!” It is essentially synonymous with “language”. Gitt takes it a step further and adds “apobetics” and “pragmatics”, which I believe are fine and good, but in my opinion are really extensions of “semantics” and already implied. So the definition above keeps it simple and straightforward.

 

Schneider’s definition of a code is an equivocation because it is too encompassing. He first starts out equating a code to a “message”, which essentially is correct, but then goes too far by calling parts of the message a code. Even a standard definition out of Webster’s shows Schneider is reaching:

 

Code: “a system of signals or symbols for communication”

 

A part of a message, say a single symbol, is not a system itself of symbols! He then uses the worst possible example, a “parity bit”. No one I know would ever call a parity bit a code. While it is true a parity bit protects a message against error (albeit partially), it is not part of the coded message. Instead it is a hang-on symbol, it does not have semantics in the true sense of any language (i.e. there is not a single human or computer language that defines syntax or semantics to a parity bit). The parity bit is in fact wasteful in message transmission - the little good it adds in partially protecting the message is easily outweighed by the waste of using the bit in the transmission. It has been completely obsoleted by more robust methods, such as checksums and CRCs[1].

 

As I mentioned in a prior post, the word syntax has been used interchangeably for both individual symbols, and their defined order. But I have never once heard code used interchangeably with its parts, until I read it on Schneider’s web site.

 

So going forward on this forum we will use the above definition I gave for a code, that way we can put an end to the endless repetition of equivocating over this. I’ll put a topic together and pin it, or perhaps come up with some sort of FAQ or glossary to handle these types of issues. I will grant that due to the “de-evolution” of our language, we do need to be clear in our definitions, and code is one of those words that has kind of taken a life of its own and why we need to clearly define it here (the word information and evolution are other obvious examples that need to be clearly defined).

 

But all of the above aside, Schneider still has not addressed the many serious problems that have been brought forth against his program. He mentioned that he cannot answer every challenge, and while I sympathize with that, he should of at least seriously tackled the holes that Dr Royal Truman pointed out, holes that I re-emphasized in my OP and are simply fatal to his claim of naturally evolving information. He continues to ignore the fact that his program uses severe truncation selection, something that does not occur in nature yet he continues to claim his program simulates nature! His mutation rate is also too high. Schneider replied to this by saying “Obviously if the rate were made lower (only 10 fold lower to reach HIV rates) the evolution would take longer, but the essential result should be the same.” The reason the result is the same, is because his program does not take into account the negative effects of an accumulating genetic load - he doesn’t let it happen by using severe truncation! As Truman pointed out “Error catastrophe would be inevitable.” [2] This alone refutes his program completely and convincingly as nothing more than an illusion. For icing on the cake, perhaps arguably the most untenable parameter in his program is the high rate of positive selection, essentially 50%! Some GA programs use 1 in 1000, and even many evolutionists will acknowledge that is a reach!

 

There is another way to look at this. Consider the fact that Schneider’s program, at least on my inspection of it, appears to be set up to work every time. It’s odds of success are unity! This illustrates the absurdity of the claim that his program shows information is added naturalistically (random change plus natural selection). Whenever a program has a guaranteed outcome, we know the information is already present from the beginning. Consider the following analogy: two engineers are given the task of writing a program that prints “hello world”. One engineer uses the easy way and his program simply prints the message. A less competent engineer uses random mutation with truncation selection to achieve the same result over many iterations. They both achieve the same programmed outcome, and hence have the same amount of Shannon information (we could rightly argue the less efficient version has less information, but we won’t here for the sake of argument). Both programs, before they run, have the same amount of information – they already have encoded the phrase “hello world” within their program. Schneider’s program already has the information Rsequence ~= Rfrequency. In the end, his program is no different than Dawkin’s “METHINKS IT IS LIKE A WEASEL” program that has long since been refuted. It is no wonder Schneider can’t see the obvious given that Dawkins was one of the reviewers of his paper.

 

I understand that it would be very hard to come to grips that something you wrote and have pushed for such a long time is seriously flawed, after putting so much work and effort into it. I hope Dr Schneider remains focused on his sequence logo work with the working genome, and eventually comes to realize he is not making productive use of his time defending his fatally flawed Ev program. As the noted physicist Richard P. Feynman once said,

 

“The first principle is that you must not fool yourself, and you are the easiest person to fool.”

 

---

1. http://www.cnet.com/Resources/Info/Glossar...rms/parity.html

2. http://www.trueorigin.org/schneider.asp

[Guest Paul C. Anagnostopoulos]

There is another way to look at this. Consider the fact that Schneider’s program, at least on my inspection of it, appears to be set up to work every time. It’s odds of success are unity! This illustrates the absurdity of the claim that his program shows information is added naturalistically (random change plus natural selection). Whenever a program has a guaranteed outcome, we know the information is already present from the beginning. Consider the following analogy: two engineers are given the task of writing a program that prints “hello worldâ€ÂÂ. One engineer uses the easy way and his program simply prints the message. A less competent engineer uses random mutation with truncation selection to achieve the same result over many iterations. They both achieve the same programmed outcome, and hence have the same amount of Shannon information (we could rightly argue the less efficient version has less information, but we won’t here for the sake of argument). Both programs, before they run, have the same amount of information – they already have encoded the phrase “hello world†within their program. Schneider’s program already has the information Rsequence ~= Rfrequency.

The chromosomes in the Ev creatures do not have Rsequence = Rfrequency when the simulation starts. Yet they do after sufficient generations are simulated. Where does the information "come from"? Obviously from the design of the program: Information is "imparted" to the chromosomes by the cyclical application of evaluation, selection, and mutation. We might say that information is transfered from the environment to the organisms. And that is exactly what we can say about real organisms, too.

 

Regarding the code that Ev evolves: It has symbols, syntax, and semantics. Your definition is fine, but you appear to have an additional unstated requirement that the code look like something you think of as a "language."

 

~~ Paul

[Guest Paul C. Anagnostopoulos]

He continues to ignore the fact that his program uses severe truncation selection, something that does not occur in nature yet he continues to claim his program simulates nature!

What would be a more realistic selection process?

 

~~ Paul

[Fred Williams 0]

FW: Both programs, before they run, have the same amount of information – they already have encoded the phrase “hello world” within their program. Schneider’s program already has the information Rsequence ~= Rfrequency.

The chromosomes in the Ev creatures do not have Rsequence = Rfrequency when the simulation starts. Yet they do after sufficient generations are simulated.

My two examples also do not have the target before the code runs! You have to run the code (or simulation) first, then you will have your target. Just like Schneider’s program, the information is there from the start, this is simply indisputable! I cannot stress enough just how erroneous the conclusions are that Schneider draws from his program. It’s really no different than Dawkins long since debunked “Weasel” program. Given all the other problems mentioned, the fact that the program already has the information encoded within it is mere icing on the cake in proving this illusion.

 

Where does the information "come from"? Obviously from the design of the program:

Here you unwittingly make my point for me. Thanks!

 

Information is "imparted" to the chromosomes by the cyclical application of evaluation, selection, and mutation. We might say that information is transfered from the environment to the organisms. And that is exactly what we can say about real organisms, too.

You can say all you want, you can say the moon is made of cheese. But that does not make it true. There is no evidence anywhere to support Schneider’s parameters, none. Zippo. Show me a study on truncation selection in nature (do a google search of this and my name and website comes up more than studies do!). Show me a study where the beneficial mutation rate is 50%.

 

Regarding the code that Ev evolves: It has symbols, syntax, and semantics. Your definition is fine, but you appear to have an additional unstated requirement that the code look like something you think of as a "language."

You are trying to do an end run around my point. Please respond with evidence and/or an explanation of how his program produced a new code (symbols, syntax, and semantics). Just don’t say it is so, that is a waste of everyone’s time.

 

Fred

[Fred Williams 0]

FW: He continues to ignore the fact that his program uses severe truncation selection, something that does not occur in nature yet he continues to claim his program simulates nature!

 

What would be a more realistic selection process?

That’s a fair question. First, the process should account for the various costs in moving a mutation. For example, even if a mutant receives a beneficial mutation, there is no guarantee it will survive to the next generation. Genetic deaths can occur due to random death, i.e. a rock falling on your head, being a prude, suffering a lethal mutation, being a homozygote when the trait is heterozygote advantage, and other costs (Haldane estimated there was probably only a 10% reproductive excess available to move favorable traits.).

 

So after a “beneficial” mutation is added, go through and account for the genetic deaths. Keep track of the genetic load while you go. Keep track of each organism’s fitness, so when you apply “selection” you do proper probability calculations and not automatically eliminate the most unfit. For example, if organism A is 51% fit compared to the rest of the population, and organism B is 49 % fit, there still is at least a 49% chance organism B could survive when selection is applied across the population to meet some the pre-programmed goal for that generation (i.e. the goal may be to keep the population at a constant size every generation; or the program can be set to allow the population to grow at 1% per generation, etc. The point is, organism B could easily survive to the next generation, unless you apply truncation selection and remove him just because he was on the wrong side of the curve, albeit barely).

 

You’ll be hard-pressed to see an evolutionist ever program a model like this, because they know it is doomed to fail, unless they use fantasyland assumptions like Schneider did and crank up the beneficial mutation rate to 50% and apply truncation selection.

 

Fred

[Modulous 0]

Just like Schneider’s program, the information is there from the start, this is simply indisputable!

 

What information is there from the start?

 

Here you unwittingly make my point for me. Thanks!

 

The program design acts like the laws of nature. Since the program was not designed to simulate the creation of rules it doesn't need to account for how the rules get there, just that there are rules. I am not suggesting that the programs rules simulate nature, but that laws of nature exist before life does. The program shows us that random mutation, coupled with a selection method can increase information content without any information theory violations or thermodynamic violations or any other violations of known physical laws.

 

The program wasn't designed to tackle Haldane's dilemma or reproductive load. They are all fair arguments, that the program was not intended to tackle. The program was not intended to show everything. It was only designed to show that random mutation coupled with a selection method can lead to an increase in Shannon information. I assume you don't dispute this?

[Fred Williams 0]

What information is there from the start?

1877[/snapback]

As I said a few posts ago (Post #37), “Schneider’s program already has the information Rsequence ~= Rfrequency. In the end, his program is no different than Dawkin’s “METHINKS IT IS LIKE A WEASEL” program that has long since been refuted.”

 

The program design acts like the laws of nature.

No, it doesn’t. Nature does not use truncation selection, nor does nature produce a 50% beneficial mutation rate, nor does nature prevent error catastrophe (extinction).

 

The program wasn't designed to tackle Haldane's dilemma or reproductive load. They are all fair arguments, that the program was not intended to tackle.

I never said it had to tackle “Haldane’s Dilemma”. Paul asked how it could use more realistic selection, and I explained how. Just because the terminology is the same that Haldane used, it is also the same terminology any population geneticist should use and itself is not “Haldane’s Dilemma” per se.

 

It was only designed to show that random mutation coupled with a selection method can lead to an increase in Shannon information. I assume you don't dispute this?

You assume incorrectly. I submit that since his program guarantees Rsequence ~= Rfrequency every time, that no new information is created since Rsequence ~= Rfrequency is encoded in the program before you run it. Now does this mean that “random mutation coupled with a selection” cannot produce new information? Of course not! That is, provided the selection method is intelligent, you can certainly produce new information. For example, a Genetic Algorithm can be given a target or a range of solutions to look for, and run ad infinitum until something useful is detected. This requires intelligence to determine when that useful thing has occurred. The aerospace industry, for example, has occasionally used programs similar to Genetic algorithms to search for various solutions. These are really no more than trial&error experiments, and the only reason any new information is ever culled is because intelligence is present (an engineer) to harness the information and put it to use. The point is this: an increase in information is impossible outside the presence of already-existing intelligence (i.e. an intelligent sender).

 

This thread is now going in circles. Unless there are any novel ideas that are added to this thread, I see no reason to not close it. For example, if Paul could explain the “code” (a communication system comprised of symbols, syntax, semantics) that Schneider’s program produced, we could discuss that. Or if someone provided evidence or any citation on truncation selection occurring in nature, etc. Otherwise we have reached the point where we are just repeating ourselves.

 

Fred

[Modulous 0]

As I said a few posts ago (Post #37), “Schneider’s program already has the information Rsequence ~= Rfrequency. In the end, his program is no different than Dawkin’s “METHINKS IT IS LIKE A WEASEL†program that has long since been refuted.â€ÂÂ

1880[/snapback]

Rsequence (information in binding sites) starts off as approximately 0 am I right? Does that not imply that it has no information?

 

 

No, it doesn’t. Nature does not use truncation selection, nor does nature produce a 50% beneficial mutation rate, nor does nature prevent error catastrophe (extinction).

 

 

You missed a part of my quote there: . I am not suggesting that the programs rules simulate nature, but that laws of nature exist before life does. I never implied that nature uses truncation selection. I have frequently agreed with you on this matter. The program has to have physical laws in order to do anything. So, yes, these laws are designed, these laws aren't what the program addresses.

 

 

You assume incorrectly. I submit that since his program guarantees Rsequence ~= Rfrequency every time, that no new information is created since Rsequence ~= Rfrequency is encoded in the program before you run it.

 

As above really. What is the information content of Rsequence at the beginning of the program?

 

Now does this mean that “random mutation coupled with a selection†cannot produce new information? Of course not! That is, provided the selection method is intelligent, you can certainly produce new information. For example, a Genetic Algorithm can be given a target or a range of solutions to look for, and run ad infinitum until something useful is detected. This requires intelligence to determine when that useful thing has occurred.

 

So nature itself cannot dictate what is useful then? I was under the impression that the ability to acquire resources and mating partners was considered by nature to be inherently useful.

 

 

Is your objection to the program the fact that Rfrequency is a pre-selected constant?

[Guest Paul C. Anagnostopoulos]

Here you unwittingly make my point for me. Thanks!

It wasn't unwittingly. I agree that the information has to be somewhere, in some form, in order for it to find its way into the genome. This is also true in nature. Don't you agree?

 

You can say all you want, you can say the moon is made of cheese. But that does not make it true. There is no evidence anywhere to support Schneider’s parameters, none. Zippo. Show me a study on truncation selection in nature (do a google search of this and my name and website comes up more than studies do!). Show me a study where the beneficial mutation rate is 50%.

Are you saying that the information does not find its way into the genome by repeated application of evaluation, selection, and mutation?

 

Ev does not simulate a beneficial mutation rate of 50%. It applies random mutations to each creature on each cycle, the number of mutations specified by a parameter (usually 1).

 

I modified Ev so it does not do truncation selection. Instead, it replicates the second-, third-, and fourth-best creatures, replacing the second-, third-, and fourth-worst ones. The best and worst creatures were left alone for good measure. Instead of taking about 800 generations to evolve a perfect creature, it took about 10,000.

 

You are trying to do an end run around my point. Please respond with evidence and/or an explanation of how his program produced a new code (symbols, syntax, and semantics). Just don’t say it is so, that is a waste of everyone’s time.

You mean don't make just-so statements like "There is no evidence anywhere to support Schneider's parameters"?

 

The code evolved in Ev is very simple, nowhere near as interesting as the genetic code. The symbols and syntax are wired in already. What evolves is the semantics: The code selects the binding sites and does not select any other sites.

 

~~ Paul

[Guest Paul C. Anagnostopoulos]

The interesting issue here is that of the information source. Evolution is a process of transferring information from the environment to the genome. The genome "learns" to survive in the environment by encoding information that complements the information in the environment, so the organism can survive in the environment.

 

There is never going to be a simulation, or a real organism, that "makes up information" out of whole cloth without that information having already been somewhere else. An exception is that a genome could evolve useful information by accident, but we all agree that accident does not explain the wealth of life around us.

 

~~ Paul

[Guest Paul C. Anagnostopoulos]

Is your objection to the program the fact that Rfrequency is a pre-selected constant?

The objection I hear most often is that the evolution of binding site recognition was built into the program. The program does not somehow evolve a genome that does "something useful" without any presumption about what that useful thing will be beforehand.

 

This is an interesting remark and is worth discussing. To make it have teeth, one would have to show that actual organisms evolve useful mechanisms without any influence from the environment concerning what those mechanisms should do. I can't conceive how one would demonstrate such a thing, except to show that it occasionally happens by chance.

 

Note that nothing stops us from changing Ev so that the number of binding sites is chosen randomly rather than prespecified.

 

~~ Paul

[Fred Williams 0]

It was brought to my attention that Dr Schnieder responded to my comments:

 

http://www.lecb.ncifcrf.gov/~toms/paper/ev/williams/

 

Specifically:

 

· definition of truncation selection:

A breeding technique in which individuals in whom quantitative expression of a phenotype is above or below a certain value (the truncation point) are selected as parents for the next generation.

The method used in the Ev program is position in the sort list, NOT the phenotype!

Tom is apparently unaware that sorting is also truncation selection! The definition he provided is not a complete definition. For example, from this evolutionist web site:

 

“In truncation selection individuals are sorted according to their fitness. Only the best individuals are selected for parents.” [link]

 

This is precisely what Tom’s program is doing! Tom, your program is using severe truncation selection, you simply cannot claim otherwise.

 

It should be noted that Paul’s adjustments still amounted to truncation selection, just not 100% truncation selection such as is implemented in Tom’s program. But it is still radical truncation selection because virtually all members are being ranked and truncated, except the worst. So it is more like 99% severe truncation! Note what happened after Paul’s adjustment:

 

“Instead of taking about 800 generations to evolve a perfect creature, it took about 10,000.”

 

It took 10x more generations, just after a minor modification where the truncation selection was still radically severe!

 

As far as Tom’s “gauntlet” I provided a reasonable starting point to remove truncation selection from his program in post 41. As I mentioned in this post, evolutionists must know such simulations are doomed to fail if they even approach realistic selection and mutation methods.

 

It is not my responsibility, or any of the other scientists, to write a new program or adjust Tom’s to prove his program is flawed. It is Tom’s program that is being critiqued, and it is up to him to address the errors. I consider this case closed until realistic selection and mutation are used in his program, plus an explanation of how his program can have a 100% guarantee of success and not have Rsequence~=Rfrequency already encoded within the program, albeit very inefficiently (see penultimate paragraph in this post)

 

This thread is hereby closed, except for any further comments by Dr Schnieder. I will email him that I have responded to his truncation blog rebuttal to this forum.

 

Fred