Random Degradation and Natural Deselection
| January 18, 2007 | Posted by Dave S. under Biology, Evolution, Science |
Truth in advertising would have random mutation and natural selection renamed to reflect what’s really happening.
Orthodox evolutionists tell us that mutations are random when plotted against fitness but this is not the case. Mutations are random when plotted against position in the genome. An error can occur anywhere and it’s pretty much an unpredictable location at this point. But the effect of errors are certainly not random. These errors in duplication are occuring in highly optimized systems and as anyone with common sense and a bit of insight into the behavior of complex machines knows, random failures seldom if ever result in improved performance. Random mutations are in fact random degradations. I have yet to see compelling evidence that any so-called random mutation resulted in improved performance except perhaps a few trivial and arguable examples of an organism escaping the effect of a toxin by minor change in an enzyme which may very well be a directed change. The vast majority of random mutations, if not all random mutations, are neutral at best and immediately fatal at worst.
Since natural selection must by definition act on changes in fitness (differential reproduction), and those changes in fitness caused by random mutation are always or almost always detrimental, natural selection is really natural deselection as it works in almost all cases to deselect organisms whose genome has wandered from its optimized incarnation through random degradation.
Orthodox evolution theory’s weakness isn’t survival of the fittest. That’s readily explainable by the theory of random degradation and natural deselection. The weakness is arrival of the fittest. If there were a law requiring truth in advertising for orthodox evolution theory the problem of arrival of the fittest would be more easily understood.
36 Responses to Random Degradation and Natural Deselection
Leave a Reply
You must be logged in to post a comment.
I’ll quickly list several implementations I can think of:
1. “methinks like a weasel” approach
2. Your approach.
3. Fitness of the word is determined by length and whether any letter out of 10 matches ANY word. (Example: “nxxcxxxxxsx” is selected for matching “narcissism”). Letters selected as matching a word can still be randomized in the next iteration, so a match can be made with a different word.
4. It’s been a while since I looked at your program but I’m assuming you’re using ASCII? Fitness of the word is determined by length and the absence of any non-alphabetical characters. As in, fitness is 0 if the generated string contains any character outside of the 52 English alphabet characters (26*2 considering upper and lower case).
As a filter, they’re listed from “narrow” to “broad”. I’m sure I could think of more. Point is, 1 and 2 are more likely to produce results. 3 and 4 still use Darwinian mechanisms but aren’t designed as well as the top two.
Now if you could think of a approach that could hit upon 500 informational bits then you’d have a challenge to ID. Especially if that approach is relevant to biology since generating 500 informational bits with no major DIRECT intermediates is the issue at hand.
3. Fitness of the word is determined by length and whether any letter out of 10 matches ANY word. (Example: “nxxcxxxxxsx†is selected for matching “narcissismâ€Â). Letters selected as matching a word can still be randomized in the next iteration.
This is certainly do-able; I imagine it will result in longer strings being found, since there are significantly more “livable” strings under this condition.
4. It’s been a while since I looked at your program but I’m assuming you’re using ASCII?
The code as written actually does not modify the binary representation of characters, but only the characters themselves; so any letter can mutate to any other letter. This means that instead of 2^8 bits of information, each character only contains 26 (~2^5) bits of information. This is represented accurately in the analysis I posted (the brute search is over the same space).
I’ll let you know if I do the analysis of #3.
Now if you could think of a approach that could hit upon 500 informational bits then you’d have a challenge to ID.
Other researchers have developed GA’s that can find long strings of multiple words forming sentences. Would these fit your description?
Oops. I just realized I wrote that one letter is 26 bits of information – of course, 26 ~ 2^5 is 5 bits of information (4.7 to be technical). I meant to say it is equivalent to searching over base-26 representations.
I repeat:
Depending on how fitness is defined/designed in an indirect pathway I think it could be done (as in, I highly doubt a direct pathway will work). But to put that into perspective:
So essentially the design of the program would (1) need to produce a positive result and (2) incorporate modern Darwinism (feel free to incorporate other proposed mechanisms besides RM+NS) and (3) need to be relevant to biological reality. #3 has proven to be the tough one.
(3) need to be relevant to biological reality. #3 has proven to be the tough one.
I agree 100% that the simulation is not an accurate model of biological reality. I only meant the analysis to show that the underlying idea of random mutation combined with a selection critera generates interesting forms faster than blind search.
If I ever write a GA to model biological reality, I’ll let you know
Sometimes people think that because I’m an ID proponent that I believe Darwinian mechanisms are completely useless. That’s not so. In the right hands I believe GA’s to be a powerful tool if designed correctly around the intended goal.
I’m also something of an ID heretic since I consider it may be possible for “emergent complexity” to occur under certain rare cases in an already complex system that’s designed in certain ways. This hasn’t been observed yet but ID may have to be modified to incorporate such scenarios just like Darwinism has been modified so much over the years.