29th Oct: I have offered a response to Gpuccio’s challenge below.
I think this is worth a new post.
Gpuccio has issued a challenge here and here. I have repeated the essential text below. Others may wish to try it and/or seek their own clarifications. Could be interesting. Something tells me that it is not going to end up in a clear cut result. But it may clarify the deeply confusing world of dFSCI.
Challenge:
Give me any number of strings of which you know for certain the origin. I will assess dFSCI in my way. If I give you a false positive, I lose. I will accept strings of a predetermined length (we can decide), so that at least the search space is fixed.
Conditions:
a) I would say binary strings of 500bits. Or language strings of 150 characters. Or decimal strings of 150 digits. Something like that. Even a mix of them would be fine.
No problem with that.
b) I will literally apply my procedure. If I cannot easily see any function for the string, I will not go on in the evaluation, and I will not infer design. If you, or any other, wants to submit strings whose function you know, you are free to tell me what the function is, and I will evaluate it thoroughly.
That’s OK. I will supply the function in each case. I note that when I tried to define function precisely you said that the function can be anything the observer wishes provided it is objectively defined, so, for example, “adds up to 1000” would be a function. So I don’t think that’s going to be an issue!
c) I will be cautious, and I will not infer design if I have doubts about any of the points in the procedure.
I am not happy with this. If the string meets the criteria for dFSCI you should be able to able to infer design. You can’t pick and choose when to apply it. At the very least you must prove that the string does not have dFSCI if you are going to avoid inferring design.
d) Ah, and please don’t submit strings outputted by an algorithm, unless you are ready to consider them as designed if the algorithm is more than 150 bits long. We should anyway agree, before we start, on which type of system and what time span we are testing.
I don’t understand this – a necessity system for generating digital strings can always be expressed as an algorithm e.g Fibonacci series. Otherwise it is just a copy of the string. Also it is not clear how to define how many bits long an algorithm is. Maybe it will suffice if I confine myself to algorithms that can be expressed mathematically in less than 20 symbols?
And anyway, I am afraid we have to wait next week for the test. My time is almost finished.
That’s OK. I need time to think anyway. But also I need to get your clarification on b,c and d.
Mark, I think you should get some understanding and compromise on the following statement from gpuccio for this challenge.
The whole point of “dFSCI” and all other versions of it, are to compare “a designer’s capabilities” versus “nature’s capabilities”.
By admitting that nature is excluded from the test makes no sense if that is what gpuccio is comparing to a “designer”.
The tilt of the Earth back and forth is very regular but generates our seasons and some very complex weather patterns that are very unpredictable.
Maybe off-topic but relates to specification of proteins. Seems functionality is not so rare as some would have us believe!
http://www.nature.com/nature/journal/vaop/ncurrent/full/nature11500.html
H/T SteveF
I’m more interested in the process of analyzing a string than in the result. I would like to see the universal method that applies to any string.
Yes. The question isn’t “how often can gpuccio guess the origin of a string?” It’s “does gpuccio have an objective, reliable method for identifying designed strings?”
A secret methodology would be unusual in science.
(d) is problematic for at least two reasons. First, it means that gpuccio’s claim to be able to determine whether or not dFSCI is present from just a string and its function is inaccurate. He also needs to know the source of the string, which makes the whole exercise unimpressive at best.
Second, this clause allows gpuccio to dismiss strings generated by GAs, despite the fact that GAs model known evolutionary mechanisms, because the GA implementation is designed.
I agree with keiths and go further: the challenge would be unnecessary if gpuccio had an objective, reliable method for identifying design. Clause (c) also makes it pretty clear that he doesn’t.
Well he has a string now. He can walk us step by step through his objective analysis.
Defining the terms of this challenge is not straightforward. If I decide on a function and then find a short algorithm to create a string that performs it – have I designed it or not? It is after all a deterministic process.
Todd Wood’s take
Why is gpuccio putting these restrictions on his challenge? Any protocol for measuring stuff has its operational limits; to cite the first example that comes to mind, C14-based radiometric dating doesn’t work so good on specimens that are 50,000+ years old. So if gpuccio truly does have a valid protocol for determining dFSCI, it makes sense that said protocol would have operational limits. So rather than “don’t bother sending me a string for which X is true”, shouldn’t it be more like “my dFSCI-determining protocol has thus-and-so operational limits, and it cannot be relied upon outside of those operational limits”?
In fact, it’d be even better if it were “go ahead, send me any string you please! any false positives my protocol yields in its current form, will help me refine its accuracy by correcting for whatever threw it off in those cases”. And it would be better still if gpuccio published his dFCSI-determining protocol in its entirety, so that any-damn-body who felt like it could use that protocol to determine the dFSCI of whatever strings they liked, and different people could use it on the same string to see if they get the same results, and…
NS is an algorithmic process – or can be written as such to deal with digital strings. But I’d doubt it can be written in less than ‘150 bits’, whatever a 150-bit program means – each instruction invokes a set of algorithmic actions longer than the instruction. It does not happen at all without the operating system, compiler/interpreter etc wrapped around it. (Nor could we generate strings without some kind of brain attached to the other end of our fingers).
So apparently the only ‘dFCSI’***-generating mechanism we would propose is excluded from the challenge.
(*** I prefer that to dFSCI, for the tiny joke it contains)
I decided to remove this comment temporarily.
All I will say about my string is that it is a subset of a highly useful and lucerative set of strings. What I want to see is gpuccio’s methodology for determining whether the string can be produced by necessity mechanisms.
I might say that one of the folks often cited by ID advocates –Hubert Yockey–is on record saying evolution can “compute” any string.
Couple of problems.
First the defined function. Nylonase catalyzes byproducts of nylon manufacture. This is a novel function. Gpuccio would probably redefine this as a general catalyst that evolved from another general catalyst. In other words, the use of the term function is ambiguous. He can and will change it to suit his conclusions.
Another problem is that he defines the dFSCI as the functional complexity of the generating algorithm or the sequence itself, whichever is least. Protein folding is extremely difficult to model, while it happens readily in nature (due to what we might call parallel computing, each atom doing its own thing). Essentially, the three-dimensional space and the properties of the molecular relationships provides information to the evolving population. If we were to model this in silico, gpuccio would say that we need to include the size of the model in our calculations. This is odd, to say the least.
I anticipate some criticism of the function – although I am not sure on what grounds as it conforms to his definition.
I can’t see any scope for taking into account the generating function when calculating complexity as the whole point is that this is meant to be a process that identifies design without knowing the generating function.
According to gpuccio, it is “dFSCI” that doesn’t care about the generating function.
When you go to attribute “design” to a string however, the generating function is then taken into consideration.
That’s where his terms “False Positive”, “False Negative” etc., come into play.
In short, he does not claim that you can know if a string is designed simply by looking at the string itself.
I think there’s a major problem quantifying information in genomes.
The article I quoted indicates that 75 percent of bases are noise — any value is equivalent to any other value. There are also numerous synonyms and some interesting change of function mutations available.
When you are counting the exponent of string length as your quantity of information, these are significant increases in the number of needles in the haystack.
Much more significant is the fact that these synonyms are available as point mutations, which means that the possibility of incremental change is confirmed in spades. Other recent studies have confirmed that change of function is possible under continuous selection.
So the search space to be considered is not the number of possible strings of length x; it is significantly less than the length of string x.
Amazingly, even Mung agrees that the search space might be less than 2**X in an analogy he gave where one bit in a string is directly related to another bit in that string, thus showing that “information” in a string is not completely “arbitrary”.
I eagerly await his objective method of detecting design that does not involve first calculating dFSCI. Isn’t that the part where the argument goes circular?
1. I deny his ability to calculate the number of bits necessary to provide a given function
2. I deny that he can determine how many bits have been modified either by evolution or by design from “noise.”
3. l I deny he can determine the history of a DNA sequence from its current configuration.
4. I deny he can set an objective standard for how many bits can be changed by known evolutionary mechanisms.
He is free to publish his objective methodology for determining any of the above. If his methods cannot be replicated by any objective experimenter, regardless of politics or religion. then he has nothing.
The issue is, no ID proponent has formulated a “principle of design” that can be identified in such a way that allows it to be repeatedly and reliably applied, in practice,
When ID proponents claim to have done this via induction, we need not address their claims directly because the idea that we actually use induction, in practice, is a myth. It’s an appeal to inductivism.
Specifically, no one has formulated a “principle of induction” that can be identified in such a way that allows it to be repeatedly and reliably applied, in practice, either. Popper has addressed the issue at length. We cannot extrapolate observations without first putting them into an explanatory framework. ID would not be an exception.
What ID needs to provide is clear, reliable guidance at every step of the way based on empirical observations. But it doesn’t. This is because induction does not provide is clear, reliable guidance at every step of the way based on empirical observations, either.
For example, proponents may think they have solved the problem of induction by directly receiving revelation from an infallible source. However,
But the idea that there is such a thing as an infallible source – and that the information it imparted was without error – is a explanatory framework. And it’s a bad explanation because it’s easily varied.
Nothing is necessary for ID’s designer because it is abstract and has no defined limitations. For example, it does not necessary predict a nested hierarchy. Any such assumption on the part of ID proponents is a smuggled bad explanatory theory, such as we were designed in the designer’s image, etc.
So, what Gpuccio needs to do is provide is clear, reliable guidance at every step of the way based on empirical observations. If he can do this for ID then he would also be doing the same for induction.
This would be quite an accomplishment from a epistemological perspective.
Yes.
I think the circularity starts with that SC in “dFSCI”.
They claim “functions” are “specifically intended” simply due to their description of that function!
While a function may be “specific”, i.e. one of X, it is not “specified” as in, “The designer intended this functionality in the “specification” of his design.
This is circular as it implies, “What I received must be what I asked for”.
critical rationalist:
Have you established that we never use induction? If so, how did you do it without reasoning from particular observations to the general? (As this is off topic, I’ll take it to the Sandbox for anyone who would like to discuss induction).
Dr who: Have you established that we never use induction? If so, how did you do it without reasoning from particular observations to the general?
Are you familiar with Popper’s work?
To highly simplify, not only is induction impossible in regards to certainty, but it is impossible in regards to probabilities with the exception of very limited cases. This is in contrast to deduction, which does give us a form of certainty.
So induction is undesirable in all but very specific and limited cases. Developing theories is not one of those cases.
Popper “solved” the problem of induction, in practice, by pointing out that deduction is preferable in nearly every case.
IOW, Critical Rationalism does not take the form of reasoning from observations to the general. It too is a conjectured explanation for the growth of knowledge, which is tested by observations, not derived from them.
Popper is often misunderstood because he did not deny that we could make progress. Rather, he pointed out that we merely thought we were using induction while doing it.
The contents of theories are not limited by past observations. They are limited by whether they solve a specific problem.
It’s unclear how this is off topic.
ID proponents argue that we have never observed knowledge (dFSCI) being created by anything but a designer. Therefore, a designer must have created it. Also, they portray Darwinism as one astronomically unlikely event after another. And they claim to have reached these conclusions based on observations alone.
But this is impossible because we cannot extrapolate observations without first putting them in an explanatory theory. So, the question isn’t *if* ID proponents are using an explanatory theory to reach their conclusions. Rather, the question is, *which* theory? Also, is it a good explanation in that it actually solves a problem? Is it hard to vary etc.?
IOW, claims that they are simply “following the evidence” are false because that is simply not possible.
Of course, just because no one has yet to refute Popper’s criticism of induction, this does not mean I am not open the possibility that it might be improved or even corrected. Deutsch has already improved Popper, as has Bartley, etc.
I agree with that. I’ll caution you, though, that you won’t persuade many people. Belief in induction seems to be deeply entrenched.
I’ll disagree about Popper. I never found him persuasive on induction.
Induction is a placeholder for whatever it is we do when inferring patterns. I would like to call it Markov inference, the summing of all dimensions of perception to form a best fit pattern. It is not a logical deduction or logical inference. It is a perception. It is seeing something new as like something known.
Mathematical induction is quite rigorous, but when we use “inductive reasoning” we mostly seem to mean Bayesian-style inference (particularly if very informal) or inferring proposition P to be true because we haven’t seen a counterexample yet. For example we might reason by induction that “No Cadillacs are SUVs“. Being that kind of inference, in many cases it is not particularly convincing.
When unsure, I’d rather stick to statistical inference, where at least we know roughly how unsure we are.
Agreed. I see mathematical induction as deductive, not inductive. I see statistical inference as deductive, not as inductive. I make that distinction because statistical inference yields only probabilities, while induction is alleged to yield truth.
Bayesian inference, if properly done, should be a form of statistical inference. Unfortunately, the expression “Bayesian inference” is tossed around quite wildly, with hugely extravagant claims often being made for it.
Gpuccio
I am delighted to see you are back. I have no problem not using GAs. I thought you meant algorithms in general.
So here are the three strings I would like to apply your procedure to:
206768813885305238342364780672957679300082009167738871270
412280557510461117341374275269279462741125744191516133118
2310050193947079811998518064317299823998576164608823460658
The function of each one is that they represent an ordered list of papers in the PubMed database. They do this by listing the PMID of the papers. Each string represents a specific set of papers so I believe there are very few digital strings of the same length that could perform that function (there might be a few alternatives by adding leading zeros in some cases). So I think they are clearly complex by your criterion.
To help you identify which papers here are the same strings with added blanks to separate out the individual PMIDs.
20676881 3885305 238342 364780 672957 679300 082009 167738 871270
4122805 575104 611173 413742 752692 794627 411257 441915 161331 18
23100501 939470 7981 19985 18064317 29982 39985 7616460 882 3460658
So evolution is excluded by definition?
I also wonder how one goes about designing a sequence if you have to know it is functional before you design it.
So how can he place conditions on how the string is generated?
Gpuccio
You write:
Why should they have anything in common? You say a function can be anything that is objectively defined by the observer. My objective definition is the list of papers I give for each string. Maybe you need to be a bit more precise about what is an acceptable function?
gpuccio:
The string is not a phrase. It is a list of neologisms mimicking several different languages. The function is the generation of potential trade names. the “words” must be pronounceable, but not in current use. This has commercial value, most notably in the pharmaceutical industry, but also in other industries.
There is no way to “weasel” this, because the specification requires that the output be currently unknown.
Gpuccio
A week or two ago I suggested that you needed to define what you mean by “functional”. I believe I suggested either “suits someone’s purpose” or “plays a role in a system”. I was then going to go on and argue that this was a source of circularity in dFSCI!. You cut this short by emphasising that this was unnecessary. Function was just something the observer objectively specified. Now you seem to be backtracking (and are in danger of reigniting the circularity issue!). Your comment suggests that there are two ways a function can be acceptable.
1) The function may be specified before the digital string was generated. However, dFSCI is meant to be a procedure for detecting design that works given a string and a function. I don’t recall reading anything about needing to know when the function was specified. How is the person doing the checking to know if the function was defined before or after the string was generated?
2) You describe an alternative acceptable “true functional specification” thus:
There seem to be two ingredients here:
* Objective – well surely an ordered set of papers is completely objective?
* Independent of the string – the function is to list a set of papers in order. This list of papers is independent of the string. The PMID is just one of many ways of referring to them. I could have conveyed the list without knowing anything about PMIDs by giving you the titles of papers or if I was close to you by giving you physical copies. It is no more dependent on the string than a list of cities is dependent on their coordinates.
The problem you have here is that a digital string can be used for an infinite variety of things. I am sure with a little imagination and time I could come up with other functions for these strings – e.g. mathematically they will each have several unique properties.
If you want dFSCI to be a usable procedure you need a clearer way of defining what is an acceptable function.
I’m defining a functional string as one that someone will pay money for. I can cite historical precedent for money being paid for lists of pronounceable neologisms.
To be worth money, a list needs to be a significant pruning of the space that includes all possible strings.
Gpuccio
I thought we were talking about developing a usable procedure here. How is the observer to know that the function was specified in advance? By evaluating the function? In that case let’s have the process for doing this.
PS I still don’t understand why a list of papers is not an objective definition of those papers. Is it subjective?
I ask you again. How do you know? If it was by inspecting the string then please give the process.
As it happens I did do that for one of the three strings. Want to tell me which one?
I didn’t design the function. I identified it from all the many things that could be done with that string. In fact the exact process was I took parts of the string and entered them into Google to see what they might be used for. I started with the whole string and progressively broke it down into smaller parts. It only took about 20 minutes. The function of representing that list of papers was a property of that string even if I had never engaged in the search. I expect there are very many other such functions should I stumble across them.
Sure I identified the function after the string was created. That’s pretty much how evolution works. A mutation takes place and if there is a function it can fulfil it does it.
Gpuccio – I am sorry I am still really confused by the process. I did give you the list of papers (I am wishing now I had given you the titles). Initially I did not explain how the list was found but rather quickly fell to explaining that. Supposing I had not done so. So now you are presented with
I thought this was sufficient to decide that dFSCI is present. Are you now saying that you also need to know something about how the function was arrived at?
Petruska: It is seeing something new as like something known.
In what way were electrons “like” something known? We had yet to “observe” them, even in an indirect manner, so it’s unclear how they were “seen” at all in the first place.
The contents of theories are not constrained by past observations, but whether they solve specific problems in a way that is compatible with the rest of our best, current explanations.
What we conjecture are explanations as to how the world *works*, which isn’t necessary like anything in the past. It’s a guess about what is happening *in reality*.
Teehee. Gpuccio doesn’t know what “weasel” means, in this context.
Gpuccio – since you’re reading this thread, why under god’s blue heaven are you not over here posting your responses/questions, instead of hiding at UD where everyone is banned?
Get your courage together, gpuccio, and show some respect for the dialog that you started by having the dialog where everyone can participate without censorship.
What are you afraid of?
Electrons are like particles and like waves. They are both and neither. Most ideas are metaphors and most language is metaphorical.
Einstein’s thought experiments involve concrete objects: elevators, trains, and so forth. Only after having an image do we begin developing the math.
My problem with gpuccio’s insistence on “someone” knowing the source or history of the string is that this is precisely what we are trying to find out. The number of bits is not interesting in and of itself.
If GP’s method doesn’t tell us anything about the history of the string, what is its value?
Gpuccion 203
My function is “refers to these papers”. You say this is ambiguous because that list of papers could have arisen in two different ways. I have to say I don’t see that at all. If I wanted to use the string to indicate the list of papers to someone else it would work perfectly well however I derived the list. But let me temporarily accept that it is ambiguous.
Well of course you could ask that question of any function e.g. refers to papers about CHD, is the first line of the Lord’s Prayer, whatever. Any of them might have been defined independently of the string or derived from the string. I might have studied the papers until I could find something in common and noticed they were all associated with CHD, or I might have decided that CHD was interesting and tried to generate a string that referred to CHD papers. So it appears that all functions are ambiguous unless you know how they are derived! Or is there something about some functions that means you don’t have to know how they were derived to use them for dFSCI while you do have to know for my functions?
The rules for defining an acceptable function for dFSCI appear to be more complicated than they first appeared!
Gpuccio 207
This all turns on the word “explicit”. Obviously you could go on asking for more and more detail about a function indefinitely. If the dFSCI procedure is to be clear than it needs to be clear about when a function has been explicitly defined. It can’t just be the tester’s opinion that they would like more information.
In particular why are my functions not explicit until I give the way they were derived while other functions such as “lists papers relating to CHD” or “is the first line of the Lord’s Prayer” seem not to require knowing how they were derived.
Let’s make the process clear.
Gpuccio
Actually rereading your 207 I grow even more confused. At the beginning of the comment you write:
Later on you write:
So is my function explicit or not?
I have not responded as to how my string was created. I only discussed its function. I am not interested in whether strings arr designed, but in whether you can tell anything a out their history or about the process by which they came about.
The question you raise regarding DNA coding sequences is whether they have subcomponets that could accrue incrementally. This is the only question of interest. I’m interested in seeing your method for determining this.
gpuccio is not demonstrating that “dFSCI” can “detect design” in this challenge.
What he believes he is taking part in is a joint effort by all of us to come with with “examples” of “known” design and non-design.
Because he believes we are really just co-authors in analogies that show examples of “dFSCI” being analyzed with known examples, any string you come up with that has unknown origins or questionable functionality is not acceptable as an example showcasing “design detection” at work.
Never has he ever said in these examples, that design can be detected in the strings, simply that these are known test cases.
What this means is that “dFSCI” cannot tell us if the “information” in DNA was designed.
It simply shows that “information” in DNA could not have been the result of purely random processes based on the improbability of the “dFSCI” in the string.
Q: Is this string designed?
A: Yes.
Result: The string is designed.