|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
Roger R
Member
Member # 200
|
posted 03. June 2003 21:36
You mean the GP who said:
quote:
I have stated my desire to hear substantial responses (not one-liners or substituted with attacks on evolutionary theory).
That GP? The GP who wanted me to take my on-topic discussion to another thread so he could continue his discussion of issues from another thread?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 01:28
This is a more extended response to John Bracht's post on the Brainstorms thread concerned with Joshua Smart's essay. I posted a partial response there, and elaborate here.
Bracht wrote in that posting quote:
Here's the problem with the Lenski et al simulation: they don't appropriately model biological evolution and hence their results have no applicability to reality. In particular, the fact that their program's "mutations" consist of swapping out entire functions, or adding/deleting them. There is no analogy in biology for the miracle appearance of fully formed, functional blocks. The closest thing I can think of is a protein that gets swapped out, but in REAL evolution those proteins have to first evolve their novel functionality. Furthermore, since real evolution occurs by mutations at the nucleotide level (point mutations, insertions/deletions etc) the mutations have a high probability of disabling the functional proteins they occur in. In the Lenski simulation, the evolving programs cannot accrue mutations at the function level or below (hence, no non-functional functions can ever evolve). This basically constrains the evolutionary process to the functional areas of sequence space--solving the problem of crossing non-functional canyons, which is the key problem for evolving ICness in the first place! (Emphases added)
The emphases in that paragraph highlight the misconceptions Bracht has about the Avida simulation. They embody Bracht's fundamental misunderstanding of the mechanics of the Avida simulation. The simulation's mutations did not consist of "swapping out entire functions ...". There was no "miracle appearance of fully formed, functional blocks." Enormous non-functional sequence space was available and accessible to the simulation, and non-functional sequences of instructions could occur and in fact, according to the knockout analyses of individual evolved programs, did occur.
Consider the language of the original paper: quote:
[At the beginning of a run] All organisms were identical and obtained equal energy to execute their genomic programs, including the copy commands by which a genome replicates itself one instruction at a time. Copying is subject to errors, including point mutations, insertions and deletions. Each mutation alters the genome and may change an organism's phenotype, including its replication efficiency, computational metabolism and robustness. Thus, genotypes vary in their expected reproductive success.
The units referred to are clearly individual primitive instructions. I may read that more easily because I have also read the Avida manual and played with the program, but if Bracht would consult the Avida manual, pages 29-30, he'd see that all of the kinds of mutations employed in the reported research are at the level of individual primitive instructions, not blocks of instructions. It is simply false to assert that mutations consisted of swapping around functional blocks of code. Period.
It is true that larger changes in genomes and phenotypic functioning can occur as a function of instruction-level mutations, but that's not at all biologically implausible. For example, a single-instruction copying mutation during replication could produce duplication of a block of code within a critter's genome, an analogue of gene duplication. There are other such "implicit" changes possible, all secondary to the effect of instruction-level mutations on replication code, on the code governing phenotypic functioning, or even on the "junk" (apparently non-functional) code that appears in the critters genomes.
There's an old saying in system administration: RTFM! I commend it to John's attention.
RBH
Added in edit: Here is the Line of Descent of the case study population, all 345 genotypes (344 mutational steps) in the lineage that evolved a program capable of performing EQU. Notice that every mutation is a single-instruction change. No blocks of code. And here is the case study lineage's knockout analysis and sequence of instructions. Notice that there are 15 non-functional instructions by the knockout criterion, 25% of the genome. And Run 129 has 275 "junk" instructions by the knockout criterion, fully 77% of its genome. Talk about wandering in non-functional sequence space!
[ 06. June 2003, 02:05: Message edited by: RBH ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 06. June 2003 02:29
RBH,
You completely misunderstand me. Perhaps I wasn't clear.
I wasn't arguing for blocks of instructions being swapped, but merely pointing out that each instruction is a complex function itself which composes many lines of lower-level code (in avida I don't know how that encoding is done, but ultimately it will boil down to some sort of machine language code).
So you've set up a strawman and knocked it down, great. But you haven't addressed my argument. You also misunderstand me by thinking I said that nonfunctional organisms couldn't be produced--but if you'll read carefully, you'll see that I said (in the part you quoted, and even highlighted!!!)
quote:
In the Lenski simulation, the evolving programs cannot accrue mutations at the function level or below (hence, no non-functional functions can ever evolve).
Notice I NEVER said non-functional programs couldn't exist. I said that evolution cannot mess up the functions (=individual instructions), since these are the base units that get swapped out or changed in "mutations". My point, once again, is that this is totally irrelevant to real biology. If proteins equate with the avida functions (instructions), then it's clear where the disanalogy is. Certainly, a point mutation never switches out complex instructions/functions like "compare two registers to see if they have the same value" or "replicate" (two real instructions from the instruction list). Can a single amino acid encode "NAND"? This is my point: in biology, you have to build up the complex functions piece by piece, while the Lenski simulation starts with complex building blocks (instructions) and just moves them around.
You want to point out that it requires two instructions to carry out NAND: the NAND instruction and the I/O instruction. This makes absolutely NO DIFFERENCE to my point. It's trivial to arrange two blocks of code (the instructions) which themselves perform complex, high-level functions, such that they produce the output. And I'm guessing that in a complex EQU-performing program, the I/O isn't required till the very end to output the final result--not after every NAND. Hence, the I/O requirement doesn't really make it much harder to evolve EQU. (As an aside, it would be interesting to see if the NAND function were broken into two sub-functions, which had to be put together by the program to get a NAND, and see if EQU ever evolves).
Perhaps the confusion was caused by my use of the word "function" when I meant "instruction". If so, my apologies. I hope this post clears this up and we can come to a genuine mutual understanding on these issues. One thing that would help: please refrain from the condescending tone and common "this reveals John's fundamental misunderstanding of...." which does little to further the discussion and much to just annoy those you are trying to dialogue with. If you think I'm mistaken, please just point it out in a professional manner without the rhetoric--I'm happy to be corrected where I'm genuinely wrong. Also--please try to take a charitable reading of my writing when you can. If you can twist it into a strawman, but that strawman doesn't make sense (like the above), you might reconsider whether you've properly interpreted my comments. I have spent a fair bit of time studying the Lenski simulation (including reading the supplemental information from Nature) and I do have a basic understanding of the programming. So while I don't have the subtleties of Avida programming knowledge you have, I'm probably not going to make the sort of gross errors you seem to want to attribute to me.
Thanks.
John
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 06. June 2003 08:17
RBH, I think you fail to understand the crux of John's point. He's merely trying to get at the fact that each primitive instruction is a well functioning operation. They, in no way, represent genetic sequence. Notice that I am not arguing against a bad biological analogue, but rather against the relevancy of the program to IC. Why? Because the primitive instructions and the Avida architecture are already too well formed to the problem at hand.
Look. The ID critics have been consistent in two things: 1. Proclaiming the amazing powers and success of the Lenski simulation and 2. Insisting that it is up to the ID people to show why the simulation is irrelevant or less than relevant to IC.
I feel differently about the situation. Since we can all agree that this simulation is FAR away from any biologically relevant system (each element in the genetic code performs a basic function itself, is the organism the complex function or is some part of the organism the complex function?, what are the well-matched parts of the IC system?, is a logic function even similar to a mechanical function?, etc.) shouldn't it be the responsibility of those who want to draw elaborate conclusions from the paper to tell us exactly why the system is relevant?
The important things for me are these:
1. Poor example of a genetic code (it would be better if alphabetic sequences coded for individual primitive functions rather than a direct mapping from letter to function)
2. What's complex, the organism or some organismal subsystem?
3. What's irreducible, the organism or some organismal subsystem?
4. What happens if we remove "p" from the genetic code?
5. I'd love to see an Avida program evolve the ability to do a simple mathematical multiply function, generalized to handle any two 32 bit input. This is not conducive to the Avida environment, whereas the EQU function certainly is.
6. There a number of 2 input functions which the system cannot come upon, not to mention 3 input functions, etc. Notice that the Avida system's architecture is limited in this way (look at registers and stacks)
http://scitec.uwichill.edu.bb/cmp/online/P10F/logic%20circuits.htm
7. The system is extremely constrained, front-loaded towards the target, and has a ridiculously high success rate >50%
8. Outside of aesthetic appeals, I don't see the strength of the project, and those who are gung ho (sp?) about it haven't given me good reason to abandon my skepticism
As I've said several times, the only instructive things I see coming out of the Lenski project are as follows:
1. The context for applying IC might be signficiant: mechanical vs. logical
2. Computer simulations testing IC should simulate the spatial and mechanical nature (or at least the spatial aspect) of IC
3. The notion of function in IC needs to be clarified and specified
4. The notion of complexity in IC could use some clarification: some sort of minimal complexity measure
5. It might be beneficial to generalize IC beyond the 3D mechanical context in which it has been formulated
[ 06. June 2003, 08:53: Message edited by: Micah Sparacio ]
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 06. June 2003 10:06
Because of the continuing interests in this post, I would like to summarize the main ideas generated here, from my perspective:
1) Largely due to my ignorance of current ID literature, I have described in the discussion a possible ID theory that can accommodate an evolutionary theory. In fact, this view may reflect certain mathematical symmetry between ID and evolutionary theory. It also resonates the language of equilibrium, due to Prof. Gould’s theory. Without a proper notation and avoiding confusion, we can denote it as “ID+E”, where E refers to “evolutionary theory” or “equilibrium” depending on your preference.
2)There is no doubt in my mind there are empirical contents and supports for “ID+E”, especially if you look hard enough. After all, many of the supports for “E” can be reinterpreted to support “ID+E”. There is only a short history of research from this perspective, and there is no reason to believe its research significance potential is absent.
3)The recurrent definition of IC/ID I described in the discussion is a general and a valid one. It can be easily “falsifiable” using the standard criterion of non-convergence to an acceptable error level. Its specificity will then be very clear when it is established. However, it does not preclude existence of a more restrictive definition of IC (that I discuss in point 5 below).
4)The capability of Avida remains in doubt. It is unlikely it will generate complex functions not composed of NANDs. The prospect of meeting my initial challenge to Avida in generating the “TESTPRIME” function without the help of a selection function is rather remote. (Indeed, the “testprime” function has been solved recently by mathematicians, after years of in-depth serious human thoughts.)
5)There are important issues that are still unresolved, related to this EQU experiment. Can an ID theory be developed that excludes a significant factor of an evolutionary mechanism, perhaps as a restrictive ID theory based on more realistic biological observations and biological evidence? The EQU experiment cannot be counted, because of the unrealistic nature of equation-like property of EQU and its limited capability. Other issues involve the detection of IC, can it be predefined at some level attainable from the level of measurements, thus obtaining an easier detection not depending on a recurrent method? There are indications that this theory can be developed, but can be considered a challenge for future works.
These summarize my current understanding.
[ 06. June 2003, 10:56: Message edited by: YZ2 ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 12:12
I have little time for this today, but I'll make two remarks. First, the confusion between "function" and "instruction" may be the case in spite of the paper's authors distinguishing explicitly between the two terms for their purposes in the second sentence of the abstract: quote:
Populations of digital organisms often evolved the ability to perform complex logic functions requiring the coordinated execution of many genomic instructions. (Emphasis added)
Seems pretty clear to me.
But that confusion does not ease the implications of John's suggestion that some kind of "high level" swapping around of instructions is occurring in a restricted sequence space such that only functional programs or functional program fragments can occur. Both the programs themselves, with long stretches of "junk" instructions, and the no-intermediates-adaptive control condition speak against that suggestion. To argue that the conditions of the experiment virtually guarantee the evolution of a program performing EQU is to accept the proposition that irreducibly complex objects can evolve quite readily.
Second, John wrote quote:
Certainly, a point mutation [in biology] never switches out complex instructions/functions like "compare two registers to see if they have the same value" or "replicate" (two real instructions from the instruction list).
"Replicate" is not a "real instruction" from the instruction list. I don't know where John gets that idea - Brig Klyce's short piece perhaps, since he makes the same erroneous claim - but he really needs to read the actual material first-hand. It's difficult (and increasingly frustrating) to address claims about the paper that don't accurately represent it.
Here is the list of 26 primitive instructions. You will notice that "replicate" is not among them.
(a) nop-A No-operation instruction ; modifies other instructions
(b) nop-B No-operation instruction ; modifies other instructions
(c) nop-C No-operation instruction ; modifies other instructions
(d) if-n-equ Test if two registers contain equal values
(e) if-less Test if one register contains a lesser value than another
(f) pop Remove a number from a stack and place it in a register
(g) push Copy the value of a register onto the top of a stack
(h) swap-stk Toggle the active stack
(i) swap Swap the contents of two specified registers
(j) shift-r Shift all the bits on a register one to the right
(k) shift-l Shift all the bits on a register one to the left
(l) inc Increment a register
(m) dec Decrement a register
(n) add Calculate the sum of the values in two registers
(o) sub Calculate the difference between the values in two registers
(p) nand Perform a bitwise NAND on the values in two registers
(q) IO Output the value in a register and replace with a new input
(r) h-alloc Allocate memory for an offspring
(s) h-divide Divide off an offspring contained in memory (specified by heads)
(t) h-copy Make a copy of a single instruction in memory (specified by heads)
(u) h-search Find a pattern of nop-instruction in the genome
(v) mov-head Move a head to point to the same position as the flow-head
(w) jmp-head Move a head by a fixed amount stored in a register
(x) get-head Write the position of a specified head into a register
(y) if-label Test if a specified pattern of nops has recently been copied
(z) set-flow Move the flow-head to a specified position in memory
Finally, I find it amusing that the main criticism of both John and Micah seems to rest on the notion that the experiment wasn't reductionistic enough, that it didn't in effect start with the individual atoms of the several elements - carbon, oxygen, etc. That is literally irrelevant to the question of whether irreducibly complex objects can evolve and to the findings of the Lenski, et al., paper. Please show that the sequence space accessible to the Avida simulation is so small as to make the occurrence of a program performing EQU high. As I've said before, if John believes that the sequence space to which the reported simulation had access is so small, let him calculate its size and report it. The information necessary to do so is available and it's his claim, so the burden is on him to provide and defend that estimate.
RBH
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 12:56
I'll add one more remark. Micah wrote quote:
1. The context for applying IC might be signficiant: mechanical vs. logical
2. Computer simulations testing IC should simulate the spatial and mechanical nature (or at least the spatial aspect) of IC
I think you want to be careful here. You seem to be moving toward giving up abstract information theory altogether, as well as giving up other kinds of symbolic representations of physical systems, making IC a purely materialistic concept. Is there a principled reason why any of the definitions of "IC" must be confined to physical structures? How do claims about the immune system functioning and the blood clotting cascade play into that restriction? Their operations are processes through time, not static physical structures.
I have no problem with the notion of confining the definition of IC to physical (= matter and/or energy) structures and processes, since I don't believe that "information" exists independent of an instantiation in matter/energy (thank you, Rolf Landauer!). But certain prominent IDists surely do, and to confine a core concept like IC to spatially extended physical structures surely rules non-physical information out of the realm of analysis.
RBH
Added in late edit. Of course, the Avida simulation is a physical system. The evolution that occurred was in a physical system, at bottom being pulses of two different voltages being transmitted around a physical substrate. The "instructions" that composed the genomes of Avida critters are various configurations of those pulses, and the evolution that occurred consisted in changes in the frequencies and patterns of those configurations. So even at the level of analysis of instructions, Avida actually is operating with physical systems.
[ 06. June 2003, 13:35: Message edited by: RBH ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 06. June 2003 14:47
RBH,
All of the Avida programs are functional in that all of the primitive instructions for any organism get processed.
All of the primitive instructions already have specific purposes in relation to the Avida architecture, and in that sense are fully functional.
Again, remove one genetic element, "p" and the system is handicapped. Genetic element "p" is essential to the evolution of all rewarded function.
Though I don't know jack about biology, this seems like it would be a descent (if perhaps naive and limited) way to program a front loaded (self-replicating) system: Give the system some primitives with generalized functionality, constrain the domain of variation and evolvabilty, and let the system take on its own life.
And by the way, I do think that IC systems evolve. I just don't think they evolve via Darwinian principles. The evolution in Lenski is so highly constrained and specified from the get go, that it is a more conservative evolution than Mike Gene's front loading thesis.
Sure, if it pleases you, I'm admitting that IC systems evolve. But that was never the question.
I'd like to see the system evolve the ability to perform a 2 input mathematical multiplication function. This seems simple enough, and I would concede that this system is IC in a very real sense that EQU is not.
Three points:
1. The EQU system is quite binary and really doesn't have to evolve the ability to handle more then two types of input (yes or no) and two types of output (yes or no). A multiplication function requires that the function handle multiple degrees of input and output).
2. We know how EQU develops in every lineage, even though we may have loads of junk, and dark alleys along the way. It evolves by the amassing of nand primitives on inputs. Take away the "p" or take away the IO primitive and the system is decapacitated.
3. Though RBH points to the lineages of organisms, we still don't have a good sense of what the "parts" are nor what the system being analyzed is. The parts seem to be the parts of the organism. So is the organism the IC system? The function seems to be something the organism does. So are we talking about "whole organism IC"?
Finally, in the philosophy of mind, people can be separated into two camps: those that can conceive of zombies and those that cannot. RBH and I exemplify two camps when it comes to simulation: I can't conceive of these things as critters, he can. While I view flagellum and their function (motility) as spatio-mechanical systems, I view the Avida and EQU systems as logical representations requiring our interpretation for their existence and thus removed from embodied reality. RBH wants us to conceive of the Lenski system as something autonomous and real, just like strong AI proponents want to conceive of the Chinese Box as understanding Chinese in a "just as real" way as the native Chinese speaker.
These types of conceptual differences often make it hard for discussion to proceed.
[ 06. June 2003, 14:56: Message edited by: Micah Sparacio ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 15:02
Micah wrote quote:
Sure, if it pleases you, I'm admitting that IC systems evolve. But that was never the question.
Um. Well, there goes Josh's assumption: quote:
I assumed that IC meant a system was unevolvable because the paper was targeted towards those who felt that this was ultimately true.
Now I've spent all the time I can afford on this for the rest of the day. Mike Gene will have to read the code for himself, I'm afraid.
RBH
P.S. What was the question, Micah?
[ 06. June 2003, 15:04: Message edited by: RBH ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 06. June 2003 16:03
RBH,
When I say that IC systems can evolve, I'm not giving away the show, and Josh doesn't have to retract his assumption. By evolve I merely mean to move from non-function to function or from one state to another. To say "evolve" doesn't preclude designed evolution. I think that designed evolution may be a promising route for engineers and computer programers to take.
The question, then, is not whether IC systems can evolve, but whether they can evolve from impartial environments, through Darwinian principles. By impartial environment, I mean an environment that is not biased for or against the IC system.
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 06. June 2003 16:51
RBH,
Please go back and read my prior post again, since you're still not getting it.
I made the point that individual instructions (with their complex functions) are guaranteed functional. It is the way they are put together that can have varying degrees of higher-order functionality. Please, go back and read my comments where I stated that no detremental changes can occur below the function (=instruction) level, implicitly assuming that deleterious combinations of instructions above the instruction level are possible!
Until you understand this basic point, I will refrain from responding to you, since it's a waste of my time.
John
PS. When I said replicate is a basic instruction, I was referring to "h-divide" which makes a copy of an offspring stored in memory. I assume that's some sort of replication process. However, even if it's not, it's clearly an incredibly high-level function and the biological irrelevance is clear.
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 06. June 2003 17:15
John Bracht, the h-divide does not make a copy, it separates an offspring from its parent.
Micah Sparacio, the transformation from an initial state (e.g. a population of individuals having N types IC systems) into a later state which in some sense is more impressive (e.g. there are now N+1 types of IC systems present) is precisely what is at issue. In Avida, this transformation occured by the accumulation of modifications and differential reproductive success. Whether the process was "front-loaded" or not is a question that only has metaphysical and theological value (or at least the analogous question for biological has), it's not a question of any scientific interest (because it doesn't help us make predictions).
Erik
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 06. June 2003 17:24
Hi all:
Behe himself in DBB says that IC can indeed evolve indirectly, and the question then becomes one of probability. I don't have the book handy, so I can't provide page number and exact phrasing, but IC has never been an absolute indicator of design. IC is a probabilistic hurdle for evolution, and therefore in an open-minded evaluation of the possibility of ID and the limits and capabilities of evolution, IC is an important concept. (Incidentally, I think the Lenski study is more important in furthering ID and determining the limits and capabilities of evolution than it is in traditional biology, where nothing makes sense except within the light of evolution, and complete capability is assumed!)
Some have said that IC is a subset of CSI. I prefer to think of them as seperate indicators. In this case, EQU evolved and EQU can be looked at as IC. But given the phase space and the fitness landscape, EQU evolved 23 times out of 50 lineages in a limited generational space. Granted more time EQU undoubtedly would have evolved even more often. So EQU is not CSI. The probabilistic hurdle for evolution in this case is not prohibitive. Future designers of evolution can be confident regarding this frontloaded path to EQU.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. June 2003 18:31
Micah wrote quote:
The question, then, is not whether IC systems can evolve, but whether they can evolve from impartial environments, through Darwinian principles. By impartial environment, I mean an environment that is not biased for or against the IC system.
Which is to say, no selection, natural or otherwise. "Darwinian principles" include selection. The most rabid Darwinian would not expect an IC system to arise under conditions of no selection. That's pure random search, and that's not Darwinian evolution.
John,
I understand your point that an individual instruction itself cannot be corrupted in the Avida implementation, and any instruction performs some operation when it is encountered. (I won't say "function" since the paper uses that term for something else.) But that operation may be (often is!) deleterious to the genome (program), it may be neutral, or on rare occasions it may be selectively beneficial. I fail to see how that rescues your repeated implicit suggestion that this somehow vitiates the inference from the experiment that IC structures can evolve from a population of functionless (except for replication) replicators.
Neither I nor anyone else I've read takes this simulation to be a full model of base-pair aggregation into strings that code for proteins that ... etc. etc. etc. I (and others I've read) take it to mean that in an evolutionary model where there are cooptable simple systems that are themselves functioning systems, a higher-level system that meets the operational definition of irreducibly complex can evolve rather easily. That, it seems to me, makes the argument that the existence of IC structures or processes in biological systems is somehow indicative of unevolvability (or cosmic improbability) much harder to sustain. It moves the issue from a general principle ("IC can't evolve") to an IC-of-the-gaps argument ("This IC system couldn't have evolved because we don't know the exact precursors that led to it.")
It further makes it clear that in a sufficiently complex (ordinary language sense) system, there likely is not just one potential pathway, there are most likely multiple pathways, which makes figuring out what the numerator of all those lovely probability estimates that much larger and more difficult to put a number to. "1" doesn't cut it. Moreover, the very different 'solutions' (programs) found in the several evolutionary runs makes it clear that there is not just one 'target' to put in that numerator; there are lots of ways to skin those particular cats and plain old evolutionary processes - mutations and selection - are clever at finding them.
Thus the paper and its implications are a threat to a number of currents of thought in ID, and I don't wonder at the defensiveness I see.
RBH
Added in edit Jack, "EQU" didn't evolve; programs that perform the logic operation EQU evolved. That's an important distinction.
[ 06. June 2003, 18:37: Message edited by: RBH ]
IP: Logged
|
|
Argon
Member
Member # 276
|
posted 06. June 2003 18:36
John Bracht wrote:
"This is my point: in biology, you have to build up the complex functions piece by piece," ...
Actually, no, one doesn't have to. One can build up new, complex functions by rearranging what is already present. For example, swapping of subunits (each with their own simpler "functionality") can create new functionality or permit an existing function to operate in a new context.
To some extent, that appears to be what is happening, both in the Avida simulation and in nature. A large proportion of genes and biochemical systems appear to have been constructed (by some mechanism to be determined) out of shuffling and rearrangements of other genes or gene fragments. This is why many people feel that genetic duplication and recombination (also including pseudogenes) play such important roles in the emergence of new functions.
What is interesting about the Avida simulation is that it really does develop functional programs that were built "piece by piece".
continuing...
"while the Lenski simulation starts with complex building blocks (instructions) and just moves them around."
The issue is not whether the Avida instruction set encodes a set of primitive functions but the manner by which the system exploits the base functionalities to assemble programs that perform new, specified operations. The Avida simulation will tell us little about the origins of its instruction set.
I don't really know how life got started or how many of the "base instructions" of biochemistry first came about. And while that's a really interesting question, it is not all of what evolution is about. What came after, and how such base instructions were rearranged in living organisms over time to give rise to new, high-order systems is the major focus of what has happened to life in the past couple of billion years. The simulation starts with instructions that have a certain level of complexity (base instructions) and moves them around into programs which perform new tasks requiring a higher level of complexity and organization. The basic building blocks available to life are not just random arrangements of nucleotides, but also large, functional subunits and a flexible system for their rearrangement and expression.
I think Jack Foster's last contribution gets it right. Although William Dembksi and others have attempted to equate IC with CSI, they should be considered as separate things. Where I might differ with Jack's assessment is whether the derived EQU program represents CSI, or more exactly, I question how one determines the level of CSI for such a program generically. It seems that only if one understands the system well enough and a good deal of the specific history of the system can one calculate the CSI. For example, Micah hopes to see a two-input multiplication function arise during simulation. Let's say that tomorrow, someone presents Micah with a program that performs the function. If that person tells us nothing about the way the program arose (whether by selective intermediates or design), how would one determine the CSI content?
IP: Logged
|
|
|
Printer-friendly view of this topic
