Uncommon Descent Serving The Intelligent Design Community

Berlinski’s Question Remains Unanswered

Share
Facebook
Twitter
LinkedIn
Flipboard
Print
Email

In a recent post I asked the following question.

I have a question for non-ID proponents only and it is very simple: Is there even one tenet of modern evolutionary theory that is universally agreed upon by the proponents of modern evolutionary theory?

Then I waited for the answers to come in. I was not disappointed, and I would like to express a hearty “thank you” to the proponents of modern evolutionary theory who participated in the exercise.

I have gone through the comments, and the proponents have nominated the following list as tenets on which all proponents of modern evolutionary theory agree:

1. Common descent

2. Modern organisms descend from very ancient ones.

3. The differences among related lineages have accumulated over many generations of change.

4. The basis of (some of) the differences among individuals is heritable.

5. The relative success of those heritable traits is the basis of evolutionary change.

6. Selection is an important evolution force.

7. Populations connected by gene flow evolve together absent very strong selection pull each population in different directions.

8. Selection is important (I am not sure how this is different from 6).

9. Drift exists.

10. Allopatric speciation is possible.

11. Most speciation processes fit somewhere between the extremes of sympatric speciation and allopatric speciation.

12. All of modern life shares common ancestry, HGT and orphan genes notwithstanding.

13. The following mechanisms of genetic change cause change in lineages, with a bias against reversal:
Mutation (point, indel, duplication and rearrangement)
Gene Conversion
Recombination (several mechanisms including HGT)

14. A parallel process of concentration/dilution of such variant alleles occurs (concentration of one is inevitably dilution of the other), by both sample error (drift) and systematic bias (selection).

15. Only frequency-dependent selection can oppose the progress of such an allele through to the fixation of one variant and elimination of the other. This is rare, so origin-fixation is the norm, long-term. [Later withdrawn]

16. Iterated occurrences of such fixations inevitably change lineages.

17. Isolated gene pools diverge.

18. Species differences are fundamentally due to the tendency of this divergence to proceed beyond the point of interfertility (isolation through prezygotic and postzygotic mechanisms).

19. Higher taxa result from ongoing divergences of historic species.

20. Transition-transversion biases in substitution are due to a biochemical tendency increasing yields of one product over the other.

21. Codon position biases occur and are due to the relative effects of selection and drift on synonymous vs nonsynonymous sites.

OK then. Let’s take a look at this list. They seem to fall into the following five categories.

Category 1: Who doesn’t believe that?

Pretty much everyone on the planet would agree with the following proposition: Animals and plants are different now than they were in the past. Thus, the proposition – while at some trivial level a tenant of modern evolutionary theory – is not that which sets it apart from other theories and accounts for its unique purported explanatory power. Even young earth creationists believe it. Therefore, these propositions cannot be the basis for any claim that the theory (as opposed to some other theory) is true. From the above list the following fall into this category: 1, 2, 3, 7, 9, 10, 12, 14, 16, 17, 18, 19, 20 and 21.

Category 2: Trivial

4. The basis of (some of) the differences among individuals is heritable.

Yes, all evolution proponents agree that genetics exists.

Category 3. It is simply not true that all evolution proponents believe it

It is simply not true, for example, that all evolution proponents believe that natural selection plays more than a non-trivial role in the process. It is also not true that all proponents believe sympatric speciation occurs. 5, 8 and 11 fall into this category. See here and here.

Category 4. I Don’t know what claim is being made

13. The following mechanisms of genetic change cause change in lineages, with a bias against reversal:
Mutation (point, indel, duplication and rearrangement)
Gene Conversion
Recombination (several mechanisms including HGT)

Is the proponent of this claim stating that genetic change and only genetic change causes change in a lineage? If so, it is clearly not the case that all proponents of the theory agree with this; indeed most of them would dispute it. Is the proponent claiming merely that genetics change occurs and somehow that gets fixed in a species? If that is the case, it would fall under category 1.

Catategory 5. Withdrawn: claim 15.

CONCLUSIONS

My suspicions have been confirmed. Proponents of modern evolutionary theory all agree on a set of propositions that even most fervent young earth creationist would also agree on. And nothing more as far as I can tell.

What I was really trying to get at was this: Is there any “core” proposition on which all proponents of modern evolutionary theory agree. By “core” proposition, I do not mean basic facts of biology that pretty much everyone from YECs to Richard Dawkins agrees are true. I mean a proposition upon which the theory stands or falls, and, as I said above, sets it apart from other theories and accounts for its unique purported explanatory power

I have in mind a proposition that would answer David Berlinski’s famous question:

I disagree [with Paul R. Gross’ assertion] that Darwin’s theory is as “solid as any explanation in science.” Disagree? I regard the claim as preposterous. Quantum electrodynamics is accurate to thirteen or so decimal places; so, too, general relativity. A leaf trembling in the wrong way would suffice to shatter either theory. What can Darwinian theory offer in comparison?

Indeed. What does modern evolutionary theory offer in comparison? How can the theory ever hope to be as “solid as any explanation in science” when its proponents cannot seem to agree on a single tenet, the falsification of which would, in Berlinski’s words, shatter the theory?

UPDATE:

In the comments eigenstate pulls Haldane’s famous “rabbit in the Cambrian” out of his hat. I will address that here:

Everyone knows there are no rabbit fossils in the strata that have been labeled “Cambrian.” First, eigen’s sputtering to the contrary notwithstanding, the primary reason a stratum would be called “Cambrian” in the first place is because of the absence of rabbit fossils.

Set that aside for the moment and consider this. The fact that there are no rabbit fossils in the Cambrian strata is a datum. It is a datum for evolutionary theory; it is a datum for young earth creationists; it is a datum for ID proponents. It is a fact on the ground in the same way that people are stuck to the earth with a force of 1G is a fact on the ground. Saying “a rabbit in a Cambrian stratum would destroy Darwinism” is equivalent to saying “if people start floating off into space it would destroy general relativity.” Well, people are not floating off into space and they are not about to. No rabbits have been found in the Cambrian strata and none ever will be. Haldane’s observation amounts to nothing more than “if the facts were different the theory to explain those facts would have to be different too.” It is trivially true and singularly unhelpful.

I take it that Berlinski’s point is very different. The mathematical models of quantum electrodynamics and general relativity make extremely precise predictions (13 decimal points). It follows that those theories have exposed themselves to an enormously high degree of “risk,” because if an observation were to vary from prediction by even an astronomically small degree it would falsify the theory.

Now consider the following exchange:

Barry: And yet unlike respectable scientific theories, there is not universal agreement on a single one of those details that sets [evolutionary] theory apart as an explanatory mechanism.

eigenstate: so what?

Well, here is what. Paul R. Gross’ asserted that Darwin’s theory is as “solid as any explanation in science.” Berlinski retorted that the assertion is preposterous. The point of the OP is that Berlinski is certainly correct. There is not even agreement among evolutionary theorists on what the model should be in the first place; far less has anyone developed a model that would make exquisitely precise predictions equivalent to those made by quantum electrodynamics and general relativity. Thank you eigenstate for confirming Berlinski’s point with your “so what.’

Comments
Still no code.Mung
May 29, 2015
May
05
May
29
29
2015
08:39 PM
8
08
39
PM
PDT
And still no code. Carpathian, if you are so concerned that your idea is so original and valuable and you don't want it copied, please hurry up and apply for the patent so we can see your code.Mung
May 24, 2015
May
05
May
24
24
2015
06:55 PM
6
06
55
PM
PDT
Carpathian, you are not modeling a random process. In fact, you are not modeling anything at all, which is why I said you suffered from map/territory confusion. You are writing a program that can search for a target. It may or may not meet that goal. If it does meet that goal, it is because it was designed to do so. The opposite of random.Mung
May 23, 2015
May
05
May
23
23
2015
05:10 PM
5
05
10
PM
PDT
Mung, Modeling random processes has been accepted by gaming commissions in Nevada and Atlantic City. The operation of these random slot and poker machines are also accepted by the customers who use them. They are also random enough to prevent people from determining beforehand what the results might be. As far as needing the source, you can select attributes that will cause it to fail to "find" the "target" simply by using command line arguments. I am now adding a special mode that does not eliminate members that don't meet the environment's requirements. It simply mutates and marries which ripples random changes through the population. This is to show that without some sort of negative feedback, random change can completely change massive amounts of DNA. If biology changed at that random speed, we would not be here.Carpathian
May 23, 2015
May
05
May
23
23
2015
12:12 PM
12
12
12
PM
PDT
No, I want to see the source code so I can point out to anyone still watching how it has the designer's fingerprints all over it and how if I tweak it just a little bit it won't work. So not only is it designed to do what it does it's also fine-tuned.Mung
May 23, 2015
May
05
May
23
23
2015
10:34 AM
10
10
34
AM
PDT
Mung:
Still no “code.”
Presumably you want to see the source to see if it would run, but when I offered to send you the running code, which runs, you declined, because you'd rather see the source, to see if it would run. Funny.Carpathian
May 23, 2015
May
05
May
23
23
2015
08:22 AM
8
08
22
AM
PDT
Still no "code."Mung
May 23, 2015
May
05
May
23
23
2015
12:10 AM
12
12
10
AM
PDT
Mung:
Design. Design. Design. LoL.
Map. Territory. LOLCarpathian
May 21, 2015
May
05
May
21
21
2015
09:43 AM
9
09
43
AM
PDT
ES, When you composed a text string to make a comment, you created functionally specific, complex organisation and associated information, FSCO/I. Similarly, when you create an algorithm and data structures then code to program a solution to a travelling salesman problem etc, you do the same. Your refusal to acknowledge the descriptive relevance of the term, the underlying reality and the implication, speaks inadvertent volumes to the force of the implications of that reality. Especially, just how strongly such points to design as key cause. KFkairosfocus
May 21, 2015
May
05
May
21
21
2015
02:26 AM
2
02
26
AM
PDT
Could you please say a bit more of what you mean by functional information? Perhaps include an example?
Sure. An accessible example would be measuring the time to visit all of the "cities" given for solving the Traveling Salesman Problem with GA. The fitness function represents the scoring metric for a given specimen at a given time. So, if permutation X yields a "time to visit all nodes" of 44 days, and permutation Y comes in at 35 days, Y is more likely to be kept/reproduced with permutations than X. This isn't "functional information" in the FSCO bafflegab sense that is often used here. Rather, we have metrics that we calculate about our candidates. It's fine and clear to call this "performance information" in this example, for many tasks I've worked on, "functionality" is a better word to describe the fitness/results.
It seems to me that, at least in your typical GA, all candidate solutions are functional in the sense you are using the term, but some are seen to be more functional (whatever that means, lol) than others. These then typically become the foundation for the next generation.
Yes, you have it.
; There is no non-functional information to be found.
If you mean that in "FSCO/I" sense, I guess not. But probably because I can't pull anything meaningful or substantive out of that... nonsense.
In many GAs I run, such mutations don’t have any effect at all, for the immediate generation, or even for many generations hence, because the mutation doesn’t ‘code for’ behaviors that the fitness function will discriminate against (at least at this time).
Right, but it's interesting when we do find novel solutions and look at the lineage, often we find "sleepers", and several of them at points in the evolutionary history, mutations that didn't come into play until long after when a copy op made an otherwise inert change many generations back "the ticket".
Do you write the code or just run the code?
The systems I work on are a mix (or collaboration, more precisely) of GA and agent-based modeling and simulation frameworks. I worked on the GA side more heavily a few years ago, am more focused on the ABM side now. But for the most part I write code, and work on system architecture for the projects. None of the systems I work with are "all written by me", though, these are all large commercial projects. ETA: blockquoteeigenstate
May 20, 2015
May
05
May
20
20
2015
09:23 PM
9
09
23
PM
PDT
Meanwhile, in another thread, Carpathian admits to his role as the designer/overseer, performing experiments in order to observe the results. The why of it all is left as a mystery. One might conjecture, however, that the reason is to determine whether the program performs according to the expectations of the designer and if not to make the necessary adjustments. Intelligent Design.Mung
May 20, 2015
May
05
May
20
20
2015
08:47 PM
8
08
47
PM
PDT
Carpathian: /c[num] – number of children You could hard code the number of children into the program as a fixed value. Why didn't you? You could have the program generate the number of children according to some algorithm. Why didn't you? What's the point of making the value of this parameter settable on the command line if it has no effect on the outcome? How did the program determine that this setting could have an effect on the outcome, or was it a design decision on your part? What if the user sets the number to 0? What if the user supplies no number? Do you provide a default? If so, what default do you provide, and why? Design. Design. Design. LoL.Mung
May 19, 2015
May
05
May
19
19
2015
10:37 PM
10
10
37
PM
PDT
Dear Carpathian, Did you give your program a choice in the matter or were these design decisions you dictated to it and forced it to follow? You keep dancing around, you're quite the accomplished dancer. Still no source code though. All you've done is make these things user selectable before running the program. They don't for that reason just up and disappear.Mung
May 19, 2015
May
05
May
19
19
2015
12:43 PM
12
12
43
PM
PDT
Mung: Here are the available options.
/c[num] - number of children /d - DNA mode 'ATCG' /f[string] - optional target entry /g[size] - generate target string /i[runs] - number of trials /j[length] - length of coding DNA /l[filename] - log to filename /m[type] - type of marriage /n - don't display individuals /p - population size /r[n] - remove 1/n mutations /s - no console output
Carpathian
May 19, 2015
May
05
May
19
19
2015
09:26 AM
9
09
26
AM
PDT
Mung:
What protocol (rule or procedure) did you implement that restricts the permissible characters of the target phrase?
A command line option, (/d), selects a 4 char DNA mode while the default is a 93 char text set for Weasel type of operation. I think ID and evolution is primarily concerned with the 'ATCG' set.Carpathian
May 19, 2015
May
05
May
19
19
2015
09:23 AM
9
09
23
AM
PDT
Mung: Protocols are agreed upon queries and responses to allow communication between active nodes. Rules, procedures and algorithms are not protocols. The target phrase is delimited by a zero. The user can select options from the command line. example:
GEN 12: 0 ATCGATCGAACGATCG #Matches: 15 ATCGATCGATCGATCG GEN 13: 27 ATCGATCGATCGATCG #Matches: 16 ATCGATCGATCGATCG -------------------------- carp /lLog.log /s /p64 /g16 /i10 /c6 /d /m5 Population Size = 64 Fitness String Len = 16 Pop Len/Fit Len = 4.000000 No Change = 1 PosChange = 16 NegChange = 0 Multiple Change = 4 % MultiBit changes = 30.77 Number of generations = 13
In a lab, a technician can use an oscilloscope to probe an electronics circuit. The scope has dials and switches on it so the user can modify parameters of the tool for whatever experiment he is doing. No one would accuse the manufacturer of "smuggling information" into the experiment because he has allowed the user to control the test. The same goes for programmers, especially "evolution" programmers. The questions you ask are of the level, "Why are stripes on a resistor different colours?"Carpathian
May 19, 2015
May
05
May
19
19
2015
09:19 AM
9
09
19
AM
PDT
Carpathian, do you honestly expect us to believe that the protocols you designed into your program were chosen for their lack of effect?Mung
May 17, 2015
May
05
May
17
17
2015
08:30 AM
8
08
30
AM
PDT
Here's one definition of protocol: a system of rules that explain the correct conduct and procedures to be followed in formal situations So you designed the rules and that is precisely what I am saying you've done. Are you going to stop denying it now? Meanwhile, you refuse to answer simple questions. What protocol (rule or procedure) did you implement that restricts the length of the target phrase? What protocol (rule or procedure) did you implement that restricts the permissible characters of the target phrase? What protocol (rule or procedure) did you implement that determines the length of your candidate solutions. And I have all the experience I need in dealing with people who avoid answering the simplest of questions and who refuse to make available their source code.Mung
May 16, 2015
May
05
May
16
16
2015
06:13 PM
6
06
13
PM
PDT
Mung: You again show your inexperience. There is a protocol to be followed in any data being loaded or communicated. That protocol is in itself not data. Data has to be in some agreed upon format otherwise Word could not load files, and the Internet would not be possible. Data content is independent of the protocol required to exchange it. My program does not know anything about the data it requires. It knows about the protocol. If the protocol was a header prepending the data or whether a string is zero terminated, makes no difference. I will coin a new term for data of indeterminate length. I will call it a Mung packet. Anyone who gets one can then say, "I've got Munged data".Carpathian
May 16, 2015
May
05
May
16
16
2015
05:25 PM
5
05
25
PM
PDT
Carpathian: The program has no information about the target when it is created. Let me quote you:
Notice that I am using 93 chars in my text char set.
Are you saying the target is not limited to this set of 93 text characters? If it is, then your statement that the program has no information about the target when it is created is simply false.Mung
May 16, 2015
May
05
May
16
16
2015
04:30 PM
4
04
30
PM
PDT
Carpathian: The program has no information about the target when it is created. So? Carpathian: The program does not need to be recompiled when changing target strings. Hardly a point that bothers me. I program primarily in a dynamic language. Recompilation isn't the issue. Carpathian: Your claim was that the program would have to recompiled when the string changed. Wrong again. And now you're changing the subject to avoid the blatantly obvious. How does the program know the length of the candidate solutions that it must generate? Are you telling us that you did not program it to look at the target string at any other point in the program than when it compares a candidate string to the target string to derive the "fitness" of the candidate? You’re just not believable.Mung
May 16, 2015
May
05
May
16
16
2015
04:26 PM
4
04
26
PM
PDT
Mung: The program has no information about the target when it is created. That is the point you and I were in disagreement on. The program does not need to be recompiled when changing target strings. Microsoft Word does not need to be recompiled when presented with a new document. Excel does not need to be recompiled when reading in a new spreadsheet.
This is all data introduced into the program
Again, the data is being processed by a program that doesn't change. Your claim was that the program would have to recompiled when the string changed. It doesn't, agreed?Carpathian
May 16, 2015
May
05
May
16
16
2015
04:07 PM
4
04
07
PM
PDT
Here's code from a Python version of the WEASEL program: # Target sequence target = list("METHINKS IT IS LIKE A WEASEL") # Alphabet of possible symbols (uppercase letters + blank) alphabet = "ABCDEFGHIJKLMNOPQRSTUVWXYZ " # Number of offspring per generation n_offspring = 50 # Mutation rate: probability that any given letter will change mut_rate = 0.09 This is all data introduced into the program You likewise have introduced similar data into your program. You've already admitted to it up thread. Further: # Construct random starting string of same length as target parent = [] for i in range(len(target)): parent.append(random.choice(alphabet)) Yes, it's true. Knowledge of the target, in this case, it's length. That's data about the target. Carpathian, you're just not believable. Sorry.Mung
May 16, 2015
May
05
May
16
16
2015
03:30 PM
3
03
30
PM
PDT
Mung:
You, the intelligent designer, have injected information into your program which drastically reduces the size of the search space. And ASCII supports 128 characters, not 93. (One of which is the NUL character.) So are you finally admitting that I was right all along?
No. I have not introduced anything into the program. The program processes data .Carpathian
May 16, 2015
May
05
May
16
16
2015
02:00 PM
2
02
00
PM
PDT
eigenstate:
Doesn’t matter for the creative process, for the origination of novel information. All you need is a source of random input (or for most computer simulations, “pseudo random” source like a Mersenne Twister. Once you have a source for novel information, functionality information obtains through iterative testing of the functionality of mutations.
Could you please say a bit more of what you mean by functional information? Perhaps include an example? It seems to me that, at least in your typical GA, all candidate solutions are functional in the sense you are using the term, but some are seen to be more functional (whatever that means, lol) than others. These then typically become the foundation for the next generation. There is no non-functional information to be found.
In many GAs I run, such mutations don’t have any effect at all, for the immediate generation, or even for many generations hence, because the mutation doesn’t ‘code for’ behaviors that the fitness function will discriminate against (at least at this time).
Do you write the code or just run the code?Mung
May 16, 2015
May
05
May
16
16
2015
08:10 AM
8
08
10
AM
PDT
@computerist
One of the problems with evo algorithms that posit to create information of functional significance, is that unless you have top-down intelligent selection, nothing you do will matter. Take for instance this fictitious compiled binary string, assume subsets (previous iterations) of this string were functional, and that step-by-step, it happened to reach a more sophisticated functional state. 1101010100101010101010101010010010101010001010101010101010 At the next iteration (generation), you have a random mutation that comes along. 1100010100101010101010101010010010101010001010101010101010 Oops. At this point, it’s no more significant (including has no more selective/fitness advantage) than any other string you can randomly pick out of a hat. The single mutation overridden the entire functional state, rendering it meaningless. Garbage “collection” mechanism is at the heart of Darwinian Theory, IMHO.
If by "garbage collection" you mean "selection" (which really, for clarity, in biology, should be called "selection-against", as it is a negative filter), then yes. Selection is a key dynamic in evolution. The origin of the fitness landscape does not matter to the rm+ns cycle. In biology, the environment, constantly changing, provides the selection filter. In a GA-based program, the fitness function may be simple and static ("find 'weasel'"), or highly dynamic and complex, and changing itself based on stochastic inputs that render the fitness filter unknowable at compile time. Doesn't matter for the creative process, for the origination of novel information. All you need is a source of random input (or for most computer simulations, "pseudo random" source like a Mersenne Twister. Once you have a source for novel information, functionality information obtains through iterative testing of the functionality of mutations. In your example, the system is already able to create novel functions -- you've stipulated in your setup ("and that step-by-step, it happened to reach a more sophisticated functional state"). So when a mutation produces the resulting string in your example, if it's non-function (something like "will die in in utero" in biology -- completely non viable, not even able to reach 'fetus' stage), that's the end of the road for that particular thread. Perhaps you are referring to some fitness filter I missed upthread or something, but while the mutation does produce a different (hence, 'mutated') string, that change is not as a general rule lethal or deleterious, just by virtue of being new/different. Ostensibly, that's how the system got where you picked up ("... it happened to reach a more sophisticated functional state."). All of which to ask, why this statement, then:
At this point, it’s no more significant (including has no more selective/fitness advantage) than any other string you can randomly pick out
That's not the case, or at least that can only be the case when the fitness function is a pure random number generator. The mutated string may (and in most cases will -- evolution works by lots of fails that get discarded and a few wins that get preserved and accumulated) be "dead on arrival", functionally. But it may be functionally inert, or advantageous. You don't know until you apply the fitness function to it -- let it "makes its way in the (digital) world" you've created . One cannot say before hand that the mutated string has "no more significant (including has no more selective/fitness advantage) than any other string you can randomly pick out of a hat". If the fitness function does have means for providing higher fitness scores for some candidates than others, that claim won't work (in other words, if your fitness function does more than produce random testing criteria). Furthermore, "The single mutation overridden the entire functional state, rendering it meaningless" may certainly be the case, but this is fitness function specific. In many GAs I run, such mutations don't have any effect at all, for the immediate generation, or even for many generations hence, because the mutation doesn't 'code for' behaviors that the fitness function will discriminate against (at least at this time). The whole point of running GAs is that one cannot say "this small mutation will ruin everything, necessarily". That may be (and usually is) true for the vast majority of permutations, but evolution, as I said above, harnesses this imbalance by discarding fails that occur in great numbers, and keeping the few wins (or 'ties' as the case may be). Perhaps the "mutated string" you've defined out there is one that enables significantly enhanced functionality. In this case, it would be much more "than any other string you can randomly pick out of a hat". Right? (hint: true by definition, which is the point of asking!)eigenstate
May 14, 2015
May
05
May
14
14
2015
08:41 PM
8
08
41
PM
PDT
haha...well not that kind of GC, that's too smart!computerist
May 14, 2015
May
05
May
14
14
2015
07:33 PM
7
07
33
PM
PDT
Yes, GA's use a mark and sweep GC. :)Mung
May 14, 2015
May
05
May
14
14
2015
07:01 PM
7
07
01
PM
PDT
One of the problems with evo algorithms that posit to create information of functional significance, is that unless you have top-down intelligent selection, nothing you do will matter. Take for instance this fictitious compiled binary string, assume subsets (previous iterations) of this string were functional, and that step-by-step, it happened to reach a more sophisticated functional state. 1101010100101010101010101010010010101010001010101010101010 At the next iteration (generation), you have a random mutation that comes along. 1100010100101010101010101010010010101010001010101010101010 Oops. At this point, it's no more significant (including has no more selective/fitness advantage) than any other string you can randomly pick out of a hat. The single mutation overridden the entire functional state, rendering it meaningless. Garbage "collection" mechanism is at the heart of Darwinian Theory, IMHO.computerist
May 14, 2015
May
05
May
14
14
2015
06:10 PM
6
06
10
PM
PDT
Still no source code.Mung
May 14, 2015
May
05
May
14
14
2015
05:38 PM
5
05
38
PM
PDT
1 2 3 11

Leave a Reply