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.
In the full context of the debate over “unguided” evolution vs design, this makes no sense.
If he cannot say anything about the history of a sequence or the process by which it was created, his discussion is vacuous.
For the record, every single character in my string was the result of the most nearly random process available to me. If I had access to a bit stream created by radioactive decay, I would have used it. It would have made no difference.
Here is another 2000 character string:
It’s specification is “designed to be incompressible”.
So UDites, is it designed? Does it have dFSCI? If so, how much?
By the way, my previous challenge had one of the two strings as a totally random set, the other not so much. No UDite could even determine that.
Gpuccio
You think things are simple when actually you are glossing over fundamental problems. Let me try it another way. Suppose I had supplied the strings and the functions and then no longer been available for further questions – maybe I died of frustration in the interim :-). So you don’t know if the papers were:
a) prespecified
b) all had something in common you were not aware of e.g. they were on my desk in that order
c) were post specified as I explained
Has the string got dFSCI or not?
To be fair to Gpuccio you need to supply the function as well. He has always said that it needed an objective and explicit function (although apparently referring to an exact list of papers is not explicit enough!)
Yes, “dFSCI” is a step that is not even required for gpuccio’s argument.
For instance, I might find a watch in the middle of the desert, take it back to my lab and calculate its “dFSCI”.
I seal my “dFSCI” findings in an envelope and hand it to you, along with the watch, and an explanation of what specifically is the function of a watch.
You take all this into consideration and then you state, “This device was designed with a conscious willful intent by an intelligent agent”.
I react with shock as I blurt out, “But you never even opened the envelope!” 🙂
Gpuccio wants to know that a string is ot the product of a GA.
I will assert with confidence that neither gpuccio nor anyone else at UD, or anyone in the ID movement can write a GA that can do what I did to produce my string, neither a working program nor a specification in pseudocode.
This is because none of them understand what evolution does.
Specifically they lack the imagination required to understand how mutation and selection can chain 10 random mutations in just a few generations to form a functional string.
They couldn’t understand how Weasel worked, and they certainly can’t understand how mutation and selection can produce completely novel functional strings. Behe’s limit of two or three linked mutations is complete rubbish. He also lacks the imagination to see how it can happen.
Gpuccio
I have to tell you at this stage I think my examples show that the process of deciding whether something has dFSCI is unclear – particularly the process for choosing a function. This is how it comes over to me.
There are also ambiguities with the assumptions you make when calculating complexity (uniform probability distribution and all that) but with the Noddy examples we are using that it is not an issue. So let’s not get into that as well.
You see there are two places where I find the process undefined. Let me back that up with examples.
You may think that all this only arises when the function is chosen because it is something the string can perform and so it is a bit a sideshow. However, the process I used is close to how evolution works. Mutation creates a gene or protein and then if there is any function it can perform that adds to the organism’s fitness it is preserved. Using your approach the choice of function for the calculation of dFSCI for a protein should be “can make any contribution to the organism’s fitness” or something like that.
JoeOctober 30, 2012 at 9:17 am
If gpuccio agrees with Joe then I guess the matter is settled! We know something is designed because everything is designed! Joe and Sherlock Holmes say so!
I think I am going into “watchful waiting mode” in the doubtful expectation that, one day there may be something useful that will eventually emerge from the ID camp.
Eugene S writes:
Yes. Szostak et al have defined function information in a biological context. They make it very clear that this number is relative to a particular function and do not make any attempt to correlate this with design. I question the way they equate “proportion of possible configurations” with “probability” but otherwise it seems a reasonable concept. However, they do not go near the question which I am raising which is how you choose the function. My problem is that if I choose the function the way I thought was perfectly OK then a high level of functionally specified information can occur in systems which not designed.
The function is to provide error correction. If a character flips it is more likely that the overall string size when compressed via gzip will get smaller rather then larger. Something like that anyway.
I agree but then have to wonder why you would allow UD to hinder our discussion by banning most of us.
Why would you support UD attempts to stop discussion between our sides by only taking part in this “public discussion” on the site where most of us have been banned?
No one but Joe has been banned here and you can find out why by directly asking him.
So, instead of supporting the censorship game, make your points on this site where onlookers can follow the debate properly.
While I would like to take part, it must look silly to the average fifth grader when adults “shout at each other from across a street”.
It also means that the history of a debate will be preserved here and not suddenly disappear.
This also applies to KF, UBP, and all the others.
Yep – in essence, make that ‘d’ stand for ‘designed’, and you have an excellect indicator of design! If you keep it as ‘digital’, you still have the ‘specified’ part, so again …
(Almost) all functional complex strings are designed. Because some are. Iteratively amended and survival-filtered copies of such strings are also designed, because the original was … probably.
Counting bits is just stamp collecting if it doesn’t support a hypothesis regarding how the bits got arranged.
It seems odd that gpuccio needs to know the history of a string in order to calculate the odds that it could have been created by a regular process.
I’d like the ID community propose a method that can say definitively whether a completely novel functional string could have been created by incremental variation and selection.
That is the biological question to be answered.
Chemistry is an “omniscient” oracle regarding function. Assuming an incremental path exists, chemistry selects. Douglas Axe know this, but he has chosen to waste his life looking for the keys in places where no one ever goes.
Gpuccio 219
There is a certain irony that your comment ends “I hope that answers all your doubts” when you have written 1648 words and not answered the two questions I asked! I asked for two rules on how to make decisions in the process of defining whether something has dFSCI and you haven’t given me any rules. Look through your comment and you will see that this is so. We are trying to define the dFSCI process and I promise you it is not nearly as obvious as you think it is.
Perhaps it will help if I try to restate the two required rules in a different way:
Rule 1
When confronted with a proposed function how do I decide if I need to know how the proposer came up with the function?
Your answer need not be long. I imagine it will be of the form “If a proposed function has features XXXXXXX then it is necessary to find out how the proposer came up with the rule”. Obviously XXXXXXX cannot include reference to how the proposer came up with the function!
Rule 2
Having investigated how the proposer came up with the function and determined that it was by inspecting the digital string and then finding a function it could perform (I assume this is right so far – feel free to correct), how do you decide what function to replace it with?
I imagine the answer will be of the form:
The replacement function has features YYYYYYYY
That is the essence of my comment. Ignore the stuff below if you are short of time.
Other comments
I think it may be helpful to be more precise in our language (your English is so good I am sure that it not being your native language is irrelevant). It may seem pedantic but it was only by this method that I realised the sense in which your statement that “everything with dFSCI is designed” not circular. Wittgenstein showed how keen attention to the subtlies of language can solve big problems.
First let’s not confuse:
For example, “carrying oxygen to the tissues” is a function while haemoglobin is a string which is capable of doing that. “Carrying oxygen to the tissues” is not a protein. It is a completely different category.
When you write:
You are making this category error. It may seem pedantic but it leads to subsequent confusion. The function is represent (or identify) papers about cystic fibrosis in the Pubmed database. There is a set of strings that can do this using PMIDs but there are other ways of performing the function.
The function I proposed was: “represent an ordered list of papers in the PubMed database”. Notice this makes no reference to the string whatever. It can be achieved by the string but could also be achieved by writing the titles of the papers or a set of URLs pointing to the papers. It just so happens that using PMIDs is the only way I can think of doing it with a string of less than 60 digits. The function is to represent these particular papers – not any old papers. I can hardly think of a more explicit description of a function. It is possible to tell with great confidence whether a string performs the function or not.
My function does not mean: “Any string that points to a set of papers that I will choose after I see the string, using the exact sequence of the string to look at the corresponding papers” it doesn’t even mean “represent a set of papers that I will choose after I see the string, using the exact sequence of the string to look at the corresponding papers”. That is a completely different function which you wish to substitute for mine.
There is also danger of confusing:
These are different things. The function is to represent those specific papers and can be performed (or not) whether I came up with the function by inspecting the string, choosing papers at random from pubMed (which I did in one case), or choosing the papers that were on my desk (to give three examples). Your dFSCI process may say that under some conditions it is necessary to know the method (see rule 1) but the method is not the function. It may also say that after knowing the method under some conditions it necessary to substitute a different function(see rule 2) – but it is not the same function.
The biological question of interest is not whether sequences are designed or not. The ID version of “design” is indistinguishable from magic. Magic is not a rational hypothesis. Human designs have histories. One can trace the history of any complex invention, even the mousetrap.
The biological question to be answered is whether incremental modification and differential selection can account for DNA sequences. So to be interesting, your measure needs to distinguish between strings that can be reached incrementally and those that cannot.
To be clear, I am limiting this to completely novel, objectively functional strings. Ones that have no history of prior existence or publication.
This “parallel posting” works as well as having a political debate on two different television channels.
Imagine Romney on Fox and Obama on CNN having a debate with each other in real time.
How does the audience follow it?
If your “dFSCI” ID argument can only be defended on a site that is ID friendly, how strong is it?
I used to post to UD but got banned because I disagreed with ID and for no other reason.
There are many of us here who disagree with you but UD is preventing us from debating you there.
Things on UD are not only ignored by mainstream biologists, they are ignored by the mainstream ID community. How may things have UprightBiped or gpuccio published at BioComplexity?
Surely the peer review standards there are favorable.
Now there’s a challenge for gpuccio and UBP ! 🙂
What would their papers look like?
How would they be accepted by other sites? !
Would other IDists find problems with gpuccio’s ideas if they were forced with accepting them as being scientifically representative of ID theory?
Mung 227
I don’t have a problem with your source code as an example of a function. Is there anyone who does? It is just that my definition “represents these papers” is equally testable and equally indepedent of the string – as I discussed above it could be performed in many different ways without the string ever coming into it. Gpuccio’s example of “Any string that has the following sequence: HHTHTHTTTHHTHTTTTHHTHHHTHTT” is not even a function and not at all similar to my example.
Let’s try another route. Here’s a history of a string, the function of which is pronounceability to English speaking people. Feel free to explain how your methodology can shed light on the change history. I particularly want to know how you account for these two points in the history:
GDJUSTIOUR
SESPECTIND
which have no characters in common (just as your precious protein domain coding sequences have no cousins). If you cannot posit a regular process that can do this, then how can you be an expert on what regular process can or cannot accomplish?
DUH
DUH
DUD
DUS
DUSS
DUNS
DUNS
DUNS
DUDS
DUDS
CURS
CURS
CURO
CURP
CAURP
SCAUR
USCOUR
USCOUR
USCOUR
AUSCOUR
PAUSCOUR
PAUSIOUR
PAUSTIOUR
SPAUSTIOUR
SPJUSTIOUR
SPJUSTIOUR
GPJUSTIOUR
GDJUSTIOUR
GDJUSTIONR
GDJUSTIONR
GDJUSTIOND
GDJUCTIOND
GDNECTIONB
LDNECTIONB
LDSECTIONB
LDSECTIONB
LDSECTIONB
LESECTIONB
LESECTIONK
OESECTIONK
RESECTIONK
RESECTIONK
RESECTIONK
RESECTIONI
RESECTIONG
RESECTION
LESECTION
LESSECTION
LESSECTION
LESSECTION
LESPECTION
LESPECTION
LESPECTIOH
PESPECTIOH
PESPECTIOV
PESPECTIOQ
PESPECTIOQ
PESPECTIOM
NESPECTION
BESPECTION
DESPECTION
WESPECTION
PESPECTION
MESPECTION
MESPECTIOK
CESPECTIOK
CESPECTIOX
CESPECTIOM
CESPECTIOR
CESPECTIOJ
LESPECTIOJ
LESPECTIOE
RESPECTIOE
RESPECTIKE
RESPECTIZE
RESPECTIAE
RESPECTILE
RESPECTILE
RESPECTIVE
RESPECTIVT
RESPECTICT
RESPECTICT
RESPECTICK
RESPECTICK
RESPECTICI
RESPECTIOI
RESPECTIOI
RESPECTIOD
RESPECTIOS
RESPECTIOS
RESPECTINS
RESPECTINS
RESPECTINS
RESPECTINS
RESPECTINS
RESPECTINV
XESPECTINV
XESPECTINS
XESPECTINH
UESPECTINH
KESPECTINH
SESPECTINH
SUSPECTINN
SUSPECTINN
SUSPECTINN
SUSPECTINX
QUSPECTINX
QUSPECTIVX
QUSPECTIVS
QUSPECTIVF
JUSPECTIVF
JUSPECTIVJ
JUSPECTIVI
RUSPECTIVI
RQSPECTIVI
IQSPECTIVI
DUSPECTIVI
SUSPECTIVI
SUSPECTIVY
SUSPECTIVY
SUSPECTIVT
SUSPECTIVO
SUSPECTIVO
SUSPECTIVD
SESPECTIND
So how do you account for the accumulation of ten or so changes to form a new function?
Gpuccio 229
Thanks for the considerable effort you have put into this.
I think you are really stretching things to say that my proposed function does not point unequivocally to a subset of the search space. It is utterly clear which strings can be used to represent that list of names whether the function be prespecified or not. In fact it is clearer than a rule such as “papers about CHD” as one could argue about which papers are about CHD. The function you want to substitute if I did not “prespecify” points to a different subset – but is clearly a different function. It is the function of representing any list of papers.
Yet another way to look at the problem. Take any function definition e.g. represents a list of CHD papers. Someone might have arrived at that function by first looking at the string, trying various things it could be used for, and then noticing that it could be used to represent a string of papers representing CHD. This would greatly decrease the complexity as by your argument the real function should be “can represent some set of papers”. The fact it was a CHD set was determined by the string. But you do not feel it necessary to ask the how the CHD function was arrived at. I can see no logical difference in the situations except a matter of degree.
Zachriel predicted this on October 25th in this thread:
And here we see gpuccio redefining functions to suit his conclusions.
Gpuccio’s most fundamental problem is he can only envision one dimension of functionality. He cannot imagine the fitness landscape having more than a vertical component.
He cannot imagine that function can have multiple views simultaneously. Hence the inability to see how complexity can evolve.
As interesting as all this ‘string theory’ is, I feel it completely misses the bus, certainly in terms of proteins, which are nothing without their 3D structure. Two protein domains can bear no sequence similarity yet have a high degree of structural congruence. And they can still derive from a common ancestor by stepwise substitution of every single part. Each amino-acid ‘letter’ is taken to be an equal distance from all others, and this is simply not the case. It isn’t the case in general, because amino acids cluster on properties, nor in specific instances, where the ‘distance’ between two substitutions, as determined by the 3D effect, is entirely dependent on the position in the broader matrix. If dFCSI takes no account of higher dimensionality, it is not likely to be a useful tool for determining protein ‘design’, even with a clear methodology for applying it to 1D structure. Elements many bits apart come together in a manner vital for ‘function’.
A further point relates to ‘function’. Function is frequently partitioned protein by protein, but the very modularity of protein domains means that the same domain can appear in proteins of widely different ‘function’. And the ‘function’ of the domain in each protein may itself be widely different, yet retaining the same 3D structure. So we have these sub-protein elements that display substantial phylogenetic congruence, some on sequence, some on structure, and some on both, scattered about the proteome. Their integration from one protein to another is entirely within the capacity of ‘RM + NS’. And yet … somehow, it isn’t. Someone had to move them, because they lend high ‘dFSCI’ to the strings which transcribe/translate/fold to make the proteins. dFSCI is not, seemingly, a transferable quality.
Such practical, chemical considerations, of course, part of the general “things that come out at TSZ […] better ignored”.
Alan Fox,
Pull up a comfy chair. I’ve been in the mode for a couple of years now. I think what we’re actually observing is the death throes of yet another variant of creationism. I don’t expect anything useful to emerge, but we may be present at the birth of whatever they come up with next to try to get their religion into public schools.
And in reference to the bit you quoted about dFSCI being “an excellent indicator of design”, didn’t gpuccio define dFSCI as only being present in designed strings? Did he change his definitions again?
It would be useful to have an example of what design is, as opposed to a list of things that it is not.
I’d like to see a before and after snapshot of an object that has a designed feature, along with some discussion of how the designer knows how to create the new function.
Human created objects seem to be the model, and we have a vast literature on how new inventions are born.
Gpuccio
I think when you start to use phrases such as “your argument is silly and useless” it is time to stop the debate. I can assure you I have similar feelings about your arguments. But, like you no doubt, I am tired of this and cannot afford any more time on it.
You know nothing about how DNA sequences were generated or selected, and yet you speak with authority about it. Unlike your precious protein domain sequences, I have given you an entire history. What can you say about how they were generated and selected?
Absolutely nothing.
You lack the imagination to see any possibility other than magic. You cannot conceive of a regular process that can assemble a completely novel string requiring eight or ten co-dependent elements, each of which is generated by a stochastic process. You cannot conceive of a finite selector that can produce novel, functional and complex sequences.
That is your problem, not a problem with biology.
The process I have modeled is not limited to eight or ten characters It can reasonably be extended to hundreds. It does not require resources beyond those of the universe. It scarcely requires a fast computer.
Why does gpuccio insist on knowing the origin of a string before he’ll try to use his dFCSI-determining protocol on it? If dFCSI genuinely is the sure-fire Design-detecting thingie that ID-pushers like to portray it as, I really don’t see any need to tell John Doe the origin of a given whatzit before Doe can use gpuccio’s dFCSI-determining protocol. Surely the results of using that protocol, assuming said protocol actually can do what ID-pushers claim it can do, would allow said protocol’s user to correctly identify whether or not the whatzit was designed, regardless of the protocol-users’ level of background knowledge about the whatzit they’re using the protocol on?
At this point, I feel confident in making a prediction about the results of gpuccio’s challenge, assuming he manages to disgorge any results thereto: For any string X, the answer to the question “has gpuccio determined that string X has dFCSI?”, and the answer to the question “has gpuccio been told, up front, that string X was Designed?”, will always be the same answer.
No, it’s to see where “dFSCI” has some sort of useful application when testing for “intelligent design”.
The whole point of “dFSCI” and all other versions of it, are to compare “a designer’s capabilities” versus “nature’s capabilities”.
Can “dFSCI” be used to detect possible design in the generation of the information in DNA that seems to have no apparent purpose, what some people call “junk DNA”?
Does DNA with no apparent function have “dFSCI”?
gpuccio,
Here’s a tough question about “dFSCI” I hope you will think about.
Imagine a friend brings you two pieces of paper.
One says, “The best food is tasted in Russia”, while the other says, “The best food is tasted in Poland”.
One was designed by your friend as a reply to the string that wasn’t designed.
What your friend had done was to send an envelope to people that he knew and ask them to change any one letter in a string of letters and then return that new string to him.
He kept repeating the procedure until he had a legitimate sentence.
No one “designer” had control over the final “specific” message that would be generated despite the appearance of intent in the outcome.
What letter to change was entirely random as people tried to steer the sentence to their “specific” conclusion and then see that it had wandered off in a new direction if they were lucky enough to receive the envelope for a second time.
Your friend of course would have a history of a sentence changing over time to what seemed to be a sure target and then drift away to something else.
Was the string the result of design or evolution?
Let us cut to the chase. Suppose we have a GA that embodies a random mutator and a selector that favors pronounceability.
We run it several times starting with the same seed string. How do you explain how the same starting point speciates and completely obliterates its origin? Even obliterates earlier versions of itself? Not only that, but the same starting point with the identical selector wanders off in radically different directions depending on the contingencies created by stochastic changes.
GP: your entire argument depends on this not being possible. Your discussion of protein domain sequences is based on this not being possible.
DUH
DUH
DUH
DUH
DUG
DUG
DUH
DUE
DUG
BUG
DUH
DUE
DUG
BUGG
DUH
DUEE
DUG
BEGG
DUH
DUEL
DUG
BEGG
DUH
DUOLL
DUG
BEGGG
DUD
DUOLLL
BUG
BEGGS
DUN
DUOLLA
BUB
BOGGS
DUNN
BHOLLA
HUB
BRAGS
DUNNN
BROLLA
HUH
BRAGS
SUNNC
BROLLA
HAH
BRAGS
SUNNC
BROLLS
CHAH
BRAGS
SUNNS
DROLLS
CHAHH
BRAGS
DUNNS
DROLLS
CHAPH
BROGS
DONNS
DYROLLS
CHAPH
BROGS
DOONIS
ODYROLLS
CHAOH
BROGS
DOONIS
ODDROLLS
CHAOH
PROGS
DOONIS
OYDROLLS
CHAOG
PROGS
DOONGS
VYDROLLS
CHANG
PROGS
DOINGS
BYDROLLS
CHANG
PPROGS
DOINGS
BOYDROLLS
CHANG
APPROGS
RDVINGS
YOYDROLLS
SCHANG
APPRODS
VRDVINGS
ROYDROLLS
CHANGI
APPRODS
PRDVINGS
ROWDROLLS
CHANKI
APPRODS
PRDTINGS
ROWMROLLS
CHANKI
APPRODS
PRDTINGS
PROWMROLLS
CHINKI
APPRODS
PRDTINGS
PROWMROLLS
CHINKI
APPREDS
PROPINGS
PIOWMROLLS
CHINKI
APPREDS
PROPINGS
PIOKMROLLS
CHINKI
APPREDS
PROPINGS
PIOAMROLLS
CHINKI
PAPPREDS
PROPINGS
PIHAMROLLS
CHINDI
PAPPRESS
PROPINGS
IHAMROLLS
SHINDI
PAPPRESS
PROPINGS
HIHAMROLLS
SHINDS
PAPPRESS
PROPINGS
UIHAMROLLS
SHIRDS
PAPPRESS
PROPINGS
DIHAMROLLS
QHIRDS
APPRESS
PROVINGS
DIMAMROLLS
XHIRDS
ADPRESS
PROVINGS
DIMAMROLLS
XHIRTS
IDPRESS
PROVINGS
DIMAMROLLS
FHIRTS
IMPRESS
PROVINGS
JIMAMROLLS
CHIRTS
IMPRESS
NROVINGS
JIKAMROLLS
CHIRTS
IMPRESC
AROVINGS
JITAMROLLS
CHIROS
IMPRESP
DROVINGS
JITAMROLLS
CHIROS
IMPPIESP
DROVINGS
NITAMROLLS
CHINOS
IMPPIESP
IROVINGS
PITAMROLLS
CHINOS
IMPPIESJ
IRONINGS
PITAMROLLS
CHINOS
IMOPPIESJ
IRONINGS
PITATROLLS
CHINES
MOPPIESJ
IRONINGS
PITATROLLS
CHINES
MOPPIES
PRONINGS
PITATROLLS
CHINNES
LOPPIES
PROTINGS
POTATROLLS
CHINNESS
GLOPPIES
PROTINGS
POTATROLES
CHINNESS
CLOPPIES
PROTINGS
POTATROLES
CHICNESS
CLOPPIES
PEOTINGS
POTATROLES
CHICNESS
CLOPPINS
MEOTINGS
POTATROLES
CHICNESS
CLOPPINS
MENTINGS
POTATROLES
OHICNESS
CLOPPINS
MENTINGS
TOTATROKES
GHICNESS
CLOPPINT
METTINGS
TOTATROKES
ZHICNESS
CLOPPINH
METTINGS
TOTATROKES
PHICNESS
BLOPPINH
EETTINGS
TOTATROKES
PHISNESS
BLOPPINM
RETTINGS
TOTATROKES
PHISNESS
BLOPPINY
RETTINGS
TOTATROKES
PHISNESS
BLOPPIND
GETTINGS
TOTATROKES
XHISNESS
ELOPPIND
GETTINGS
TOTATROKES
CHISNESS
ELOPPINY
GETTINGS
TELATROKES
YHISNESS
GLOPPINY
IETTINGS
TELATROKES
WHISNESS
GLOPPINK
IUTTINGS
TELATROKES
LHISNESS
GLOPPINK
PUTTINGS
TELATRONES
LHIMNESS
GLOPPINW
PUTTINGS
TELATRONES
LLIMNESS
CLOPPINW
PUTTINGS
TELATRONES
PLLIMNESS
CLAPPINW
PUTTINGS
TELATRONES
PLLIGNESS
CLAPPINW
PUTTINGS
TELATRONES
PLLIGNESS
CLIPPINW
JUTTINGS
TELATRONES
PLLIGNESS
CLIPPINW
JUTTINGS
TELATRONES
PLLIGNERS
CLIPPINK
RUTTINGS
TELATRONES
MLLIGNERS
CLIPPINO
RUTTINGS
WELATRONES
ELLIGNERS
CLIPPING
RUTTINGS
WELATIONES
ECSIGNERS
CLIPPING
RATTINGS
WELATIONES
ECSIGNERS
CLIPPING
RATTINGS
WELATIONES
ECSIGNERS
SLIPPING
RATTINGS
DELATIONES
EASIGNERS
SLIPPING
RATTINGS
DELATIONES
BASIGNERS
SLAPPING
RASTINGS
DELATIONE
SASIGNERS
SLAPPING
RASTINGS
QDELATIONE
SESIGNERS
SLAPPING
RASTINGS
PQDELATION
CESIGNERS
LAPPING
RASTINGS
PJDELATION
JESIGNERS
CLAPPING
JESTINGS
PDDELATION
JISIGNERS
QLAPPING
JESTINGS
ODDELATION
JDSIGNERS
KLAPPING
JESTINGS
ODMELATION
JLSIGNERS
FLAPPING
JESTINGS
ODMELATION
BLSIGNERS
FLAPPING
TESTINGS
ADMELATION
BOSIGNERS
FLAPPING
TESTINGS
ADVELATION
NOSIGNERS
ULAPPING
TESTINGS
DVELATIONS
WOSIGNERS
ALAPPING
TESTINGS
DVELATIONS
FESIGNERS
ALAPPING
ZESTINGS
DVELATIONS
VESIGNERS
SLAPPING
ZESTINGS
MVELATIONS
BESIGNERS
SLOPPING
WESTINGS
MAELATIONS
PESIGNERS
SLOPPING
WESTINGS
BAELATIONS
AESIGNERS
SLOPPING
WESTINGS
GAELATIONS
DESIGNERS
RLOPPING
WASTINGS
GIELATIONS
DESIGNERS
JLOPPING
WASTINGS
HIELATIONS
DESIGNERS
ULOPPING
WASTINGS
WIELATIONS
DESIGNEES
LOPPING
WASTINGS
WIERATIONS
DESIGNEES
LAPPING
FASTINGS
WIERATIONS
DESIGNEES
HAPPING
FASTINGS
WIERATIONS
DESIGNEES
HAPPING
FASTINGS
WIERATIONS
DESIGNEES
NAPPING
FASTINGS
WIERATIONS
DESIGNEER
NAPPING
FASTINGS
WITRATIONS
DESIGNEE
CAPPING
FASTING
WITRATIONS
DESIGNED
CAPPING
BFASTING
XITRATIONS
DESIGNED
QAPPING
BNASTING
JITRATIONS
DESIGNED
ZAPPING
NASTING
JMTRATIONS
SESIGNED
ZAPPING
HASTING
JTTRATIONS
PESIGNED
PZAPPING
HASTING
JTERATIONS
RESIGNED
BZAPPING
HASTING
JJERATIONS
RESIGNED
JUAPPING
BASTING
JAGRATIONS
RESIGNED
JWAPPING
BASTING
JAGRATIONS
NESIGNED
JWAPPING
BASTING
JAGRATIONS
LESIGNED
WAPPING
RASTING
JAURATIONS
FESIGNED
KAPPING
RESTING
UAURATIONS
GESIGNED
KDPPING
RESTING
TEURATIONS
GESIGNER
KEPPING
RESTING
CEURATIONS
HESIGNER
KEPPING
OESTING
BEURATIONS
HESIGNEA
DEPPING
YESTING
BETRATIONS
HESIGNEAA
LEPPING
YESTING
BOTRATIONS
HESIGNEAE
LEPPING
HESTING
JOTRATIONS
HESIGNIAE
PEPPING
HESTING
JNTRATIONS
HESIGNIAY
PEPPING
VESTING
TNTRATIONS
BESIGNIAY
AEPPING
VESTING
CNTRATIONS
BESIGNIAH
BAEPPING
TESTING
CNTCATIONS
BESIGNIAT
BSIPPING
JESTING
CNTCATIONS
BESIGNIAJ
BSIPPING
QESTING
CNTCATIONS
RESIGNIAS
SIPPING
KQESTING
CNTUATIONS
CRESIGNIFS
SAPPING
KAESTING
CATUATIONS
DRESIGNIFS
SAPPING
KAESTING
CATTATIONS
DRESIGNIFG
SAPPING
AESTING
CALTATIONS
RESIGNIFG
PSAPPING
PAESTING
MALTATIONS
Well I have to complain to the management. the site software obliterated my table after displaying it correctly in the preview. As displayed, each group of four strings should display on one line, representing the same generation in four runs.
Gpuccio 238
I was going to drop this thread but I suddenly had an extraordinarily simple insight. All you have to do is modify the definition of dFSCI to add one more condition:
That’s reasonably clear to me. It follows that you can only deduce dFSCI if you know how the function was arrived at – but that seems clear. My only objective is to come with a clear set of rules for determining dFSCI and that seems to me a satisfactory rule 1. It also makes rule 2 unnecessary. If it is post-specified then you simply declare that it is not dFSCI you don’t attempt to substitute another function. See I am really trying to help you! Mind you it only shifts the argument because I think the functions found in life are post-specified!
I don’t see the relevance of your Bayesian challenge – I thought we were trying to define a process with 100% specificity – not estimate which hypothesis is most likely. But it is quite interesting to try and answer it. It is quite a lot of work so I will save it for another comment.
I have two comments in moderation.
Can anyone see this?
Gpuccio 238 Bayesian Exercise
This turns out to quite a demanding task involving probabilities but also psychology. I think the process is more interesting than the result which involves a subjective judgment. It has very little relevance to our previous discussions.
Bayes theorem is about determining the probability of a prior condition given some data (where the prior condition may be a conjunction of many conditions). First its let’s clear up what the prior conditions are. I don’t recognise function (a) and can’t make much sense of it. My function was “refers to 5 specific papers in the PubMed database” – as I have said they can easily be listed even if the string was never thought of. In fact I don’t see how you could create a list such that “We can explicitly list the entries, if and only if the string is already known”. There will always be other ways of listing them – such as the titles or the URLs.
I have an alternative definition which I think you will find acceptable and which makes it much easier to do the analysis.
First I will abbreviate: “the string was generated in a Random String Generator, in one single attempt” to S – it saves typing.
Let’s start with these three conditions, which with the exception of (a) match yours:
a) S & “The function is to point to 5 entries in the PubMed databases with these titles: A,B,C,D,E.” (I will not write out the titles)
b) S & “The function is to point to 5 entries in the PubMed databases, all of them indexed by the keyword: “disease”.
c) S & “The function is to point to 5 entries in the PubMed databases, all of them indexed by the keyword: “elaprase”.
Any of these might have been prespecified or postspecified. So we can break them down into conjunctions thus:
(a1) (a) & was prespecified
(a2) (a) & was postspecified.
(b1) (b) & was prespecified
(b2) (b) & was postspecified
(c1) (c) & was prespecified
(c2) (c) & was postspecified
Prespecified and postspecified are mutually exclusive so: Pr(a) = Pr(a1)+Pr(a2), Pr(b) = Pr(b1)+Pr(b2), Pr(c) = Pr(c1)+Pr(c2)
The outcome is the same in all 6 cases – the string performs the function.
As you well know, there are two parts to the Bayesian calculation: the probability of the outcome given the prior and the probability of the prior.
Consider just the probability of the outcome given the prior (in what follows I will abbreviate outcome to “o”).
First consider the postspecified cases. Pr(o|prior) is 1 for all of them. Given that the function has been selected to match the string then it follows that the string must match the function.
Pr(outcome|prior) is extremely small for all the prespecified cases but we can say pr(o|c1)>>pr(o|b1)>>pr(o|a1).
At this stage you can see that really the prespecified options are effectively irrelevant to the best decision.
Take the example of (a)
Bayes formula says: Pr(B/A) = Pr(A/B)*Pr(B)/Pr(A)
So Pr(a1|o) = Pr(o|a1)*Pr(a1)/Pr(o) and Pr(a2|o) = Pr(o|a2)*Pr(a2)/Pr(o)
Therefore Pr(a|o)= Pr(o|a1)*Pr(a1)/Pr(o)+Pr(o|a2)*Pr(a2)/Pr(o)
We don’t know all the values but we know Pr(o|a1) is extremely small and Pr(o|a2)=1. So effectively unless Pr(a2) is also extremely small we can forget about the case where the function is prespecified. The same argument applies for b and c. So for all three cases, all that matters is the probability of the function being postspecified – the realistic alternative being that the string is not randomly generated.
Which has the highest probability of being postspecified? Without any context it is extremely hard to make any sensible estimate. For example, if it were you judging functions and strings I had provided you could make a psychological estimate as to which case was I most likely to postspecify . It somewhat harder work to postspecify (b) and (c) than (a) so I guess you would plump for (a) – but of course I might have been trying to bluff you into choosing the wrong one!
Mark is the only one who can see it, since this is his thread.
The spam filter seems to be going crazy today. On one of my threads, both Zachriel and I have had comments in moderation for no apparent reason.
Moderation seems to be the rule now. Weird.
Even my own comments are going into moderation!
Gpuccio 245
Thank you for your comment but I really don’t see what all the fuss is about. I defined a function “refer to these 5 papers”. The function might have been post-specified or pre-specified. Pre-specified would have mean’t a chose the 5 papers some way independent of the string e.g. by entering key words at random into the database. Post-specified would have mean’t defining the string and then using it to choose the papers. I admit the probability of getting a match if pre-specified was negligible.
My only point is that this is true for the other three functions as well – not as negligible as for the 5 papers – but negligible compared to the probability that someone decided to either design them or postspecify them.
So what it comes down to is which one is the most likely to be a post-specified (as opposed to designed). That’s a pyschological assessment. As I say it is good deal more work to postspecify b or c than a so I would put my money on a (although of course someone might a special motivation for post-specifying b or c – like fooling the observer). I would do this rather reluctantly as I think all three are more likely to be designed.
Where is the problem in this argument? I am not even sure if we disagree. I am just to clarify what you are asking me to do and then show how I would do it.
Gpuccio
More to the point – I think I have another example which would give you reason to refine your dFSCI process if you want to preserve 100% specificity. Before I do the work let me check the function is acceptable:
“The string identifies for each month over a period of 120 months whether the London monthly mean high temperature is above or below long-term average.”
As I have given you the function before working out the string you can see that I am prespecifying it!
Gpuccio
That’s false. For the other two examples if they were post-specified this would be something like taking the string, studying the papers it points to, and seeing what you can find that they had in common. As all papers have something in common (even if it is just a distinctive phrase somewhere in the text) then the probability of success is 100%. That’s why I suggest you simply amend the process to say no post-specified functions. Any function could potentially be post-specified.
I was thinking of the last 10 years – 2002 to 2012 – I could do a longer period but it would be tedious. I was going to use http://www.holiday-weather.com/london/averages/ for the averages. Although the values for 2002 to 2012 are known I was not going to use them to generate the string. That’s why I said “identify” rather than “predict”. I will not even look at the actual temperatures until after I have generated the string – although I won’t be able to resist checking it has worked when I have finished. The string will simply be a string of 120 bits with 1 for above average and 0 for below average. I realise you want 500 bits but that would be really tedious to look up all the data, so I hope 120 will be sufficient to prove the case.
To anyone following this thread!
TSZ is behaving so badly I plan to move my responses to Gpuccio to my own blog temporarily.
Sounds a bit like tree ring width. Or pulse code modulation of a sound wave.
I still have a comment in moderation. It may be causing trouble due to its HTML. Perhaps deleting it would fix th problem.
Where’s Lizzie
I don’t think gpuccio can make the circularity of his thesis any more clear:
Mark:
gpuccio:
Which is to say: phenomena that arise by natural means do not exhibit dFSCI by definition.
It follows that, regardless of other properties an object may have, it cannot be concluded that it exhibits dFSCI until its causal history is known. No procedure or calculation performed upon the object can alone warrant the conclusion that the object exhibits dFSCI absent knowledge of that causal history.
It further follows that to claim that dFSCI present in an object is evidence for a particular kind of causal history (it was designed) is patently circular, as you cannot assert that dFSCI is present until that causal history is known.
And, lastly, it follows that no enumeration of supposed of objects displaying dFSCI, defined in this way – and the absence of counter examples in this collection – has any empirical bearing upon the question of whether the exclusion of natural objects by definition is in fact appropriate.
http://www.uncommondescent.com/education/the-tsz-and-jerad-theread-iii-900-and-almost-800-comments-in-needing-a-new-thread/#comment-438088
I think there is some confusion (my fault – I have explained it badly). My plan was to point to the list of London temperatures not as the string but to help define the function. The function was to predict whether those temperatures were above average without looking at them. Another string, of which I would tell you nothing, just give you the string would be the one that did the predicting. Your challenge would then be tell me whether that string was designed.
In fact that string would be the based on the temperature record of an adjacent location – but I wouldn’t tell you that. So it would be:
Complex
Digital
Functional using a prespecified function
OK. I will be more precise. So the point of dFSCI is to detect design origin when there is a positive case. This is not much use if you have to know the origin to determine if there is dFSCI in the first place. Right? You always have to watch for circularity with dFSCI because it is used to determine whether something has a design origin but in order to decide whether something has dFSCI you have to assess whether it has other origins. We only agreed a set of conditions when it was not circular by being extremely precise about the definition and recognising it was relative to a particular observer at a particular time with that observer’s knowledge.
As a matter if interest can you point to a real example of someone using dFSCI to detect design when they didn’t already know the answer because they knew the origin was designed for other reasons?
gpuccio, Mung, Joe,
But wouldn’t a necessity mechanism be the purest source of “specific” information?
A string of “information” can be “digital”, “functional” and “complex” but the key attribute is “specified” which has to do with its origin.
How can you say a necessity mechanism is NOT generating something specific?