Mathematics, Information, Genetics, and Evolution

by Chris Cogan (Copyright 2005)

Feedback, discussion, comments, questions: Chris Cogan, ccogan@ou.edu

 

Abstract: Active genetic information is carried by the sequencing of the DNA nucleotides, so changing that sequence changes the information, whether creationists want it to or not, and the changed information is inherently new information: It wasn't there in the parent genes, and it is there in the offspring genes. The essay is mainly the development or the explaining of several points pertaining to this idea, such as how genetically useful new information can be obtained from informational noise by a sufficiently systematic filtering process.

Preface

This essay should really not be needed, but, judging from the number of creationists and ID-advocates who think Dembski's abuse of information theory is brilliant, and from recent gross misrepresentations of observational facts by creationists claiming that mutations are always "deleterious," I conclude that these people are woefully ignorant (or worse) in this area.

So, I decided to write up a summary of some ideas that I've been developing for years (since the early 1990's). It was a remark of Dawkins about genetic information (in one of his earlier books) that got me started on developing this general line of thinking in my own case. I would give some credit to Johnson (for his nonsensical claim that information is a kind of "stuff") and Dembski (for his gross misrepresentations of information theory generally, and its application to evolution in particular), but the fact is, were it not for the fact that both Johnson and Dembski actually get approving readership among the ignorati, I wouldn't bother, any more than I would bother trying to refute a flat-earther's claims. There are so many crackpot theories that one has to restrict one's extended consideration to those that are popular enough to be or contribute to a social problem (such as the prospect of removing science from public school science teaching). I have decided to take on the information-theory issues because I have had an ongoing interest in the subject since reading Shannon's book while I was still in my teens (over forty years ago). Despite my interest in information theory and my love of mathematics, I will not inflict any actual mathematics on you, as Dembski would (in his attempts at hiding the fundamental wrongness of his claims, his arguments, and his approach). I won't inflict the actual mathematics on you, and I have been rather more informal in my use of some of the ideas than would be ideal, but I hope the main lines of reasoning are clear enough (one thing I haven't even touched on at this general level is the definition of information, except insofar as it applies to genetics (a sequence of genetically active DNA, where it is the sequencing that carries the information, because a different sequence of the same length would carry different genetic information).

Introduction

Mathematics can be used to prove quite a wide array of things, if it is correct mathematics and if it is applied correctly.

However, while figures may not lie, liars may figure, as the old saying has it. This is relevant to the issue of whether (genetically) new information can get into the genotype of the offspring of another genotype during reproduction.

The really short answer is simply yes.

However, let's consider a slightly longer answer. Note before continuing that, for now, I will only be considering the simpler case of non-sexual reproduction.

Let me also note before going on that information is not some kind of mysterious "stuff" (as professional liar Phillip Johnson claimed it was). Information is simply the arrangement of parts of something, or the attributes of it, depending on context. An electron carries information in the form of charge, "spin," mass, etc. A sequence of ones and zeros carries information by virtue of that sequencing (not by virtue of some additional "stuff" other than the ones and zeros).

DNA carries/stores information as a sequence of nucleotides, and the sequencing is the important part, genetically. If the nucleotides are in one order, the information content is different from the information content (in non-genetic terms) of any other order. For example, deleting ATA from the middle of GATATACCA produces GATCCA.

We can relax this slightly (for the benefit of the creationist/ID crowd) by saying that two genetic sequences are informationally different only if they are functionally different in their phenotypical expression. That is, we can restrict our consideration to genetically active or effective information, and refer to it as genetic information, or even just information (as long as context makes it clear that we are using the term in this sense). This restricts us specifically to genetically functional information, which is what we are really interested in most of the time. This way, we can temporarily set aside questions about "junk" DNA. However, I will note again before going on that, in a technical sense, any change is still an informational change, even if the information is ignored by the reproductive or other processes. The pages of an encyclopedia that you never read still contain information, even if you don't read them; they are just not informative to you.

What a Mutation Is

What is a mutation? It is a change in the DNA sequencing of nucleotides in a genotype from the parent genotype. That is, a mutation is a change in information, producing new information. That information wasn't there in the parent genes (that is, that sequence, in context, was not present in the parent genes, and it is present in the offspring genes: It wasn't there, and now it is there.

This applies to duplication (unless it is immediately "junk" DNA) and to rearrangement (if it has an effect on what happens) and rearrangements, and insertions, as long as the result is genetically significant to the offspring. It is also true of the actions of HOX and other homeobox genes. If one of these genes is modified so that its effect on other genes is modified, this is new information: It wasn't there, and now it is there.

Finally, this even applies to deletions (removing the genes for pigmentation is equivalent to the new instruction to produce an albino, and removing just a part of a gene may produced a functionally very different new gene).

If you don't think deletions can change information, consider two sentences. The second is made from the first by deleting four letters:

I sent the catalog to Bob

I sent the cat to Bob.

If you don't think the second sentence is new information, imagine just reading the first sentence, and ask yourself: Does this tell you that I sent the cat to Bob? No, from reading the first sentence, even if you actually did think of the question of whether I sent the cat (as well as the catalog) to Bob, you wouldn't be informed about it from reading the first sentence.

But the second sentence does tell you this. It's new information: It wasn't there, and now it is there.

How do we know? Well, how do we know that genes carry information at all? One way is by seeing the effects. Until recently, this was the only way we had. If an organism gives birth to offspring that are phenotypically different from itself (still assuming non-sexual reproduction), and if that difference is then passed on to the offspring of the offspring, then the genetic change that produce that phenotypical difference is new information: It wasn't there in the parent, and it is there in the offspring.

Further, we can look at the genes themselves, and by experiment, observation, and testing, determine that there is genetic information in the offspring that is not present in the parent's genotype.

What is a mutation in informational terms? A mutation simply is the introduction of new genetic information, and we know this because of its effects on the organism and its offspring.

That's precisely the problem with mutations. If they were not new information, mutations would have no phenotypical effects at all, and there would never be any harm in them. But the fact is that it is precisely because of their informational content that they become problematic in a large percentage of cases. The effects on the offspring can only occur because mutations are new information: It wasn't there in the parent, and it is there in the offspring.

Creationists who admit that mutations affect the offspring but deny that there is any new genetic information in them contradict themselves. When they admit that phenotypically effective mutations do occur, the admit that there is in fact new genetic information, but when they then deny that there is new genetic information, they are denying, by implication, that genetically effective mutations occur.

Now, let's consider genetic recombination. In order to avoid having to insert qualifiers too frequently, I'm assuming that we are talking only about active genetic material, not "junk" DNA.

Recombination nearly always produces new information except in extremely bizarre cases where the genes of the offspring are an exact functional duplicate of the genes of one of the parents, and in those cases where what is different as a result of recombination is truly "junk" DNA. All other recombinations that have an effect on the carrier's phenotype produce new information. ALL, without exception, since that simply is what new information is, in physical terms. That is, if two people give birth to child whose effective or active genes do not exactly duplicate all of the genes of just one of the parents, there is new information: That combination of "words" (or whatever) wasn't in either of the parents, and it is there in the offspring.

We know it's new, even in functional terms, because it produces offspring that are phenotypically different in ways not due to environmental effects on development. For example, a child may be said to have its mother's eyes and its father's hair, producing a combination that is a new combination (it wasn't in either of the parents, but it is there in the offspring). The same genetic considerations apply to all (or virtually all) sexually reproducing species (I think it even applies to Cheetahs and other animals that have gone through a genetic "bottleneck" that reduced their genetic variation to nearly zero (all alleles the same except for gender-specific ones). Why do I say this? Because even Cheetahs are not perfectly genetically identical (as far as I know). Also, some lab mice are specially bred to be as nearly identical as possible, so they might count as a true exception (but, of course, even their genes are subject to mutation).

Are Mutations Ever Beneficial?

Is the new information from recombinations and mutations ever beneficial?

The short answer, again, is yes.

(Note that, even though there is no special objective reason, at this level, to separate out mutations from recombination, I do so for the moment to humor creationists, since they claim there is some fundamental difference between them.)

Recombinations obviously occasionally produce beneficial results: The offspring of people fairly frequently have traits that are clearly beneficial, clearly genetically based, and that are the result of combining genes from each of their parents.

Is this sufficient for evolution? Only up to a point, because recombination (as normally conceived) does not include genetic changes that encompass less than a gene (such as a single base-pair substitution from one parent into a gene that is otherwise entirely from the other parent; I think this would always be considered a mutation rather than a recombination), and most especially after complex genomes are established, when lots of genes may contribute to more or less "single" phenotypical differences.

Mutations are defined, for the moment, as any change in the genes other than those that are purely recombinations. For example, if the genotype of the offspring is longer or shorter than that of its parents, there has been a mutation of some sort. As we know, most mutations are either junk or harmful, so the question is whether there are any that are beneficial to the organism.

Yes there are. One example: Experiments with yeast in labs has shown that yeast living in a sugar-scares environment can develop additional genes for more-efficient sugar-scavenging, and we know that bacteria introduce new genetic information that's beneficial (to them, not us) as a response to antibiotics. Another, and more commonly noted, example: Bacterial adaptation to antibiotics.

It is sometimes irrelevantly claimed that these changes always result in some other effect that is negative, such as less efficient protein construction. First, I don't think this was true in the case of the additional gene in yeast, but, in any case, it's irrelevant because the definition of "beneficial to the organism" has nothing inherently to do with efficiency in the production of protein (or efficiency in any other process, for that matter). The claim that it does is creationist misrepresentation of both biology and the theory of evolution. What counts as beneficial in evolutionary terms is ongoing reproductive success, not any particular level of protein-making efficiency. Even the fact that these bacteria do not do well in competition with non-resistant bacteria when they are put into an environment without antibiotics is irrelevant because that's not the environment the adaptations were for. Humans would not do well in competition with birds as flying organisms, but that doesn't mean we are not well adapted to the environment we live in and to the resources we have access to.

The creationist argument in this case is like saying that the adaptation of most birds to flight is deleterious because it means they can't live well as burrowing animals. Whether or not an adaptation is deleterious or not depends on the environment that the organism is living in, not the environment their ancestors lived in.

The dishonesty of the creationist's claim in this respect is shown in the way they characterize such changes as "deterioration" even though it clearly benefits the bacteria (reproductively) to have this "deterioration," as long as they remain in their environment. Creationists further characterize the non-antibiotic environment as the "normal" one, even though the fact that the bacteria had to evolve to deal with the antibiotics means that that environment is normal for them.

The implication of the creationist claims is that there is some sort of Platonic Idea of what is "normal" that is to be applied regardless of the changes in the environment and in the organisms themselves. For them, "normal" is simply arbitrarily taken to be: Whatever the environment was like before the evolutionary process occurred. This is arbitrary, like saying that normal for humans is living without antibiotics merely because there were (presumably) no antibiotics in the Garden of Eden (pretending for the moment that Eden ever existed). At least as far as biology and science generally are concerned, normal is determined by facts, not by some floating abstractions about what constitutes normal or about what normal was at some irrelevant past time.

Further, if there are negative side effects of beneficial mutations (and there are negative side effects of many genes anyway, possibly even of the "normal" genes that code for efficient protein production in bacteria), further evolution may remove at least some of them. Evolution works in particular directions by a process of refinements as well as by introducing new alleles. Thus, a bacterium that has adapted to become resistant to antibiotics may later also adapt to regain the efficiency of protein production that was lost in the first round of mutations. Obviously, this is not always possible, because the side effects may be inherent in the modifications (if, for example, it was somehow the very inefficiency of protein production that was responsible for the resistance to antibiotics), but, it is typically the case that an organism that has adapted to some environmental change can also recover any needed functionality that was lost in the initial adaptation.

This is because further mutations will occur, and those that tend to refine the organism's functioning so as to retain the new features while recovering the missing ones will be favored. The proto-giraffe that evolves a somewhat longer neck may pick up genes for a bigger heart from another proto-giraffe and thus get rid of one of the main side effects of the longer neck (greater strain on the heart, and possibly insufficient blood flow to the head, etc.). A further mutation that increases the strength of the blood vessels going to and from the head may overcome any tendency to bursting caused by the increased blood pressure.

[Note: Significantly, despite the claims of supporters of "Haldane's Dilemma," if there is a population of many organisms, they need not establish each beneficial mutation before the next one comes, because one member of the population can have a longer neck while another member has a larger heart and yet another has sturdier blood vessels. Because they recombine genes, there is a chance that the ones with the longer necks will breed with the ones that have the larger hearts, and so on. Since these combinations will typically be beneficial (in the case where each trait separately is beneficial), they can easily be favored by reproductive success for transmission to future generations, thus eventually becoming "fixated" in the population.]

But the basic fact is that not all mutations have negative side effects, especially if the main function of a mutation is to correct a side effect of an earlier mutation. Such additional mutations will typically be smaller than the initial one because they only need to introduce enough of a modification to overcome or partially overcome the side effect. Thus, if resistance to antibiotics brings reduced protein-making efficiency, it may well be the case that this protein-making efficiency can be recovered by a mutation (or a series of mutations) smaller than the one that produces the antibiotic resistance. This will not always be the case, of course, but, unless the reduced protein-making efficiency and the resistance to antibiotics are exactly the same at some physiological level (if, for example, the antibiotics are tuned to efficient protein-making as such and somehow take advantage of it to kill the bacterium -- by making it overproduce some proteins, I suppose), then there is no reason that further mutations can't fix the problem, given enough time. Since we tend to switch antibiotics when one becomes less useful due to resistance, we rarely keep a consistent evolutionary "pressure" on the bacteria to evolve both resistance and efficient protein making.

But, suppose we continued using an antibiotic at the same rate even after bacteria were almost completely immune to its effects. Then what? I predict that, if there is any significant advantage to the more-efficient protein-making processes, there will be evolution to recover this efficiency despite the side effects of the antibiotic resistance (again, unless the protein-making inefficiency and the resistance are really the same thing physically and genetically).

It would be possible to do such an experiment with animals and a selected bacteria species that has already developed resistance. Basically we would keep the animals frequently dosed with this antibiotic so that it became a constant in the environment of the bacteria, and thus providing plenty of opportunity for beneficial mutations to occur to that would eliminate or reduce any negative side effects of the antibiotic resistance. Would anyone like to bet that this will never happen (I can always use a few extra bucks)?

These mutations can include mutations to the same genes that introduced the benefit to begin with. Suppose that there is something about the antibiotic resistance gene that causes the reduced protein-making efficiency, but that that aspect of the gene is not really essential to its antibiotic resistance, because it is merely a holdover from some previous function of the gene from which this new gene evolved (by duplication, perhaps). Now, suppose that this antibiotic resistance gene itself is modified over time so that it retains the antibiotic resistance effect but simply loses its negative effect on protein making.

Although I say "suppose" a lot in the preceding paragraph, it's because I'm only giving a hypothetical example. Such modifications upon modifications do in fact occur, both in nature and in the lab. This is an observational fact, not a theoretical one.

Because all of this talk about mutations modifying the effects of previous mutations is actually information-theory talk in disguise. Every mutation that has any effect at all introduces new information, so there is no reason to think that beneficial mutations will just magically stop altogether if one has already occurred. If a beneficial mutation for antibiotic resistance can occur, then so can one for improved protein-making efficiency, and one for any side-effects of that mutation, and so on, to the limits of their benefit to the bacterium's reproductive success.

The claim that beneficial mutations are always deleterious in some other way is simply false, except by reference to irrelevant ideas of what is beneficial and what is not. In objective terms, a mutation that removes a side effect may be purely beneficial, because all it may do is remove the code that is producing the negative side effect.

The claim that mutations can only have negative side effects, or worse, that they are always, deleterious is simply delusional wishful thinking on the part of creationists. It's wishful for obvious reasons, and delusional because, in order to maintain this bizarre anti-science in their own minds, they have to deny or corrupt all of the key concepts and facts.

-- Along with mathematics. Many mutations are either duplications of existing genetic material or modifications of existing genetic material. For sufficiently small mutations, it is mathematically absurd to claim that none of them can be beneficial overall to the organism, even if we grant the silly premise that the environment before antibiotics (for example) is the "normal" to be used as the standard (etc.). It's like claiming that no one can ever win an (honest) lottery. If there is any physical possibility of a win occurring at all, then, given enough lottery cycles, someone will win. It's also like saying that no random changes to a book can ever make for an improved book, even if all it does is add a needed word or letter, or delete a typo.

The only way this creationist claim could be true would be if there were simply no genetic change that can be beneficial in principle even if we tried to construct it nucleotide by nucleotide, a claim that is simply empirically false in the real world. Creationists would very much like it to be true that mutations are never beneficial information, but they are simply factually wrong. They would like it to be true because it fits their theory, not the facts of the real world.

This is why they have to use incorrect definitions of terms like "normal" and "beneficial." Without this "stopper" claim, the door is open for evolution to continue making small beneficial changes, until, viola! (and voila!, too, of course), the ever-dreaded monster of "macroevolution" would occur, and their case would be exposed as the pitiful excuse for a theory that it has always been (since long before Wallace and Darwin). If they didn't make up their own biologically nonsensical definitions of such concepts as "normal" and "beneficial," they would lose an entire major category of unsound arguments.

What makes it seem plausible to them at all, aside from their desire to have it be so for propaganda purposes (or is it plausible to them at all?)?

I think it's this: Since most mutations are not beneficial but are either neutral or harmful, it's almost (in their minds) the same as all mutations being either neutral or harmful, as if we could simply ignore anything that doesn't occur with an extremely high relative frequency (relative to the population).

But, this is precisely what makes their math such bad math: We can't do this in biology, diseases, fire control, and computer-virus control. If ninety-nine out of a hundred bonks on the head are quite definitely harmful (or, at best, neutral), we can pretty much ignore the one bonk that might be beneficial (because we don't want to pay for it with, on average, 99 harmful bonks). But, if we are trying to prevent fires, the fact that most campers in the forest don't start forest fires is, at any particular time, unimportant if one camper does start a forest fire, no matter how small, because that initially tiny blaze may grow and burn down the entire forest. Similarly, if you eat food that is not in good condition but which has no bacteria dangerous to you in it, you may do just fine. But, if you eat food that has even just a few deadly (to you) bacteria in it (even if the food is otherwise in perfect condition), they may kill you, even though there are just a few of them (initially).

Why? Because, like forest fires, bacteria reproduce. Fires "reproduce" by spreading to new fuel. Bacteria do it by cell-division and growth of the resulting cells.

In reproductive biology, even one mutation that is genuinely beneficial cannot rationally be ignored, because the fact that it is beneficial gives it a better chance (on average) than other mutations or even existing gene alleles at the same locus. Given time, it can spread to become universal in the population of the species, even if it started out as just an insignificant statistical "fluke" in the eyes of creationists.

Why can this happen? Why does it happen? Because such mutations are both beneficial to the organism that has them and they are informational. Genetic information differs from the configuration of, say, a long nylon polymer, because it can be (generally) copied, thus passing the benefit on to further organisms. Once the initial beneficial change has been "paid for" (by all the other mutations that have to be accepted in order to find the one beneficial one), the further use of that benefit is (at least in this respect) "free" (that is, Dembski's "lunch" has been paid for).

Incidentally, if you doubt any of this, I suggest you write a computer program to emulate the (actual) principles of evolution on a population and give it a try. After setting up everything when the program begins execution, have a beneficial gene introduced into just one member of the population, but make it such that this single lone beneficial gene provides a fitness/reproductive advantage to that single lone "organism," so that it's own chances of dying before reproducing plentifully are lower than the chances of other otherwise similar organisms for the same reproductive success. In "fluke" cases, the gene will get wiped out (just by dumb bad luck). But, in other cases, and over the long haul, such beneficial genes will often spread through the population, as the members of the species without this trait are gradually edged out of the reproductive race. At first, the effect on the larger population will be small simply because the number of individuals with the new trait is small. But, as they become progressively more common by virtue of their different attrition rates, the effect will become more noticeable and will spread faster until "complete" fixation of the gene occurs.

It will spread through the population faster as time goes on because there will be, for those organisms that don't already have the trait, fewer opportunities to breed with others that also don't already have it. Suppose, for simplicity's sake, that there is an initial population of a hundred organisms of species S. Suppose just one of them currently has the beneficial mutation, but that it is not recessive and so has a good chance of being expressed in this organism's offspring. And suppose it is in fact expressed in the offspring. And suppose it provides a significant reproductive fitness advantage as compared to the rest of the population.

At this point, only a few members of the population have it, and so, most of the time, any randomly-selected organism that doesn't have this trait will be one that will breed with another organism that also does not have this trait (simply because of the proportion of those that don't have it to those that do). However, because this also applies to those few that do have this trait, they will usually be helping to spread the trait. When their percentage gets to be higher, the number of organisms that don't have this trait will be lower (assuming a constant total population), and so now the chances that an organism that doesn't have this trait will breed with one that does have it is significantly higher. When the new trait has spread throughout half the population, there will be about a fifty percent chance than the breeding partner for an organism that doesn't have the trait will be one that does have the trait.

This is a big increase from the days when the trait was only possessed by one lone organism. Thus, the rate of spread will increase even more now, and even more in the next generation, until, very rapidly, the new gene has become almost completely universal to the population.

Put another way, the rate at which a beneficial trait spreads is partly a function of the percentage of the population that already has it, so the rate itself changes over time, increasing until it reaches the point where the only members of the species that don't have it are ones that have a mutation at that locus, or until a point is reached where there is diminishing returns because of an increase in the value of alternatives due to the reduction in their frequency (if you have a gene that inclines you to eat a certain food, the amount of that food available to you may be greater if other members of your species do not have an inclination to eat that food, even if their disinclination is due to a gene that is otherwise beneficial up to its current frequency).

If you write the program suggested above and use reasonable assumptions, such as that the trait is not one that directly increases mating attractiveness (then, the effect is even stronger), you can prove it for yourself. You don't even have to emulate actual evolution by actually having software or data structures to represent individual organisms. You only need to represent the mathematics correctly for each generation. This will be better, in some ways (if you do it right) because it will give you information about averages directly instead of requiring that you monitor an actual evolutionary process and keep statistics on it, and do it many times to ensure that the fluky cases are averaged into the whole.

What is happening in the evolutionary scenario described above (and which occurs with many variations in the real world)? Genetically new information enters the genotype and then spreads through a population via information duplication.

Evolution Is the Evolution of Information

In fact, evolution is the evolution of information, and it is information which is, in a metaphorical sense, information "about" how genetic information can survive in its environment (including the physical medium of DNA). If this information is sufficiently "correct," the information does survive.

If it is not, it is filtered out. Natural selection is an information filtering process. It takes as input large number of "messages" represented in a phenotypical form, many of which have some "noise" in them, and filters out those "messages" that don't meet the survival requirements of information in that environment. The organisms are "selected" (or not) directly by the environment, but the information is already present in the genes, so selecting indirectly selects genetic information as well. Because of the correlation of phenotypical traits with genetic information, it is easy to assume that selection is the selection of phenotypical traits, but the core of selection is the selection of information.

Noise can become useful information in such a context because that's the way filtering works on random modifications to a message. It becomes mathematically almost certain that some instances of such "noise" will produce useful information if there are a lot of trials, a lot of messages that may be modified each with its own bit of noise. This is standard Shannon information theory combined with the fact of many repetitions of the "messages." For each individual message, the probability is that noise will not be beneficial to it. But, mathematically, it becomes all but absolutely certain, over a large enough number of repetitions and sufficiently nearly random modifications ("noise"), that there will be some that are beneficial.

To illustrate, consider the "message": MTAXYBALLY, and suppose this message is in competition with many others of a similar sort, such that those that have certain combinations of consonants and vowels will tend to be favored, while the rest will tend to be removed from further repetitions. Suppose that the environment consists of a set of statistical rules that happen to be the same as those of English with regard to combinations of letters, and that these rules are applied blindly, with no intelligence at all, like laws of physics. Will this system be able to evolve "MTAXYBALLY" into the word "MATHEMATICALLY"? Yes, even though, nowhere in the entire system of rules are there any words (except insofar as they are inherent in the statistics of letter combinations, as, for example, the word "ion," which is included only because it is a frequently used three-letter combination at the ends of words (as it is in the word "combination," for example)). Why can a word like this evolve in such a clearly "dumb" system (where the rules are mechanically, unintelligently, applied, and where the rules themselves are mere probabilities of letter combinations)? It can happen because of the large number of repetitions, and because each "success" establishes a new "base-line" from which future changes start. We never have to try to randomly select all the letters of the final long word all at once, but only by a gradual, evolutionary process of mechanical trial and error.

[Note, please, that this system does not work because we have somehow "implanted" our own intelligence into the system (besides, how many creationsts/ID-ers would settle for a designer no smarter than the average computer programmer?!). All we have put into it are probability values for different letter combinations, which involve no knowledge of English as a language, but only as strings of letters. We have to have some intelligence to write such a program, but that's different from having the program itself be intelligent (as many of my own programs prove). If we were to try to produce evolution by emulating a physical world, it would take intelligence to create the emulator, but the emulated physical events themselves would be no more intelligent than the falling of a rock in the real world.]

Because the signal to noise ratio is high enough, genetic modifications don't normally destroy the "message" entirely. Most copies of the information are fairly faithful to the "original(s)" (but only in a piecemeal way in the case of diploid reproduction). But there are enough modifications that are make the signal "better" for the local filtering process that there can be gradual change to a species (or a subpopulation of a species) over time.

Summary

All genetically active mutations are new information. Information is obtained from "noise" by filtering, filtering in such a way as to keep changes that constitute improvements.

The direct application of standard Shannon information theory to evolution shows that, given almost any standard for what constitutes improvement (as long as it is not some bizarre step function, etc.), and given a sufficiently nearly randomized introduction of changes in the signal, some of the changes have to be improvements (by that standard).

It is mathematically improbable in the extreme that no such changes will ever be beneficial. The environment is an information filter that works by filtering the organisms themselves and thereby indirectly filtering their genes, and thereby filtering genetic information, leaving (mostly) only such bits of "noise" that turn out to be evolutionarily better information (for survival in that very same information-filtering environment)

Arbitrary Platonic ideas of what is normal and what is beneficial are simply not relevant to whether evolution is true or not in the real world. What is beneficial evolutionarily is whatever promotes reproductive success.

Feedback, discussion, comments, questions: Chris Cogan, ccogan@ou.edu