(Sorry this is so long – I am in a hurry)
Gpuccio challenged myself and others to come up with examples of dFSCI which were not designed. Not surprisingly the result was that I thought I had produced examples and he thought I hadn’t. At the risk of seeming obsessed with dFSCI I want assess what I (and hopefully others) learned from this exercise.
Lesson 1) dFSCI is not precisely defined.
This is for several reasons. Gpuccio defines dFSCI as:
“Any material object whose arrangement is such that a string of digital values can be read in it according to some code, and for which string of values a conscious observer can objectively define a function, objectively specifying a method to evaluate its presence or absence in any digital string of information, is said to be functionally specified (for that explaicit function).
The complexity (in bits) of the target space (the set of digital strings of the same or similar length that can effectively convey that function according to the definition), divided by the complexity in bits of the search space (the total nuber of strings of that length) is said to be the functional complexity of that string for that function.
Any string that exhibits functional complexity higher than some conventional threshold, that can be defined according to the system we are considering (500 bits is an UPB; 150 bits is, IMO, a reliable Biological Probability Bound, for reasons that I have discussed) is said to exhibit dFSCI. It is required also that no deterministic explanation for that string is known.”
(In some other definitions Gpuccio has also included the condition that the string should not be compressible)
These ambiguities emerged:
Some functions are not acceptable but it is not clear which ones. In particular I believe that functions have to be prespecified(although Gpuccio would dispute this). Also functions which consist of identifying the content of “data strings” (a term which is itself not so clear) are not acceptable because the string in question could have been created by copying the data string.
The phrase “no deterministic explanation for that string is known” is vague. It is not clear in how much detail and how certainly the deterministic processes have to be known. For example, it appears from above the possibility that the string in question might have been copied from the string defining the function by some unknown method is sufficient to count as a known deterministic explanation. This implies that really it is sufficient to be able to conceive of the very vague outlines of a determinist process to remove dFSCI. I think this amounts to another implicit condition: no causal relationship between the function and the string.
Lesson 2) dFSCI is not a property of the string.
It is a relationship between a string, a function and an observer’s knowledge. Therefore, it may be that dFSCI applies for a string for one observer with a certain function but not for another observer with a different function. The rules for deciding which function are not clear.
Lesson 3) The process for establishing the relationship 100% specificity of dFSCI and design is not commonly found outside examples created by people to test the process.
Gpuccio says thisabout the process:
“To assess the dFSCI procedure I have to “imagine” absolutely nothing. I have to assess dFSCI without knowing the origin, and then checking my assessment with the known origin.”
When challenged he was unable to name any instances of this happening outside the context of people creating or selecting strings to test the process as in our discussions. This is important as the dFSCI/design relationship is meant to be an empirical observation about the real world applicable to a broad range of circumstances (so that it can reasonably be extended to life). If it is only observed in the very special circumstances of people making up examples over the internet then the extension to life is not justifiable. To give a medical analogy. It might well be that a blood test for cancer gives 100% specificity for rats in laboratory conditions. This is not sufficient to have any faith in it working for rats in the wild, much less people in the wild. Below I discuss what is special about the examples created by people to test about the process.
A Suggested Simplification for dFSCI
dFSCI says that given an observer and a digital string where:
1) The observer can identify a function for that string
2) The string is complex in the sense that if you just created strings “at random” the chances of it performing the function are negligible
3) The string is not compressible
4) The observer knows of no known deterministic explanation for producing the string
Then in all such cases if the origin eventually becomes known it turns out to include design.
Given the rather lax conditions for “knowing of a deterministic mechanism” that emerged above, surely (2) and (3) are just special cases of (4). If (2) or (3) were present then deterministic mechanisms would be conceivable for creating strings.
So the dFSCI argument could be restated:
Given an observer and a digital string where:
* The observer can identify a function for that string
* The observer cannot conceive of a deterministic explanation for producing the string
Then in all such cases if the origin eventually becomes known it turns out to include design.
Conclusion
There are two main objections to the ID argument:
A) There are deterministic explanations for life.
B) Even if there were no deterministic explanations it would not follow that life was designed
For the purposes of this discussion I will pretend (A) is false and focus on (B)
No one disputes that it is possible to detect design. The objectors to ID just believe that B) true. The correct way of detecting design is to compare a specific design hypothesis with alternatives and assess which is provides the best explanation. This includes assessing the possibility of the designer existing and having the motivation and ability to implement the design. If no specific hypothesis is available then nothing can be inferred.
So is the dFSCI claim above true and if so does it provide a valid alternative way of detecting design?
The trouble is that there is dearth of such situations. One of the reasons for this is that digital strings do not exist in nature above the molecular level. At any other level it is only a human interpretation that imposes a digital structure on analogue phenomena. The characters you are reading on this screen are analogue marks on the screen. It is you that is categorising them into characters. So all such strings are created by human processes. It follows that design is a very plausible explanation for any such string. People were involved in the creation and could easily have designed the string. If you add the conditions that the function must be prespecified and there should be no causal relationship between the function and the string then design is going to be by far the best explanation. It goes further than that. It also means there almost no real situations where someone is confronted with a digital string without knowing quite a bit about its origin – which is presumably why Gpuccio can only point to examples created/selected by bloggers.
What about the molecular level? Here there are digital strings that are not the result of human interpretation. Now human design is massively implausible (except for a few very exceptional cases). The problem now is that carbon chains are the only digital strings with any kind of complexity and these are just the one’s we are trying to evaluate. There are no digital strings at the molecular level with dFSCI except for those involved in life.
So actually the dFSCI argument only applies to a very limited set of circumstances where a Bayesian inference would come to the same conclusion.
I understand and have even explained that very point to others.
The problem is that an argument cannot be followed when the definitions of terms are not consistent.
gpuccio should have one term, “dFSCI”, for the configuration of a string which is independent of origin, and a separate one when determining the means by which that string was generated.
This is the frustration that always comes up when talking with IDists, and that is that they somehow assume we must know what they mean by their own unique terms, even as we ask for clarification.
As far as gpuccio’s “dFSCI” goes, a string that does not seem to be generated by a random process but can be generated by any “non-design mechanism” such as a “necessity mechanism”, simply means that the “search space” for that mechanism is of a non-random distribution, as it is for design.
Neither mechanism is a “roll of the dice” and the UPB improbability argument is therefore not valid in those cases.
Toronto,
But just a couple of comments ago you were still conflating “origin” and “explanation”. You wrote to gpuccio:
You were referring to this statement of gpuccio’s:
You complained that
That’s not correct. Criterion 1c doesn’t refer to the origin of the string, it refers to explanations of the string (hence the word “explains”).
Gpuccio is certainly off base in accusing you of lying, and he has done a poor job of explaining himself, but he does draw a distinction between origins and explanations that you seem to have missed.
Alan writes:
Alan, dFSI and dFSCI are not the same. In the comment you linked to, Gpuccio confirms only that dFSI and FSC are the same:
The difference between dFSI and dFSCI is that every functional string has some value of dFSI, large or small, but a string only has dFSCI if 1) the dFSI is greater than a threshold that is computed based on the system and time span in question, and 2) no known “necessity mechanism” can explain the string. The first qualifier excludes strings that could conceivably have been produced by pure random variation within the allotted time, and the second excludes strings that could be produced by unguided evolution and other known “necessity mechanisms”.
None of that changes the fact that dFSCI is useless, of course, but it is conceptually distinct from dFSI.
An extremely silly argument from gpuccio:
Even setting aside his tendentious interpretation of the data, the logic is laughable. He is saying, in effect, “If you can’t give me lots of information about the intermediates, I’m allowed to assume that there are none. Zero.”
Applying the same logic to gpuccio’s designer gives amusing results: “If you can’t give me lots of scientific information about your designer, I’m allowed to assume he doesn’t exist.”
Should gpuccio attempt to argue that life itself “shows us” the designer, I would remind him that the evidence is trillions of times better for unguided evolution (and consequently for the presence of selectable intermediates) than it is for his hypothetical designer.
But the explanation is the origin, it’s the only thing left! 🙂
Let’s say that we have a string that meets the requirements of being digital, functional, complex and specific.
We find it is above the UPB threshold and now it is a candidate for a design explanation.
What else can you tell me about it other than why it has that configuration?
gpuccio uses explain and origin as equivalent terms here and so have you.
In one case the origin is “you”, a designer, and the other case it is a “D flip-flop”, a necessity mechanism.
If your claim is that it might be due to design or it might be due to a necessity mechanism, you’ve already said enough, which is that a certain information-rich string, can be due to a “non-design mechanism”.
gpuccio’s “dFSCI” attribute , is withheld from strings that in all other respects, would be classified as “dFSCI” if their origins, i.e. how they came to be in the bit configuration that we observe, were not considered.
Toronto,
Let me try putting it differently. For any given instance of a string, such as my “10101010…” example, there is one actual origin. The string may have been designed or it may have been produced by a non-design mechanism. In my case it was designed. I thought it up and typed it in.
If you didn’t know the actual origin of a string, then you would have to consider possible origins, aka “explanations”. Looking at my string, you would see that both design and non-design explanations are possible. Gpuccio wants to avoid false positives, so if he knows of both design and non-design explanations of a string, he declares that the string does not exhibit dFSCI. This means that some strings that are actually designed will be assessed as not exhibiting dFSCI. Gpuccio will not infer design for them. They are false negatives — an unavoidable consequence of using the explanatory filter.
Of course, false positives remain a problem for gpuccio because he will only disqualify strings for which he knows of a mechanism that he considers to be a satisfactory explanation. His argument boils down to this: If I (gpuccio) don’t know of a non-design mechanism that I consider sufficient to explain this string, then there isn’t one and design should be inferred. dFSCI is just a smokescreen that obscures this obviously bogus argument.
The key point in all of this is that the actual origin of a string is not used in assessing dFSCI, even when the actual origin is known. The mere existence of a known “necessity mechanism” that can produce the string is enough to disqualify the string from exhibiting dFSCI, even if that particular instance of the string happens to be designed.
The actual origin of a string is not used to assess dFSCI. Potential origins (aka “explanations”) are.
Gpuccio 638
For the moment I will just take the last part of this comment.
I disagree and that is what I have tried to explain.
That is partly true. There is a rational process for inferring design and you might rationally come to different conclusions depending on your a priori beliefs about reality. However, these rational processes are not ID theory as usually expounded and do not include the argument from dFSCI. I am trying to explain why that is so.
Also, we should not hide the requirement for those prior beliefs in order to make that rational deduction. They are prior beliefs, not something that you can deduce from the evidence. Your prior belief is that there is a designer with the appropriate motives and abilities. Mine is that nothing is proven about there being a designer or not.
Well of course I don’t accept they are valid scientific theories so I can’t try to falsify them at the scientific level.
I don’t think I have ever argued dFSCI is not politically correct! I believe there are deep methodological problems with your argument from dFSCI independent of evolutionary theory. As someone with a philosophy of science/statistics background I am vaguely qualified to discuss that. I am not a biologist and am not qualified to discuss the evidence for modern evolutionary theory although I am impressed by what I read.
If you only want to discuss the evidence for “darwinism” I am not your man. If you are at all interested in perceived methodological problems with dFSCI then I am interested to do my best to explain them. I know I have tried it many times but there are always new ways of putting the problems.
Yeah, I don’t see how anyone could possibly get dFSCI, FSC and dFSI mixed up! 🙂
Gpuccio 638 (cont)
Now for the second part of the response. I know you will disagree with this. All I ask is you believe that it is a thoughtful serious response. You may well decide to ignore it – which is fine by me.
I continue to learn more about your definition of dFSCI and find there are things I have misunderstood. As a result my account of the problems does vary. Here is my most recent understanding and account. In what follows I have labelled key propositions (A), (B) and (C) so I can easily refer back to them.
As a result of all the test examples and discussion (which I note has been going on since September!) I now believe you have described a process for detecting design that is equivalent to:
“in the case of digital strings with a function, if there is no known deterministic explanation why the string should happen to have a configuration that performs that function then you can infer design”. (A)
It seems to me that the other conditions – complexity and non-compressibility – just rule out special cases of deterministic explanation (where selecting a string at “random” from all possible strings is one kind of deterministic explanation). That still leaves some uncertainty over the phrase:“no known deterministic explanation”. I now think you mean:
The majority of people in our society at this time cannot conceive of even the outlines of a deterministic explanation. (B)
You “prove” (A) mainly by pointing out that there are:
For obvious reasons neither of these prove much! No one denies that there are lot of digital strings that have been designed and a good proportion of them will have no conceivable deterministic explanation. The second part is trivially true.
What would count as evidence for (A) would be lots of situations where:
There is a digital string with no known deterministic explanation and the explanation is not known then later the explanation becomes known (C)
If (C) happened a lot and it turned out that every time the explanation included design then it would be a correlation worth exploring (although I would still want to understand why before blithely applying it to another domain). Unfortunately (C) appears to be extremely rare. Can you think of any? I can only think of analogue examples and in these cases the explanation often turns out to be deterministic. In the case of digital strings where the explanation eventually becomes known the context virtually always makes it pretty clear what the explanation was from the start. I tried using my imagination to construct hypothetical examples such as the Fibonacci series but you didn’t like that.
Given your reluctance to pursue hypothetical examples the alternative was the game we played. We try to construct examples of strings where you can’t conceive of a deterministic explanation but we know of one. It is not exactly the same as (C) but is close. This proves to be hard because you can raise the bar of “cannot conceive of deterministic explanation” very high. Perhaps the best example was the London temperatures. I described how to generate a digital string that could be used to tell you whether London temperatures were higher or lower than average. You dismissed this because the string could have been copied from the pattern of temperatures in some undefined way. Basically we have to think of a way that digital strings can be determined that you could not possibly conceive of in even the vaguest way – quite a challenge.
So in practice we are unable to think of/construct a test case which would provide evidence for (A). There are other problems – even if the test cases flowed and they always resulted in a designed explanation I would stick to my guns that you cannot rationally transfer a correlation like this in one domain (man-made digital strings) to another (molecular digital strings) without understanding why the correlation happens.
PS No a priori principles or worldviews were used in the construction of this argument!
Agreed 100%
That is exactly what gpuccio is doing.
If a potential origin is not acceptable, gpuccio withholds the designation “dFSCI”.
The set of potential origins contains as an element, the actual origin.
This has to be the case since the actual origin must qualify as having the potential to be true.
So here’s the set:
ORIGINS_Potential { Design, Necessity Mechanism, .. }.
However, gpuccio wants a set like this:
ORIGINS_Potential { Design }.
This is his test criteria for real world strings which he will use to infer design.
Note that necessity mechanism is gone.
What he wants to end up with is this:
ORIGINS_dFSCI { Design } instead of this:
ORIGINS_dFSCI { Design, Necessity Mechanism, .. }.
A potential origin should be treated as such and not waved away as an explanation.
Gpuccio 650
You move from irony to sarcasm! But I don’t see how it relates to my comment. I was declaring my prior beliefs and I emphasised that they were not something you could deduce from evidence. You respond by mocking my use of evidence in unrelated questions – very strange.
Anyhow, having told you my prior beliefs, I am interested to know whether you admit that the ID argument is not valid unless you have a prior belief that there is a designer with the appropriate powers and motivation i.e. ID is not evidence for such a belief.
Please clarify what you are explaining if not the origin of the string.
A string that is the result of a necessity mechanism has an origin, and if that string meets complexity and functionality requirements, that string has “dFSCI”.
I must admit I only realised what Gpuccio was getting at with the necessity clause yesterday. I hope I captured it in my comment to him below when I wrote:
The majority of people in our society at this time cannot conceive of even the outlines of a deterministic explanation.
On reflection I suppose deterministic explanation could also cause confusion! How about – a deterministic explanation is chain of cause and effect from some natural origin which does not involve design – such a chain may have stages which are stochastic i.e. the effect was plausible but not certain given the cause)
Really these definitions need great care to avoid barren circular arguments.
You have a very bad design detection tool and that needs to be clarified for anyone who actually thinks the ID movement has any credibility as science.
I think gpuccio (and all ID advocates) conflate agent and process. This confaltion becomes clear when we discuss whether humans creat a string via design or via a necessity process. I think this whole line of argument is rubbish.
The issue to be decided is not whether sky fairies exist or whether mysterious agents meddle in the history of life, but whether evolutionary processes are sufficient.
This is the same issue at the heart of OOL. Not whether we can determine the one true origin story, but whether regular pocesses are sufficient.
The issue in any scientific theory is not true history, but sufficiency of regular processes.
Which is why Id advocates shy away from any attempt to model evolution and evolutionary processes.
My own argument has been for some time that evolution is the only known process for creating and modifying biologically relevant information. There is simply no competing conjecture except poof. This is true even if we conceed the possibility of a designer.
Process is always relevant. The only known designers of abstract information do not poof things into existence. They build incrementally, modifying existing inventions by increment change or by horizontal transfer.
Is this why “necessity mechanisms” are ruled out?
For conscious process, read poof.
What has that got to do with building a signal generator?
The flip-flop in this analogy is used as a pattern generator which is the necessity mechanism that will build the content of a data string that we will evaluate as “dFSCI”.
gpuccio is withholding “dFSCI” because the content looks like it could be generated by a necessity mechanism and true enough, it was in our analogy.
I find this staggering hubris.
Gpuccio, perhaps you would be better off having the conversation with other ID supporters. Explain to them how to determine “design” from start to finish. Then ask them to re-do that process against other data and show their working.
As it’s you that has failed to explain in sufficient detail what the process is, it’s not other peoples fault that you are unable to express yourself in terms that others find unambiguous.
This is a comment you made to Mung a while back.
As a test, if I could come up with a “software string” below the UPB that could sel-replicate, would this qualify as an invalidation of “dFSCI”?
Hidden within GP’s declarations regarding naturalness and the requirement of consciousness is the assumption that consciousness is not a natural process. GP believes that consciousness is non-physical.
Exactly what UprightBiped would make of non-physical manipulation of information, we’ll never know, because ID advocates never discuss their mutual discrepancies.
There’s really no point in having a discussion with GP until you clear up his contentions regarding consciousness.
It’s at the heart of everything he has to say about evolution. It’s why he dismisses GAs out of hand, regardless of what they can do.
GP is not interested in whether evolutionary processes are sufficient. In fact, he is uninterested in anything that can be “reduced” to a process.
Having said that, I tend to agree that the current state of AI is such that “automated” code generation is mostly limited to parameterized variations on known algorithms. You don’t see non-trivial software applications being generated by automated processes.
GP would see this a permanent and necessary state of affairs. He simply does not believe that invention is a process that can be modeled and replicate in silicon. Or that it is a process at all.
I don’t have a reference to hand but I’m sure I read recently about a physical lab that was used by a computer program to perform experiments via automation. The results of those experiments are obviously fed back and the program creates a new hypothesis and tests it in the next set of experiments.
From what I recall it surprised it’s creators with it’s ability.
Humans are built to predict the future. When the future is impossible to predict, e.g. what will be the product of this reaction, experimentation has to take place. It seems Gpuccio et al never want to get to that stage.
Yes and I believe all IDists think that way.
You can see it in all their discussions whether origin of life, evolution, fine-tuning of the universe, ba77’s NDE’s, etc.
The odds of that are beyond the UPB! 🙂
I think you are right that experimentation is not something they want to get involved in.
I think IDists believe that the future, in a sense, has already been mapped out by the “intelligent designer” towards a specific goal.
What would happen to their world-view if their experimentation pointed to a result they weren’t ready to accept?
Exactly so. Why would the “designer” intervene over billions of years if not to move things towards a pre-defined goal. Of course the absurdity is that the “designer” created the universe yet can’t do anything more in it then that which is hidden behind “randomness”.
As whatever the outcome of the dFSCI debacle, even if Gpuccio explains it 100%, that won’t say a thing about the “origin” of that dFSCI just like Upright is silent on the actual origin of his “codes”.
You’re the gift that keeps giving Mung.
If state educators want a good example of how allowing students to hear “both sides of the debate” works, UD is a good model. 🙂
keiths:
gpuccio:
No, you are not saying that. Read what you wrote:
It’s good that you recognize that your statement is ridiculous, and that you now wish to retract it, but be honest about it. Don’t pretend you were saying something else all along.
The evidence for their existence is literally trillions of times stronger than the evidence for your hypothesis.
Gpuccio 651
Because yours are unclear to me. So I try rewording them in my own language and playing them back to you. It is a perfectly standard way of checking understanding.
I could dispute the difference between probabilistic and deterministic but it makes little difference to my argument so I will go with your versions of A and B. .
Ah the difference between origin and explanation. This is key. It is another thing which is not as clear as you might think and deserves some detailed analysis. To me an explanation is an origin plus the mechanism that goes from the origin to the outcome. For example, an explanation of a sonnet appearing on my screen might be you copying (the mechanism) the sonnet from a book (the origin). Neither part by itself counts as an explanation. If you disagree please say so and give an example.
Both origins and mechanisms can either be designed or not. For me a “necessity mechanism” is just a mechanism with no element of design. So this defines four types of explanation:
a) Designed mechanism, designed origin
b) Designed mechanism, non-designed origin
c) Non-designed mechanism, designed origin (the sonnet got accidentally copied when it fell on the copier)
d) Non-designed mechanism, non-designed origin
I believe d) corresponds to a deterministic explanation.
I think this analysis makes it clear that it is false that everything with dFSCI has a non-designed origin. There are plenty of cases of b) which I think you would recognise as having dFSCI. For example any kind of artificial selection process such as the Weasel programme starts with a random origin. If you want to make it even remotely plausible that everything with dFSCI is designed then you must include a, b and c as cases of design. So really what you are talking about is whether there are any cases of dFSCI which are known to be d) i.e. deterministic explanations. You claim there are none. But the definition of dFSCI is no known deterministic explanations i.e. d)
The “no known necessity mechanism” clause just amounts to “don’t infer design if a perfectly good non-design explanation is known.”
Gpuccio’s error is in assuming that if a non-design explanation isn’t currently known (to his satisfaction), then none exists and it is safe to infer design.
I can’t get interested i the debate about design until the process of design is illuminated.
Poof is not a process.
Let’s see someone in the design community discuss a plausible process.
No, because a self-replicator below the UPB means kairosfocus can no longer claim OOL is not possible without a designer.
It’s pretty remarkable that after all this time you still do not understand the reason your side came up with the UPB.
Since ID approaches the problem from a UPB viewpoint, a replicator below the UPB proves that “random search” is capable of solving kairosfocus’s replicator challenge.
Ask kairosfocus about the significance of a self-replicator if you don’t understand how important it is to the ID position.
gpuccio, kairosfocus, Mung,
Step 1: A self-replicator springs into existence whose total information content is below the UPB, thus making it likely that it could “randomly” come together without the conscious efforts of an intelligent designer.
The UPB is thus not a barrier to such a simple object.
Step 2: Imperfect replications that have more information, less information, wrong information or any combination of the three.
Step 3: The possibility that imperfect replication will result in a growth of information beyond the UPB, that is functional, complex and unforeseen.
Condition 1: No UPB barriers will be broken in any single parent/child gain in “dFSCI” information.
Strings can be generated by this process that would pass gpuccio’s “dFSCI” test.
If I can design such a string whose length is below the UPB, your side can not claim random processes cannot result in its configuration also.
Mung wrote (quoting me):
But all of the above are genetic algorithms of one sort or another. There is thus no issue of whether or not GAs resemble GAs. And what I don’t know what gpuccio would say is a model of evolution. And neither do you, as gpuccio has not presented any model.
It already is an example of the GA. It implements the Wright-Fisher model of population genetic processes. And that is a GA, as you must know if you understand these things.
No, I did’t realize this, and still don’t realize it. The models I mentioned, all of which are GAs, are modeling evolutionary processes. Then we see what they do, and we compare it to any mathematical theory that we have for that case. My one-locus teaching program PopG, for example, is for students to find out what is the outcome of particular combinations of evolutionary forces, ones that they invent. There is no need for a “goal”.
The reason I brought up my (decades of) experience with models of evolution and with GAs was that I was regularly being lectured by the UD posters on these subjects. I know how you model evolution, and I have experience with GAs.
It is less clear from Mung’s comments that Mung knows what is and is not a GA, and it is unclear what gpuccio would do about modeling evolution, as gpuccio has presented no such models.
Gpuccio 690
You are right that the relationship between explanation, origin and mechanism needs deeper thought. It is crucial to your dFSCI argument.
I am happy to go with whatever definitions you want to use, provided they are clear. But right at the moment I am bit confused. Please don’t answer “the ordinary English sense of the words”. They are extremely broad in use and susceptible to many interpretations. (In the following a “design explanation” is any explanation with involves an element of design and “entails” means logically entails according to your definition, not empirically happens to be true)
Here are some questions that might help elucidate the meaning.
Joe, I think you intended this comment in the gpuccio thread so I moved it here.
Oops, yes I did. Many thanks.
No one has said that.
In the following, kairosfocus clearly shows that any argument in these discussions must address physics and chemistry and cannot simply be fought with virtual “information”.
Thanks for drawing attention to the fact that theories such as gpuccio’s and Upright BiPed’s are not acceptable as arguments against evolution, unless they address actual biological models, instead of assumptions, probabilities and “information”.
Kairosfocus wrote: F/N: JF needs to recognise that GA’s START inside islands of function, and so at best explain some form of microevolutionary adaptation to niches, they do not explain origin of body plans starting with the first. That is, the fitness function has a generally uphill pointing slope of course with room for various local peaks etc. But the problem is that the vast majority of reasonable config spaces for multi component systems depending on well matched and organised parts to work, will be flat zero, non functional. There is no slope info to get one pointed right with a blind process. We don’t even have the sort of neutral drift possibility in an already working entity to play with until we hit an island. And, given the complexity involved for FSCO/I, drifting is equivalent to blind random walks in the full config space. Which gets you nowhere. KF
That is a succinct statement of how Michael Behe’s argument might appear in the context of GAs, but it does not address my point.
The issue is whether there is some way of proving the assertion that CSI cannot be out into the genome by natural selection. Note the “cannot”. The assertion that it cannot be put into the genome by natural selection is regularly made by posters and commenters at UD. And yet no one has a reasonably flexible mathematical model of evolution for which it can be proven that CSI cannot be put into the genome by natural selection.
I have posted here, last March a simple model in which it can be seen that functional (or specified) information is put into the genome by natural selection. It was for a 100-locus genome, so it does not get enough SI into the genome to constitute CSI, but if you simply take the same model and give it 1000 loci, it is obvious that it will show that natural selection will put CSI into the genome.
That was a counterexample. I do not need to remind readers that one counterexample is enough to refute a proposed generality. A similar genetic algorithm, with 1000 loci and the same fitness scheme, would also refute the assertion that natural selection cannot put CSI into a genome. If the proposed general rule that natural selection cannot put CSI into a genome were true, my example would not work, but it does work.
I am a bit surprised that Professor Felsenstein’s OP back in March generated so little attention. I am ashamed to admit I can’t recall reading it! I have now done so and need to read it again slowly.
I suggest mung does the same so he does not waste too much energy attacking straw men! 🙂
Gpuccio then:
Gpuccio, now:
Gpuccio,
You’ve backed off from your earlier extreme statement — “the only realistic model of NS is to attribute to it no role at all” — and substituted a more moderate one — “the only model he can really build at present is to give it no rile at all.” Your later statement is an improvement on the first, but it is still wrong.
You wrote:
We know that natural selection happens, so giving it no role would be foolish and unrealistic. Most IDers (including you, I thought) acknowledge that natural selection happens — they just maintain that it is far less powerful than evolutionists think it is.
The right way to model NS is exactly the way Joe does. He includes fitness parameters in the equations. By substituting appropriate values for the fitness parameters, you can model situations in which there are selectable intermediates as well as those in which there aren’t.
There is both direct evidence (e.g. Szostak, Lenski and others) and indirect evidence (see my next comment) for the existence of selectable intermediates. As far as I can see, you dismiss this evidence wholesale simply because it leads to an undesired conclusion. That’s bad science, gpuccio.
P.S. Does your browser have an English spell checker?
keiths:
gpuccio:
Because it is directly relevant. We have two competing hypotheses, and the argument laid out in that thread shows why one is vastly superior to the other.
According to one hypothesis, selectable intermediates exist and unguided evolution makes its way from one to the next. If this hypothesis is correct, then life should form an objective nested hierarchy (except among organisms such as the prokaryotes for whom horizontal transfers are common). This is exactly what we find. The objective nested hierarchy exists (out of trillions of other possibilities), and the case for the existence of selectable intermediates is enormously strengthened.
According to the other hypothesis, selectable intermediates don’t exist and evolution is stranded on “islands of function.” If this hypothesis were true, we would not expect to find an objective nested hierarchy. Yet we know the objective nested hierarchy exists.
The only way to make the second hypothesis fit the data is 1) to assume that a designer exists who can bridge the gaps where there are no selectable intermediates, and 2) to assume that this designer acts exactly as unguided evolution would have if there had been selectable intermediates! As I put it in my thread on common descent:
It takes three ridiculous and unwarranted assumptions to make the designer hypothesis fit the evidence. By contrast, no ad hoc assumptions are required to make the other hypothesis fit the evidence.
The evidence of the objective nested hierarchy supports the existence of selectable intermediates trillions of times better than it supports your hypothesis, in which selectable intermediates are absent and a designer is required to bridge the gaps.
I’m sorry, I should be more solemn and respectful, but it is really funny to see the UD commenters bad-mouthing natural selection in widely-discrepant and off-the-wall ways:
Mung: Natural selection doesn’t put things into the genome. It is incapable of putting things into the genome. It is not a “creative” force. Frankly, it’s not even proper to speak of natural selection increasing or decreasing the frequency of some trait. The actual increase or decrease in frequency is caused by differential reproduction, not natural selection.
Eric Anderson: Natural selection is just a convenient label — a convenient placeholder for our current state of ignorance. It doesn’t do anything. …. it is nevertheless still quite true that in most cases references to natural selection operate as useless circular tautologies.
gpuccio: NS is a process derived from the information that makes living beings which reproduce and have metabolism possible. Being a consequence of that information, it can do very little beyond the scope of the information that already exists. Microevolution is all that can happen by RV + NS.
Joe: And whatever you call it, it still doesn’t do anything. So why not just call it snake oil?
And Mung insists that “natural selection can only increase or decrease the frequency in a population of some already existing property”.
(Mung seems not to be noting that increasing the frequency of some genotypes at the expense of others, one can move the population further out into the tail of a distribution of fitness, and Mung also seems not to realize that by increasing the frequencies of rare alleles one can give rise to new combinations of those alleles that did not exist before.)
There are interesting discrepancies here: it either shouldn’t be called NS, or it can be. It does nothing, or it is (somehow) known that it can only do so much. It is a convenient label, or it is snake oil.
But the common thread is that some sort of incantation is invoked to just make it go away.
All this avoids the real question: what proof do the UD commenters have that natural selection cannot result in Complex Specified Information being in the genome? None, in fact — they have no model, no simulation to show that. For that they need to show that in cases where we do not know the detailed fitnesses and the detailed mutational events, that it is in principle impossible to get CSI into the genome.
No such proof exists. And yet there are steps in the Design Inference flowchart that require such a proof, otherwise one cannot infer Design.
Gpuccio 706
Fine with this … we agree on something!
I have some more detailed responses below – but maybe I have learned enough to ask this crucial question.
Consider these statements:
Are they true?
No need to respond to the items below – unless they help you understand what I mean!
OK. I see I need to modify it a bit. If we are trying to explain the configuration of a digital string then does an explanation entail describing the origin?
I find this confusing. It seems to me that the airplane was the origin! I am unsure how to recognise an origin when I see one. I think maybe the business of inferences being right or wrong is confusing the issue. An explanation may be right or wrong and it may be right or wrong about the origin but if it is a design explanation can it avoid making a statement about the origin (perhaps wrongly). Perhaps I should I have phrased my question thus:
Does a design explanation entail an assertion about a designed origin?
OK. I think that will do.
Time once again for my favourite quote, from Graham Bell’s Masterpiece of Nature: “Natural selection is a simple theory because it can be understood by everybody; to misunderstand it requires special training.”
Mung is particularly priceless:
it’s not even proper to speak of natural selection increasing or decreasing the frequency of some trait. The actual increase or decrease in frequency is caused by differential reproduction, not natural selection.
Like saying that a breeder does not alter the frequency of favoured and disfavoured characters in his stock – that is instead attributable to their differential reproduction! Like NS, AS presumably ‘does nothing’.
30 bits is only 1 in a billion.
If Axe is lowering the threshold because of chemistry/physics, then no improbability arguments from the ID side are acceptable that don’t do the same.
I hope you understand what this means for your argument and also Upright BiPed’s.
Your “dFSCI” and UBP’s semiotic codes are no longer arbitrary and cannot be considered improbable purely from a statistical perspective, if actual chemistry is involved.
The argument I am putting forward for a self-replicator however, is purely statistical like yours, Dembski’s and UPB’s.
It is not a trick of any kind.
Please try to understand what it means when the ID side puts forward probability arguments based on pure arbitrary randomness.
Any bit configuration under a certain probability threshold has a possibility of existing without the aid of a conscious designer, according to ID proponents themselves.
If I can design a 7 bit string of bits that represent the ASCII letter ‘A’, that configuration is very likely to occur randomly.
Show me where the “trick” is there.
Any string under a threshold X, has a possibility of occurring 1 in X configurations, regardless of whether that string is identical to some designed string.
That is the whole point of Dembski’s and other IDist’s improbablity arguments, that something above a certain threshold needs design, not a configuration below that threshold.
If I design a self-replicator below the UPB, that string has a possibility of existing due to random processes.
I intend to generate a significant number of bits above my threshold.
I continue to be amused by the fact that IDists want the respectability of mathematics, but deny the one mathematical route to modelling the process they wish to discredit.
This is what puzzles me. How do you know a string is not a random string that just happens to coincide with a designed string? A random DNA sequence can stiil be translated into a protein. How do we know without actually doing the synthesis and checking the resultant protein for properties and functions?
Can gpuccio distinguish between a DNA sequence that codes for a functional protein from a sequence just pulled from the air? Of course that arbitrary sequence could by chance code for a protein with some activity! I’ve asked him and am interested to see the answer.
Gpuccio 709
If something is logically possible (even if empirically false) it should be possible to imagine it. So please could you describe what it would be like to have:
(You quote the example of the cloud that was designed by an airplane a few minutes ago. Surely the airplane creating the cloud is the origin? In any case that is not the configuration of digital string).
Thanks