Uncommon Descent Serving The Intelligent Design Community

Why we shall have to wait for a real biography of Stephen Hawking

Share
Facebook
Twitter
LinkedIn
Flipboard
Print
Email

In a review of Kitty Ferguson’s Stephen Hawking, Ed Lake “examines how Stephen Hawking gets the world to sit up and take notice” (Telegraph, January 12, 2012):

When he speaks, as he has this week on his 70th birthday, the world takes notice. That’s partly down to his distinguished career but, let’s not be squeamish, partly because his motor neurone disease and voice synthesiser have made him a convenient symbol for the life of the mind. That aura of mystical detachment doesn’t quite stand up to examination, however. “Was it just an accident that he always seemed to come up with attention-getting statements whenever public and media attention appeared to require a boost?” asks Kitty Ferguson in her starry-eyed biography. As one of Hawking’s assistants told her: “He isn’t stupid, you know.”

One starts to suspect that his real genius may be for judging the appetites of the public.

Well, how about the appetites of self-conscious urban elites – people who feel knowing about “imaginary time” and space wormholes, who could not point to and name a single star visible in their own region.

Indeed, there’s so little that’s dark or sad about her Hawking, the effect is almost sinister. Perhaps he really is just a permanently upbeat and sunny chap. On the other hand, …

On the other hand, that’s highly unlikely. Such people exist, to be sure, but they don’t think, say, or do the things Hawking has. Which is why we shall have to wait for a real biography of Stephen Hawking.

Stephen Hawking at 70: What would revolutionize our understanding of the universe

Follow UD News at Twitter!

Comments
Peter: OK, and I leave you with the following link to a Wikipedia page: http://en.wikipedia.org/wiki/Probability_distribution The article linked is 18801 characters long.gpuccio
January 20, 2012
January
01
Jan
20
20
2012
08:07 PM
8
08
07
PM
PDT
Peter asks "why not use the same terminology?" as if that is some type of argument. But anyway, they do use the same terminology Peter:
Biological specification always refers to function. An organism is a functional system comprising many functional subsystems. In virtue of their function, these systems embody patterns that are objectively given and can be identified independently of the systems that embody them. Hence these systems are specified in the same sense required by the complexity-specification criterion (see sections 1.3 and 2.5). The specification of organisms can be crashed out in any number of ways. Arno Wouters cashes it out globally in terms of the viability of whole organisms. Michael Behe cashes it out in terms of minimal function of biochemical systems.- Wm. Dembski page 148 of NFL
In the preceding and proceeding paragraphs William Dembski makes it clear that biological specification is CSI- complex specified information. In the paper "The origin of biological information and the higher taxonomic categories", Stephen C. Meyer wrote:
Dembski (2002) has used the term “complex specified information” (CSI) as a synonym for “specified complexity” to help distinguish functional biological information from mere Shannon information--that is, specified complexity from mere complexity. This review will use this term as well.
And what is this magical "selection" you evos keep talking about?Joe
January 20, 2012
January
01
Jan
20
20
2012
06:00 PM
6
06
00
PM
PDT
Peter G: “Then presumably the same can be said of human beings, who are really just programs too.”
See Emperor's New Mind by Sir Roger Penrose for a good treatment of why that's probably not true. Human's appear to exhibit insight that is non-algorithmic, i.e, not "just programs." Moreover, there is no evidence whatsoever that our consciousness is a program, unless you know something I don't. I'd like to hear it if you do. Knock yourself out.mike1962
January 20, 2012
January
01
Jan
20
20
2012
03:38 PM
3
03
38
PM
PDT
It's late so I'll just leave you with this, for now. A
work chairs in clothing when crayons cheat to loiter because papers cheat
B
feeling my totem I considered my fingers throughout my clothing
Peter Griffin
January 20, 2012
January
01
Jan
20
20
2012
03:32 PM
3
03
32
PM
PDT
Peter: Stilla lot of consfusion, but at least it seems we are arriving somewhere. So, let's go in order. So why not use the same terminology? We are independent people, with independent minds, not a party. Maube in the future the terminology will be unified. I have chosen my terminology mainly for clarity in the debate here. Exactly so. You can’t identify any reason to calculate it and so won’t. There is no reason to calculate it because the answer derived is not actually useful in any way. The main purpose in computing dFSCI for protein families is to use it in the debate about design inference. There are obviously other scientific purposes, for instance to understand the relation between functional density and structure function relationship in proteins. The values given by Durtson have been immensely useful in the debate about design inference, and i have used them many times. I would be very happy is new values were computed for other protein families, but as you probably know, that is not exactly my job: I am a medical doctor. Absurd! What if you discover a function, and assign dFSCI then. And then you discover another function, what then? Does the dFSCI of the two functions get added together? No. I supposed it should be clear at this point, but I will repeat it just the same: a) If I cannot recognize no function in the protein (or other kind of sequence), then I say that the protein does not exhibit dFSCI, and I don't infer design. It is a negative. It could be a true negative (protein not designed) or a false negative (protein designed, but function not recognized, or protein designed but function too simple). b) If in a second time a function is recognized, the judgment is simply reviewed (as always happens in science) in the light of the new data, and dFSCI frp that function can be computed. c) If a second function can be described for the same sequence, dFSCI can be computed for that other function. That does not get added to the other value. As said many times, each value of dFSCI is specific for a defined function. It's the complexity of the function, not of the object. A conscious intelligent agent can run programs! Learning to drive for example, are you consciously driving the car or is there some “muscle memory” going on too that you don’t consciously think about? I.E. You want to accelerate and trigger the “accelerate program”. Aren't you simply changing what you said. What you said was: "Then presumably the same can be said of human beings, who are really just programs too." (Emphasis mine) It's definitely not the same as "A conscious intelligent agent can run programs!" Please, decide yourself. Sure, it’s “just an example” but typically examples can be used to apply the procedure on a different text. No. You obviously did not follow the reasoning. The example on your summary was an introduction aimed to demonstrate that dFSCI increses inevitably as the length of text increses. Then I have made a more detailed discussion arguing, that I quote again here, that applies to all possible texts: "Now, let’s say that we have a segment of text, dense and specified, of say 100 characters. Let’s say that the functional combination of intelligible text conveying the meaning of the orioginal text are no more that 1:1000 (something that is not obly obvious, but easy to prove by any computer simulation). Let’s say, even contrary to the evidence, that they are 1:1000. therefore, the dFSCI for that segment will be approximately 10 bits. According to the previous reasoning, if we increase the length of the text, keeping it dense in meaning, we will gain about 10 bits of dFSCI for each new 100 characters. So, for 1000 characters dFSCI will be about 1000 bits, and for 10000 characters it will be about 1000 bits, that is well beyond Dembski’s UPB. That result will never be obtained by a computer by a random search. And this is only a lower threshold of dFSCI for text, based on an assumption of target space that is evidently too big." That proves my point, that it is possible to give a higher threshold of length that guarantees that any text longer than that has certainly more than 1000 bits of dFSCI. I have given 10000 characters, computing from thhe simple assumption, easily verifiable by a compute simulation, that the functional space/search space for a text 100 characters long is less than 1:1000. Do you question that? Just choose a 100 characters long text, with specific dense meaning, and run a random search for a string of 100 characters in a computer program, say 10000 times. No great number here, no great resources implied. And see if you get one sequence that intelligibly conveys the meaning of the original text. So, if that assumption is confirmed, any text longer than 10000 characters will exhinit at least 1000 bits of dFSCI. For any such text, we can safely infer design. yet again it’s “possible in principle but not practice” No. I have made it in practice. I have demonstrated that any english text that is dense and meaningful and longer than 10000 characters exhibit dFSCI beyond the threshold of 1000 bits, and supports a design inference.
Ah, here is the heart of the matter. What relevance is it to discuss the probability of a protein randomly occurring if proteins don’t randomly occur? What relevance is it to make calculations based upon random probability when the thing you are trying to disprove (evolution) does not work that way at all? Proteins don’t randomly appear, 747s don’t randomly assemble in junkyards for good reason. Yet that straw man persists.
I have given you a detailed answer to that, through my link to that other thread. Please, read it. I will not copy everything here. Such small probabilities can be greatly increased via selection. Such as the example I already gave you of random generation of Shakespeare. I never use dFSCI where selection can be proved. Only for transitions where no selection has ever been shown to be possible. So, no straw man at all. Please, again, read my posts in the other thread. If the complexity is probabilistic then why does knowing if the artifact in question have function come into it at all then? I am really surprised that you still don't get it. Any random string has the same probabilistic complexity of a functional string, if the length is the same. But in the functional string, we measure the probabilistic complexity necessary to implement the defined function, not absolute probabilistic complexity. That's why knowing if the artifact has function is of the utmost importance. It's not difficult, after all, if you drop for a moment your prejudices. Two proteins, one with known function and one with no known function, with the same length. The dFSCI for A is calculated at X, and B at 0 (presumably?). OK. But wait, I’ve made a mistake! The labels were switched on the vial, and it seems that we had the samples mixed up. And so? You have made a mistake. It happens. Now A and B have not changed at all but somehow they’re gained/lost dFSCI based solely on factors external to the samples themselves. I really don't follow you any more. You have made a mistake. Therefore your data and your conclusions are wrong. What has that to do with A and B "gaining or losing" dFSCI? Just redo your evaluation correctly, and you will find the right answers. Then what relevance does the function of the protein have in that case? Surely the function of a protein is, in this metric, the same as the meaning of a given example of text? Yes. Where is the problem? The function of a protein has the same role as the meaning in a text. In both cases, what we want to know is the probability of getting that function, or that meaning, from a random search in a random system. That 's extremely relevant to the discussion about neo darwinian algorithm, where RV is suppose to be the origin of any new function. The role of NS, as explained in my othe rthread, must be evaulated separately, if and when it is explcitly shown to exist. If a protein with not much function has a lower dFSCI value then a protein with loads of function then texts with little meaning should have less dFSCI then texts with loads of meaning? Yet you are saying that’s not the case! The correct form is: "A protein with a function that is simpler to be implemented (which requires a lower number os specific bits) has a lower dFSCI value then a protein with a function that is more complex to be implemented (which requires a higher number of specific bits); just as texts a meaning that is simpler to be conveyed (which requires a lower number os specific bits) have less dFSCI then texts with a meaning that is more complex to be conveyed (which requires a higher number of specific bits)" In any case, it is not the conceptual difficulty or originality of the function that is being measured, but the specific functional bits that are necessary to implement, or convey, it. The reason is simple. What we are measuring, as said many times, is the probability for that sequence to arise in a random system. Now, if Einstein has said in 10 word a concept so deep and important that it has changed the history of science, and that it is difficult to understand for normal people, I am happy of that, but still the probability of those ten words to come about in a rando search are the same: the probabilities of a ten word sentence (with the individual variability ascribable to different robustness of specific words to random vairation, not to the depth of the message that is being conveyed). I disagree, somewhat. The function is present (and therefore dFSCI) regardless of if the observer knows it or not. I can agree, but I am speaking of a procedure, not of philosophy. The function must be recognized and defined objectively by a conscious observer, so that it can be objectively measured, and its specific complexity evaluated. Yet we need more then the information in the object in question to be able to calculate the dFSCI in it, as you have just said! OK, and so? That “definable function” does not change the amount of information in the protein itself. Certainly not, and nobody has ever said anything like that. To define objectively the function, and how to measure it, is simply a necessary step in our procedure for evsluating dFSCI and inferring design. Does it? I have not seen that anywhere so far, please give an example! Again, read my posts in the other thread about modelling the neo darwinian algorithm. And yet you presumably infer design? Why? Because basic protein domains exhibit dFSCI, that is a marker of design, and any algorithmic explantion of their origin based on RV and NS is not supported by any facts, and indeed falsified by what we know. Again, read the other thread and then we can deepen some aspects of this discussion (by the way, I must still write my last post in that thread, to complete the discussion about positive NS. I will do that as soon as possible).gpuccio
January 20, 2012
January
01
Jan
20
20
2012
03:00 PM
3
03
00
PM
PDT
gpuccio,
But the same thing is being measured: the ration betwenn the target space and the search space.
So why not use the same terminology?
Why should I calculate it? Durston has done it.
Exactly so. You can't identify any reason to calculate it and so won't. There is no reason to calculate it because the answer derived is not actually useful in any way. But why should you calculate it? You tell me!
Sure, if a protein has no definable function, we cannot say that it exhibits dFSCI.
Absurd! What if you discover a function, and assign dFSCI then. And then you discover another function, what then? Does the dFSCI of the two functions get added together?
Ah, very kind of you to inform me of that. Stupid me has always thought human beings are conscious intelligent agents…
Who "run programs". There is a reason we use the word "deprogramming" in relation to people. A conscious intelligent agent can run programs! Learning to drive for example, are you consciously driving the car or is there some "muscle memory" going on too that you don't consciously think about? I.E. You want to accelerate and trigger the "accelerate program". If you can't "program" yourself (via learning) then what do you do when you learn?
Well, I have done that in my next post. I thought you had read it too.
I did, but there you said
and that we have determined that the number of intelligible strings of that length that still convey the full meaning is, say, 1.6 x 10^60, that is about 200 bits. It’s not important is that is true or not, it’s just an example.
Sure, it's "just an example" but typically examples can be used to apply the procedure on a different text. There are so many unknowns in your procedure (how do you work out how many ways there are of saying the same thing for example?) that yet again it's "possible in principle but not practice" What would convince me is if you developed a procedure that two independent people could follow and determine the same value for the same text, and then again for a different text.
I don’t think you have been discussin ID for so long, and still don’t understand that simple difference.
Ah, here is the heart of the matter. What relevance is it to discuss the probability of a protein randomly occurring if proteins don't randomly occur? What relevance is it to make calculations based upon random probability when the thing you are trying to disprove (evolution) does not work that way at all? Proteins don't randomly appear, 747s don't randomly assemble in junkyards for good reason. Yet that straw man persists.
Not if you need much nore than the whole resource of the univers to just get a small probability of success.
Such small probabilities can be greatly increased via selection. Such as the example I already gave you of random generation of Shakespeare.
Yes, it is. It is measuring probabilistic complexity, not conceptual complexity. Nobody has ever said anything of that kind. The complexity is probabilistic, not conceptual. The definition of the function can include sone perspective about its conceptual value, but that is not implied in the computation of the probabilistic complexity.
Exactly. But you contraction yourself, I think. If the complexity is probabilistic then why does knowing if the artifact in question have function come into it at all then? Two proteins, one with known function and one with no known function, with the same length. The dFSCI for A is calculated at X, and B at 0 (presumably?). But wait, I've made a mistake! The labels were switched on the vial, and it seems that we had the samples mixed up. Now A and B have not changed at all but somehow they're gained/lost dFSCI based solely on factors external to the samples themselves.
You have obviously never understood the meaning of dFSCI, or of CSI, if you think it should be related to the conceptual depth of the content.
Then what relevance does the function of the protein have in that case? Surely the function of a protein is, in this metric, the same as the meaning of a given example of text? If a protein with not much function has a lower dFSCI value then a protein with loads of function then texts with little meaning should have less dFSCI then texts with loads of meaning? Yet you are saying that's not the case!
The function is defined by a conscious observer, and measured in some objective way, for instance in a lab, for proteins.
I disagree, somewhat. The function is present (and therefore dFSCI) regardless of if the observer knows it or not.
What has that to do with adding information to the claculation? What we are measuring is the information in the objetc, not the information necessary to recognize the information in the object. Please, be more clear.
Yet we need more then the information in the object in question to be able to calculate the dFSCI in it, as you have just said!
Sure, if a protein has no definable function, we cannot say that it exhibits dFSCI.
That "definable function" does not change the amount of information in the protein itself.
dFSCI takes into account all that is known about the process that created a protein.
Does it? I have not seen that anywhere so far, please give an example!
Nothing is known about the process that created basic protein domains.
And yet you presumably infer design? Why?Peter Griffin
January 20, 2012
January
01
Jan
20
20
2012
09:00 AM
9
09
00
AM
PDT
Peter: Except I don’t believe he has, really. He’s calculated something and named it functional sequence complexity. You say that dFSCI is exactly the same thing? Then why not use the same name for the same thing? It's the same thing. You read the paper, and use your intelligence, and you will see it. Durston has used the term FSC (functional sequence complexity), deriving it from Abel, to indicate the complexity that contributes to function. I have used the term dFSCI to indicate the same thing. "d" is for digital. Dirston does not say explictly that the sequences are digital, but they are. I have made that explicit in the concept. That's all. Functionally specified information is the same as functional sequence complexity. The main difference in the concept of dFSCI and FSC is that FSC is a continuos measure, while dFSCI is used coupled to a threshold of complexity, to infer design. But the same thing is being measured: the ration betwenn the target space and the search space. Please, refer to the Durston paper, table 1. Let's take the values for ribosomal S12, just as an example. The length of the sequence is 121 AAs, that gives a search space of 20^121, that is about 523 bits. That's exactly the value Durston gives as "null state": the total uncertainty in the random state. The FSC (the functional complexity) is given as 359 fits. That is obtained by comparing all the sequences in the family, and evaluating how much each AA can vary in the family. If an AA is always the same, it will "reduce the uncertainty" of 4.32 bits. That is the same as saying that that aminoacid cannot vary, that the functional information of that position is the highest (4.32 bits). If a protein were built only of AAs that cannot change, the reduction of uncertainty would be equal to the complexity of the random state, the functional complexity would be as big as the search space, and the probability of getting the function by a random search would be 1/search space, because only one sequence in the search space would be functional. If a site can vary, the reduction of uncertainty is computed according to Shannon's principles, according to how much it can vary. If any AA can be at that site with the same probability, it just means that the site has not effect on the function, and the reduction of uncertainty is 0. In the case of Ribosomal S12, the mean functional density per site is very high: 3.0 bits. The search space is 523 bits. The functional complexiti space is evaluated as 359 bits. That means that the functional space is about 164 bits. My follow up question as to when you *will* be calculating it and *why* you would calculate it still remain! Why should I calculate it? Durston has done it. So, you would agree that a protein could have no dFSCI then? Sure, if a protein has no definable function, we cannot say that it exhibits dFSCI. Is it? Then presumably the same can be said of human beings, who are really just programs too. Ah, very kind of you to inform me of that. Stupid me has always thought human beings are conscious intelligent agents... Given that computer techonlogy is really in it’s infancy (compared to, say, cars) I’d not go around making such claims just yet. Well, I go around and take my risks. Different personalities, maybe... Then your claim that dFSCI can be used to determine the presence of consciousness or intelligence in computer generated text is not supported. Well, I have done that in my next post. I thought you had read it too. You can find anything with a random search if you have sufficient time and searchers. Logically, you can. But not empirically. Not if you need much nore than the whole resource of the univers to just get a small probability of success. I don't think you have been discussin ID for so long, and still don't understand that simple difference. Then dFSCI measures nothing except how long a given string is and the size of the possibility space that string occupy s given all other possibilities. If dFSCI will give the same (or very similar) value for a children book as it would to a paper by Einstein if both were the same length then it’s not really measuring “complexity” at all is it? Yes, it is. It is measuring probabilistic complexity, not conceptual complexity. Nobody has ever said anything of that kind. The complexity is probabilistic, not conceptual. The definition of the function can include sone perspective about its conceptual value, but that is not implied in the computation of the probabilistic complexity. If I can write out “The wall is blue, the wall is red, the wall is green, the wall is blue and red, the wall is green and red” a million times that has “meaning”. So it has dFSCI. And if I write it out so it’s longer then Einsteins seminal paper the it’s dFSCI will also be higher. Well, forst of all it is repetitive, therefore algorithmically compressible. But if you are writing a detailed description of some not too brilliant machine, yes, it can have the same complexity of some paper from Einstein of similar length. I don't understand why you expected anything different. You have obviously never understood the meaning of dFSCI, or of CSI, if you think it should be related to the conceptual depth of the content. Which is absurd. It's not absurd at all. It's only you who have not understood. The probability of getting the Einstein paper and the machine description thorugh a random search are similar. Sure, but I’m getting the impression here that much of what you are talking about is “possible in principle but not in practice”. Would it not have been better to work out those problems before making the claim that dFSCI can determine if a given text was written by a conscious entity or not? The reason why dFSCI increases with the length of text was intuitively obvious to me. But, as you have requested, I have given a quantitative approach in my following post. Sure, but how to quantify that? Can I write a program that does it? Well, it is rather easy to write a text in a way that losing words compromises intelligibility. Your summary was a good example of that. We are looking at measurements of orders of magnitude anyway, we need not complete precision. A demonstration would probably be useful here then. In my following post I have given a demonstration of how dFSCI depends on length. That different "resistance" of particular words or phrases to random variation can be a cause of variability, given a certain length of text, is rather obvious. Yet you have to include bits (information) not present in the thing you are measuring in order to be able to make that determination. Does the additional information (of *knowing* what the function is of a given artifact) get included in the calculation of dFSCI for the artifact? If not, why not? I don't understand what you mean. The function is defined by a conscious observer, and measured in some objective way, for instance in a lab, for proteins. What has that to do with adding information to the claculation? What we are measuring is the information in the objetc, not the information necessary to recognize the information in the object. Please, be more clear. Yet a string could have many functions, some more complex then others. A strings function could also depend on other factors then the string itself (before 12 it unlocks the door, after 12 it sounds an alarm). That's perfectly OK. We can compute dFSCI for each defined function, and the value will be different for each function. That's not a problem at all. You see, the value of dFSCI can be seen only in a definite epistemologic and methodologic context. For instance, when we compute the dFSCI of a protein in relation to a specific biochemical function, it's because we want to explain the emergence of that function. It's the function that is complex, not the object. If an object has differnt functions, each function will have its complexity. If all those functions are implemented in the object, what is the problem? We are interested in the complexity of the function that is interesting for our reasoning. But then in relation to dFSCI proteins cannot be treated in that way. Proteins are not found via a random search in actual biology. Sure there is a random component but it’s not a random search. It’s not a tornado in a junkyard. That is all another question. I have applied dFSCI to the same problem studied by Durston and Axe: the emergence of basic protein domains. For a detaile discussion about that (almost complete), please look here: https://uncommondescent.com/intelligent-design/evolutionist-youre-misrepresenting-natural-selection/ Posts 33 and following. Anyway, I apply the concept of dFSCI only to the transitions for which no algorithmic explanation has ever been shown, and that can, at the actual state of our knowledge, be explained only as the product of RV or design. That's one of the points in my original definition of dFSCI. If dFSCI does not take into account the *actual* process that created a protein of what relevance is it to the protein itself? dFSCI takes into account all that is known about the process that created a protein. Nothing is known about the process that created basic protein domains. The paper you quote, finally, is about evolution in a superfamily, not about the emergence of a superfamily.gpuccio
January 20, 2012
January
01
Jan
20
20
2012
08:36 AM
8
08
36
AM
PDT
gpuccio,
Just some more support for my concept that dFSCI increases with length increase, this time empirical and regarding the protein space.
I do believe you. But that's the problem. I can write a book with minimal meaning and as long as it's sufficiently long it'll apparently be more complex (have a higher value for dFSCI) then a much shorter text which much more inherent meaning. Harry Potter might be a long book but it can't have more dFSCI then the special theory of relativity, or can it? Or to put it another way, is a very very large protein that does almost nothing more complex (higher dFSCI value) then a very simple one that does alot?
SO, empirically and for the protein space, values of dFSCI are strongly related to sequence length.
But I'd expect that given everything you've said so far. What I'd like to see is two proteins of identical (or close) length, where one does almost nothing and the other that does a lot, and presumably the dFSCI values for them would be very different?
I wish I could show you the scatterplot, but I am not sure is I can upload a picture to a post.
Upload it to imageshack.us and just post the URL. I suppose the problem I have with any purported measure of dFSCI is this. We examine X. We find no function for X We calculate the dFSCI as zero (presumably). Then suddenly we discover it's function. Then suddenly it has a dFSCI value greater then 0. Yet nothing whatsoever has changed with X, it's only external to X that something has changed. So dFSCI is not really measuring anything inherent in X at all. It's measuring "X+what we need to know about it in order to determine X is a thing that has function and so dFSCI". But everything after X is not used in the calculation, and it's the most crucial part!Peter Griffin
January 20, 2012
January
01
Jan
20
20
2012
07:53 AM
7
07
53
AM
PDT
Joe,
Yes Peter, a protein that can be any amino acid sequence would have zero or close to zero dfsci.
Please show your working! That's simply a claim, until you demonstrate it is but an empty claim.
Even something that doesn’t exist?
Depends on what you mean by "exist" really. Do books exist before they are written, out there in configuration space?
Is that supposed to be a refutation? LoL!
No, it's a simple question as to why a quantity in one context must be renamed in a different context if they in fact represent exactly the same metric. What's your response? Do you have one?Peter Griffin
January 20, 2012
January
01
Jan
20
20
2012
07:44 AM
7
07
44
AM
PDT
Yes Peter, a protein that can be any amino acid sequence would have zero or close to zero dfsci.
You can find anything with a random search if you have sufficient time and searchers.
Even something that doesn't exist?
Except I don’t believe he has, really. He’s calculated something and named it functional sequence complexity. You say that dFSCI is exactly the same thing? Then why not use the same name for the same thing?
Is that supposed to be a refutation? LoL!Joe
January 20, 2012
January
01
Jan
20
20
2012
07:19 AM
7
07
19
AM
PDT
gpuccio,
Durston has actually calculated dFSCI for proteins.
Except I don't believe he has, really. He's calculated something and named it functional sequence complexity. You say that dFSCI is exactly the same thing? Then why not use the same name for the same thing?
FSCI is not my invention at all. I have only given a detailed definition of the digital variant, including some clarification of how the function can be defined, and discussed many times aspects of its application to the biological models.
Sure, but you have not used it to calculate the dFSCI for a protein. And that is all I asked. My follow up question as to when you *will* be calculating it and *why* you would calculate it still remain!
If I cannot recognize a function, my empirical conclusion would be that there is no dFSCI, or, more correctly, that I cannot recognize any dFSCI, and therefore I cannot infer design.
So, you would agree that a protein could have no dFSCI then?
Still, it is an empirical fact that computers cannot output long meaningful and original text of any kind, not even by algorithmic search, least of all by a random search, least of all any specified text with any specified meaning.
Is it? Then presumably the same can be said of human beings, who are really just programs too. There are many people working on the problem, and many resources available. http://www.evolutionzone.com/kulturezone/c-g.writing/index_body.html Given that computer techonlogy is really in it's infancy (compared to, say, cars) I'd not go around making such claims just yet.
And yes, I have not done that.
Then your claim that dFSCI can be used to determine the presence of consciousness or intelligence in computer generated text is not supported.
least of all by a random search
You can find anything with a random search if you have sufficient time and searchers. http://www.jesse-anderson.com/2011/09/a-few-million-monkeys-randomly-recreate-shakespeare/
No, it will probably be similar. dFSCI depends essentially on length and on how much language can undergo random modification while still conveying its meaning. The nature of the meaning is not important.
Then dFSCI measures nothing except how long a given string is and the size of the possibility space that string occupy s given all other possibilities. If dFSCI will give the same (or very similar) value for a children book as it would to a paper by Einstein if both were the same length then it's not really measuring "complexity" at all is it? If I can write out "The wall is blue, the wall is red, the wall is green, the wall is blue and red, the wall is green and red" a million times that has "meaning". So it has dFSCI. And if I write it out so it's longer then Einsteins seminal paper the it's dFSCI will also be higher. Which is absurd.
What can nake a difference is if the emaning is “dense” in the text or not. In your two phrases, even the loss of one or two words would imply loss of part of the meaning.
Sure, but I'm getting the impression here that much of what you are talking about is "possible in principle but not in practice". Would it not have been better to work out those problems before making the claim that dFSCI can determine if a given text was written by a conscious entity or not?
But in a text there can be parts that are not really essential for the function as we have defined it.
Sure, but how to quantify that? Can I write a program that does it?
But if the text we start with is conveying the desired information in the simplest way it can be conveyed, the values of dFSCI should be similar for texts of given length, with some variability due to the different “resistance” to random variability in specific words or in specific phrases.
A demonstration would probably be useful here then.
The important point is that dFSCI does not depemd on the nature of the defined function (either it is to describe a law of physics or the desitny of a fictional character), but it does depend on how many bits are necessary to implement that function. That is perfectly true both for proteins and text.
Yet you have to include bits (information) not present in the thing you are measuring in order to be able to make that determination. Does the additional information (of *knowing* what the function is of a given artifact) get included in the calculation of dFSCI for the artifact? If not, why not?
dFSCI is a measure of the functional complexity of a digital string in relation to a function recognized for the string.
Yet a string could have many functions, some more complex then others. A strings function could also depend on other factors then the string itself (before 12 it unlocks the door, after 12 it sounds an alarm).
That is measured as the probability to get a string that implements the defined function by a random search.
But then in relation to dFSCI proteins cannot be treated in that way. Proteins are not found via a random search in actual biology. Sure there is a random component but it's not a random search. It's not a tornado in a junkyard. If dFSCI does not take into account the *actual* process that created a protein of what relevance is it to the protein itself? What might seem massively unlikely when viewed from the top might be inevitable front he bottom. Some recent research might help illustrate my point:
We show, contrary to current belief, that the ancestral NR was a ligand-activated transcriptional activator that existed in the earliest period of animal evolution. Our analysis reveals how the extraordinary functional diversity of modern receptors was generated by subtle tinkering with this ancestral template—slightly reshaping the ligand cavity, stabilizing the protein's active conformation so it no longer required a ligand, or disabling the protein's capacity to activate transcription without affecting its other properties. We predict that, when sufficient data are gathered to allow detailed evolutionary reconstructions in other protein families, it will become apparent that most protein functional diversity evolved by tinkering with ancient functions; invoking the evolution of wholesale “novelty” will seldom be necessary.
http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1000497 The last line is especially relevant. If you have a precursor that can be tweaked, of what relevant are random searches across the entirety of possibility space?Peter Griffin
January 20, 2012
January
01
Jan
20
20
2012
07:09 AM
7
07
09
AM
PDT
Peter: Just some more support for my concept that dFSCI increases with length increase, this time empirical and regarding the protein space. I have taken the data from Durston's table giving values for 35 protein families, and I have performed a linear regression of dFSCI against sequence length. Here are the results for the regression: Multiple R-squared: 0.8105, Adjusted R-squared: 0.8047 F-statistic: 141.1 on 1 and 33 DF, p-value: 1.836e-13 SO, empirically and for the protein space, values of dFSCI are strongly related to sequence length. I wish I could show you the scatterplot, but I am not sure is I can upload a picture to a post.gpuccio
January 19, 2012
January
01
Jan
19
19
2012
08:40 AM
8
08
40
AM
PDT
Peter: As you have prompted me, I would like to propose a simple reasoning to show how, in the case of text, dFSCI necessarily increases with length. Let’s take your summary of Hamlet. The play, set in the Kingdom of Denmark, recounts how Prince Hamlet exacts revenge on his uncle Claudius, firstly for murdering the old King Hamlet (Claudius’s brother and Prince Hamlet’s father) and secondly for then succeeding to the throne and marrying Gertrude (King Hamlet’s widow and mother of Prince Hamlet). It is 315 characters long. It conveys a very dense meaning about the Hamlet play (indeed, a well done summary of the plot at a very general level). Now, let’s divide it into two shorter parts. The first one, up to “father)” is 196 characters long. The second part is 119 characters long. Both parts are necessary to convey the meaning as defined (a complete summary of Hamlet, with all the information we can derive from it as you have proposed it). Now, assuming an alphabet of 30 symbols, the search space for the first part is 30^119, that is about 961 bits. The search space for the second part is 30^119, that is about 584 bits. The total search space is 30^315, that is about 1545 bits. Now, let’s say that we have calculated the functional space for the first part, defining as function the transmission of the meaning of that part only (or of any equivalent message of the same length), and that we have determined that the number of intelligible strings of that length that still convey the full meaning is, say, 1.6 x 10^60, that is about 200 bits. It’s not important is that is true or not, it’s just an example. Let’s say, similarly, that the number of functional sequences for the second part is 1.4 x 10^45, that is about 150 bits. Now, according to my definition, the dFSCI of the first part is 761 bits, and the dFSCI of the second part is 434 bits. If we define the function as conveying the full meaning of the full text, we can reason that both parts must be intelligible if we want to achieve the function. Therefore, the number of functional strings that convey the full function will be the product of the number of intelligible functional strings for the first part by the number of intelligible functional strings for the second part. IOWs, the number of functional strings will be 2^350, and the functional space will be of 350 bits. Therefore, dFSCI for the whole text will be 1195 bits. As you can see, dFSCI is bound to increase with the increase of legth, which was exactly my point. Now, let’s say that we have a segment of text, dense and specified, of say 100 characters. Let’s say that the functional combination of intelligible text conveying the meaning of the orioginal text are no more that 1:1000 (something that is not obly obvious, but easy to prove by any computer simulation). Let’s say, even contrary to the evidence, that they are 1:1000. therefore, the dFSCI for that segment will be approximately 10 bits. According to the previous reasoning, if we increase the length of the text, keeping it dense in meaning, we will gain about 10 bits of dFSCI for each new 100 characters. So, for 1000 characters dFSCI will be about 1000 bits, and for 10000 characters it will be about 1000 bits, that is well beyond Dembski’s UPB. That result will never be obtained by a computer by a random search. And this is only a lower threshold of dFSCI for text, based on an assumption of target space that is evidently too big. So, I believe that my “safe” proposal of 100000 characters was even too “safe”, and that my statement about 1500 characters was more than reasonable. Moreover, as there is no limit to the length of meaningful, dense text that humans can easily output, you can easily see that computers certainly cannot output dFSCI, while humans easily do it.gpuccio
January 19, 2012
January
01
Jan
19
19
2012
08:01 AM
8
08
01
AM
PDT
Peter Griffin: So, please correct me if I’m wrong, but you have not actually then calculated dFSCI for anything at all, text or protein? Durston has actually calculated dFSCI for proteins. FSCI is not my invention at all. I have only given a detailed definition of the digital variant, including some clarification of how the function can be defined, and discussed many times aspects of its application to the biological models. I refer to my definition, because I know it, it is detailed and IMO complete and consistent. But the fundamental concepts are derived from Dembski and others. Then dFSCI depends on the observers knowledge and that alone, whereas the meaning in the data remains the meaning regardless. For example. A block of data is received at Nasa HQ from a point in space very far away via radio. It is very large. It is very uncompressible. It is repeated many times over many days. You can’t make sense of it in any way.Would you conclude this data block contained no dFSCI? That's perfectly correct. If I cannot recognize a function, my empirical conclusion would be that there is no dFSCI, or, more correctly, that I cannot recognize any dFSCI, and therefore I cannot infer design. That would be one of the many false negatives. I searched the paper mentioned and found no FSCI or dFSCI mentioned. Why is that? Durston uses the term FSC (functional sequence complexity). It's exactly the same thing. You certainly gave the impression that dFSCI could be calculated easily to test the output of computer programs for consciousness. Yet here we are! It is very easy. I have given you a safe length threshold for text (100000 bits) that IMO implies dFSCI beyond 500 bits. For text, that's not too much: about 20000 characters. I have even stated that 1500 bits is enough. That no computer can generate a recognizable specified text of that length can be empirically verified. Obviously, it would be desireble to understand better the structure function relationship for text, and that can certainly be done, by studying how deeply text can be varied and still convey its meaning, and apply combinatorics to that. And yes, I have not done that. Still, it is an empirical fact that computers cannot output long meaningful and original text of any kind, not even by algorithmic search, least of all by a random search, least of all any specified text with any specified meaning. Sure. How long do you have? It might take some time! OK, go on... If you really count that as empirical confirmation that your inference is correct I’m bloody glad you are not my doctor! Well, I am bloody glad you are not my patient. When you calculate the dFSCI will it be different for the two strings? How different? No, it will probably be similar. dFSCI depends essentially on length and on how much language can undergo random modification while still conveying its meaning. The nature of the meaning is not important. What can nake a difference is if the emaning is "dense" in the text or not. In your two phrases, even the loss of one or two words would imply loss of part of the meaning. But in a text there can be parts that are not really essential for the function as we have defined it. But if the text we start with is conveying the desired information in the simplest way it can be conveyed, the values of dFSCI should be similar for texts of given length, with some variability due to the different "resistance" to random variability in specific words or in specific phrases. That would probably be, for long texts, a random effect. The important point is that dFSCI does not depemd on the nature of the defined function (either it is to describe a law of physics or the desitny of a fictional character), but it does depend on how many bits are necessary to implement that function. That is perfectly true both for proteins and text. What is dFSCI really measuring I wonder? You wonder without reason. I have clearly said it many times. I repeat it here: dFSCI is a measure of the functional complexity of a digital string in relation to a function recognized for the string. That is measured as the probability to get a string that implements the defined function by a random search.gpuccio
January 19, 2012
January
01
Jan
19
19
2012
01:54 AM
1
01
54
AM
PDT
gpuccio,
Not true. I have given the values of dFSCI for many family proteins, taking those values from Durston. I have never been engaged in the operative study of text, as I have often said
So, please correct me if I'm wrong, but you have not actually then calculated dFSCI for anything at all, text or protein?
If I cannot recognize and define a function, I will conclude that there is no dFSCI.
Then dFSCI depends on the observers knowledge and that alone, whereas the meaning in the data remains the meaning regardless. For example. A block of data is received at Nasa HQ from a point in space very far away via radio. It is very large. It is very uncompressible. It is repeated many times over many days. You can't make sense of it in any way. Would you conclude this data block contained no dFSCI?
Because I know how it can be done for proteins, that is my field of interest, and if I worked with language, I would find ways to do it with language.
I searched the paper mentioned and found no FSCI or dFSCI mentioned. Why is that? You certainly gave the impression that dFSCI could be calculated easily to test the output of computer programs for consciousness. Yet here we are! Can I ask another question? When will you be calculating the dFSCI for a protein next? Why? What will you do with that information? Will you be publishing your work anywhere? I'm getting the impression you've never actually calculated dFSCI for anything you know!
You can easily falsify my inference by exhibiting a randomly generates text of that length that conveys all the meaning of the original text. If you cannot, I will receive empirical confirmation that my inference is correct.
Sure. How long do you have? It might take some time! If you really count that as empirical confirmation that your inference is correct I'm bloody glad you are not my doctor!
A) Confirm the text is understandable B) Calculate a possible “search space” based on the length of the string. C) If search space > X then design = True.
So, one last question. Here are two lines of text, A and B: A: The cat jumped over the mat and back and forth again! B: Energy equals mass times the speed of light squared. Let's say those strings continue in that vein up to the required size threshold to allow us to reliably infer design if crossed. When you calculate the dFSCI will it be different for the two strings? How different? What is dFSCI really measuring I wonder?Peter Griffin
January 18, 2012
January
01
Jan
18
18
2012
02:52 PM
2
02
52
PM
PDT
Peter Griffin: Yet any string of equivalent length will have the exact same value, 1500 bits, correct? The value of the search space is the same for that length of english text. It depends on the length of the string and on the alphabet. Yet your measure of complexity seems to give the same value for both the summary of Hamlet and a selection of random characters. I think you are a little bit confused here. The search space of that length oftext is about 1500 bits. The specified complexity of your summary of Hamlet is defined as the rate between the target space (the number of sequences of that length that can convey the full meaning) and the search space. A random text has no meaning and no function, therefore it has no specified complexity.
So what is the function of Hamlet? I asked you that and you said: It is to convey a specific meaning, what Abel calls “descriptive information”. We can define it at various levels of detail. That’s not a function, that’s your description of what a function is. So, specifically, what is the function of Hamlet?
I have defined the function explicitly. It is to convey a specific meaning. I ahve also suggested how to measure it: for exmaple, by verifying the understanding of the specific meanings in a group of reader, for example by a questionnaire. If you have read my definition of dFSCI, you should know that the function must be defined objectively by a conscious observer. Any explicit definiton is fine, and the dFSCI will be measured in relation to that definition. TRhe function of your summary is to convey the basic meanings it contains. The function of Hamlet is to convey the full sum of the information, meaning, emotion and beauty that it can convey. As I have said from the beginning, descriptive information is more indirect than prescriptive information. Presciptive information generates a functional object, whose function can be measured in the lab. Prescriptive information conveys meanings, and requires conscious observers to understand them. So given string A: The summary of Hamlet And string B: A string made of English words of equal length but selected randomly Can you please calculate the dFSCI of string B. I don't understand your question. If string B is random, I suppose it doe nor mean anything. Then we cannot define a function for it. If you can, please make an example, give your definition, and we will see. Nowhere have you calculated a specific dFSCI value! Not true. I have given the values of dFSCI for many family proteins, taking those values from Durston. I have never been engaged in the operative study of text, as I have often said. The principles are the same, but you ahve to work to those specific problems of language to give answers. I have suggested a reasonable approximation for language, that can be verified empirically. Yet if you can’t measure dFSCI how can you tell if it’s being generated by a computer or not? You are really confused. I have suggested that a threshold of 100000 bits is enough to guarantee that dFSCI is there for language. Therefore, I am explicitly assuming that computers cannot generate 100000bits of functional language. As I am giving an approximation by lenth, I am taking the risk: take any text of 100000 bits of length, and I will infer it is designed. Obviously, the true dFSCI will be lower, because the target space is not one. That should make it easier for a computer to generate a text that conveys the same meaning as the text we have chosen. But as I am safe that my approximation is sound, and that at those levels of search space the order of magnitude of the exact dFSCI is not much lower than the search space, I infer design. You can easily falsify my inference by exhibiting a randomly generates text of that length that conveys all the meaning of the original text. If you cannot, I will receive empirical confirmation that my inference is correct. If you can’t calculate an exact value you have no actual idea if it is beyond a given threshold. All you’ve done so far is to A) Confirm the text is understandable B) Calculate a possible “search space” based on the length of the string. C) If search space > X then design = True. Nowhere have you calculated a specific dFSCI value! Again, I have given specific approximations for proteins. I have given you a general approximation based on length for text. It works. It will be empirically confirmed in all contexts. You don't understand that dFSCI is a tool to distinguish designed things from random outputs. It works, even in approximation. Ah, now we get to the heart of it. So a random string of the same length of Hamlet has the same dFSCI value. Are you kidding? dFSCI is the complexity necessary to convey a defined meaning. What is the meaning your random string conveys? If you cannot define a menaing, your text has no function, and does not exhibit dFSCI. Which means that dFSCI tells us absolutely nothing about Hamlet other then it’s length. Again, completely wrong. Have you at least read my posts? But why does a “specific meaning” come into it at all? That would seem to indicate that dFSCI is not an independent measure of complexity at all. Indeed it isn't. I have said a lot of times that any measurement of dFSCI is relative to an explicitly defined function. If you change the definition of the function, the measurement of dFSCI will change. But I really think that you have not read my posts. If I gave Hamlet to a person who had no written language at all it would have no specific meaning and if asked them to calculate the dFSCI they would say there was none. Exactly! Now you are beginning to understand. If I cannot recognize and define a function, I will conclude that there is no dFSCI. That is one of the reasons for the many false negatives. The other is that many designed things are simple. But, do you remember? the purpose of dFSCI is to be specific, not sensitive. It gives no false positives, but a lot of false negatives. So in summary, you must know that something is designed (has a specific meaning that is meaningful to you) in order to apply dFSCI and determine it is designed or not. Complete nonsense! I must recognize a function (or meaning) in an output, and verify that it requires enough complexity to be implemented that it cannot be generated in a random system. Then I infer that the output is designed. Would you say that it is the same as what you wrote? And I’m rather positive that I’ve invented a cold fusion machine, but until I demonstrate it my claims are empty! But my tool works: a computer cannot geberate your summary. That can be empirically verified. Sure. I’ll generate it in Kitanemuk, a long dead language. But it’ll have meaning. The second example will be a random string, also in Kitanemuk. Calculating the dFSCI values will no doubt allow you to determine which is which, correct? You are repeating your misunderstandings. I hope I have clarified why you are wrong. But if you like to go on being wrong, be my guest. Give it a go anyway? If you can’t do it then who can? And if you can’t do it as an expert on dFSCI how can you then say with a straight face “you will find dFSCI a wonderful tool for research.” Because I know how it can be done for proteins, that is my field of interest, and if I worked with language, I would find ways to do it with language. A suggestion: start with short words or phrases, and study how much a short phrase can randomly change remaining fully recognizable. That wouls be similar to the study made by Axe for proteins. Good night...gpuccio
January 18, 2012
January
01
Jan
18
18
2012
02:32 PM
2
02
32
PM
PDT
I would like to know what is the shortest functional coding sequence in living things. Also, what is the shortest change in a copied sequence that would give rise to a different function. Also, what is the shortest change to a random sequence that could give rise to function. Also, what is the shortest sequence that could be functional in a regulatory network. Obviously, those who infer design must have the answers to these questions, else they couldn't calculate probability bounds.Petrushka
January 18, 2012
January
01
Jan
18
18
2012
10:43 AM
10
10
43
AM
PDT
Gpuccio,
That’s easy. Assuming an english alphabet of 26 characters, plus spaces and punctuation marks, it is about 1500 bits.
Yet any string of equivalent length will have the exact same value, 1500 bits, correct?
No. That a measure is difficult in some specific cases is in no way a limitation of its value.
Yet your measure of complexity seems to give the same value for both the summary of Hamlet and a selection of random characters.
In principle dFSCI can always be measured, because you have to gove an explicit definition of the function and of how to measure it.
So what is the function of Hamlet? I asked you that and you said:
It is to convey a specific meaning, what Abel calls “descriptive information”. We can define it at various levels of detail.
That's not a function, that's your description of what a function is. So, specifically, what is the function of Hamlet? I asked:
If you can only “in principle” measure empirically the functional space how can you possibly know that computers cannot generate new dFSCI if you can’t actually measure it in practice?
You answered:
Well, if a computer can randomly generate your summary of Hamlet, or any variant of it that conveys objectively all its content to ten readers, I would be impressed.
That is not responsive to my question. So I ask again, how can you possible know that computers cannot generate new dFSCI if you can’t measure it in practice?
dFSCI is a function of the length, certainly, but not only of it. It is a function of the functional space too, and of the rate of the two. And of the form of information (scarcely compressible).
So given string A: The summary of Hamlet And string B: A string made of English words of equal length but selected randomly Can you please calculate the dFSCI of string B.
It is not necessary to compute an exact value of dFSCI for each case.
Yes, but the question I'm asking is can you calculate it exactly just the once! All you have said is:
Your summary, with a search space at 1500 bits, is almost certainly already there.
What is the specific dFSCI for the summary of Hamlet? Search space is one thing, it's just a function of the length of the text. But what is the specific value of dFSCI for that summary?
I am rathr sure that your summary is already long enough to infer design
You being sure is not science. When I can independently calculate the dFSCI in that piece of text and come up with the exact same answer (as you would expect) then that's science. Or at least a start towards it.
After all, we are not trying to know if an object was designed in all cases. We are happy to affirm the design inference in some cases. We want specificity, not sensibility.
Yet in this case I would like to see a specific dFSCI value for the text given.
So, I invite you to show some piece of english language of at least 100000 bits of length that coveys a very specific meaning, and which was generated by a computer starting from random variation. That would falsify my empirical threshold.
Yet no mention of dFSCI in your challenge? Once you have explained to me how to calculate dFSCI for a given example of text then I will get to work on your challenge.
Why? It’s enough that it is beyond an appropriate threshold, empirically verified. I have faith that kindergarten inhabitants are reasonable people, and would understand that.
If you can't calculate an exact value you have no actual idea if it is beyond a given threshold. All you've done so far is to A) Confirm the text is understandable B) Calculate a possible "search space" based on the length of the string. C) If search space > X then design = True. Nowhere have you calculated a specific dFSCI value!
. I don’t know if any similar procedure is available for text, but anyway we can take a reasonable length threshold, empirically tested, and use it.
Yet a few posts ago you made that exact claim!
I strongly believe that the failure of computers in generating true creative language output is an empirical demonstration that they cannot generate new dFSCI.
And
One of the most outstanding behaviours of conscious intelligent beings that cannot be simulated by computation is the generation of new dFSCI.
Yet if you can't measure dFSCI how can you tell if it's being generated by a computer or not?
dFSCI measures only one thing, in all cases: the probability to get a digital sequence that satisfies the objectively defined function by a random search.
Ah, now we get to the heart of it. So a random string of the same length of Hamlet has the same dFSCI value. Which means that dFSCI tells us absolutely nothing about Hamlet other then it's length.
The only thing I cannot do is c), but as already said I am not a text scholar. I believe that reasonable ideas about how to do that can be developed. In the meantime, I am sure that a threshold of, say, 100000 bits of pure length, conveying a specific meaning in the simplest way, will be found to be a criterion that empirically gives no false positives in designed text recognition.
But why does a "specific meaning" come into it at all? That would seem to indicate that dFSCI is not an independent measure of complexity at all. If I gave Hamlet to a person who had no written language at all it would have no specific meaning and if asked them to calculate the dFSCI they would say there was none. So in summary, you must know that something is designed (has a specific meaning that is meaningful to you) in order to apply dFSCI and determine it is designed or not.
I am rather positive that your summary of Hamlet already exhibits dFSCI, with its 1500 bits of condensed length.
And I'm rather positive that I've invented a cold fusion machine, but until I demonstrate it my claims are empty!
You say those values are not useful. Prove it! Show me any example of text of that length and functionality that is generated by a computer from random variation.
Sure. I'll generate it in Kitanemuk, a long dead language. But it'll have meaning. The second example will be a random string, also in Kitanemuk. Calculating the dFSCI values will no doubt allow you to determine which is which, correct?
But if you want to focus on text, yio will find dFSCI a wonderful tool for research.
Once you explain how to calculate it for the summary of Hamlet, I'm sure I will!
The only thing I cannot do is c), but as already said I am not a text scholar.
Give it a go anyway? If you can't do it then who can? And if you can't do it as an expert on dFSCI how can you then say with a straight face "you will find dFSCI a wonderful tool for research."Peter Griffin
January 18, 2012
January
01
Jan
18
18
2012
04:01 AM
4
04
01
AM
PDT
Peter Griffin: Sure, so we we can turn that round then and say “if this example output has dFSCI it *must* be from a conscious entity”? Yes, but the correct way to say it is: "If this example output has dFSCI, a design inference is the best empirical explanation for it". Design inference is an empirical inference, not a logical deduction. So, words like "must" are not appropriate epistemologically. So while we cannot use it to determine the lack of consciousness we can absolutely use it to determine the presence of consciousness, correct? Yes, in the empirical sense outlined above, definitely yes. What is the function of Hamlet then? It is to convey a specific meaning, what Abel calls "descriptive information". We can define it at various levels of detail. I must remind you that dFSCI is computed according to a specific definition of function, and its value is relative to the definition. Any functional definition can be used, provided it is objectively defined and measurable. Agreed? Yes. That would not be Hamlet, but a general description of the plot, with not much detail in it. If that's your function, to convey that level of detail, it's finne with me. Of course, that summary is much less complex then Hamlet itself. Please do that for that summary then, that would be useful. That's easy. Assuming an english alphabet of 26 characters, plus spaces and punctuation marks, it is about 1500 bits. Yet now you fall back to “in principle it’s possible”? That’s a very different claim to the ones you have made earlier in this thread, as just quoted. No. That a measure is difficult in some specific cases is in no way a limitation of its value. In principle dFSCI can always be measured, because you have to gove an explicit definition of the function and of how to measure it. Obviously, in the case of very big search spaces, an individual measure of each state is practically impossible. But there are other ways to bypass that. First of all, we can investigate the relationship between search space and functional space in shorte sequences, and get an idea of its formal properties and behaviour. That is certainlt possible not only in principle, but also in practice. Obviously, it requires time and work. For the protein space, many are working at it, on both sides. Moreover, we can fins theoretical approximations that are reasonable and useful. For proteins, Durston'r method is one such approximation. As I told you before, I have not focused much on text, because my interest is protein function. But for the reasons I have given it seems rather obvious that most text beyond a certain length is well beyond any conventional threshold for dFSCI, be it my 150 bits for biological systems or Dembski's UPB of 500 bits. Your summary, with a search space at 1500 bits, is almost certainly already there. If you can only “in principle” measure empirically the functional space how can you possible know that computers cannot generate new dFSCI if you can’t measure it in practice? Well, if a computer can randomly generate your summary of Hamlet, or any variant of it that conveys objectively all its content to ten readers, I would be impressed. So dFSCI is simply a function of how long the original text is? That does not seem very useful to me. dFSCI is a function of the length, certainly, but not only of it. It is a function of the functional space too, and of the rate of the two. And of the form of information (scarcely compressible). But it is true that, increasing the length of the sequence, the ratio between functional space and search space becomes quickly smaller. That effect can have different rates in different functional spaces, but it happens anyway. The reason is simple: while random states increase exponentially with the increase of sequence length, functional states are constrained by many independent rules of necessity. How does that help you determine if a given example text is produced by a conscious entity or not? It is not necessary to compute an exact value of dFSCI for each case. What we need is a reasonable approximation for each search/functional space, and a reasonable threshold. For text, we can try to derive that threshold theoretically by investigating the functional space of language, or just assume a minimal length as a reasonable threshold. I am rathr sure that your summary is already long enough to infer design, but if we want to be safe we could put the threshold at 100000 bits, just to make an example. After all, we are not trying to know if an object was designed in all cases. We are happy to affirm the design inference in some cases. We want specificity, not sensibility. And any threshold we assume can be empirically verified (or falsified). So, I invite you to show some piece of english language of at least 100000 bits of length that coveys a very specific meaning, and which was generated by a computer starting from random variation. That would falsify my empirical threshold. So it seems that dFSCI is like obscenity – you know know it when you see it. I’m afraid “a lot of it” would not pass muster at kindergarten as an answer. Why? It's enough that it is beyond an appropriate threshold, empirically verified. I have faith that kindergarten inhabitants are reasonable people, and would understand that. If you can’t calculate the specific value for dFSCI for any text example can you really claim that it has *any* utility as a metric? It is very useful. It can categorize outputs according to a threshold. For proteins, we can use Dirston's method to measure different values of dFSCI in proteins of similar length. I don't know if any similar procedure is available for text, but anyway we can take a reasonable length threshold, empirically tested, and use it. Because, as said, dFSCI however increases with the increse of the length, both in language and in protein space. If dFSCI is simply a measure of the length of the text then it is useless for what your are proposing. As said, it is not "simply a measure of the length of the text". It is a measure of the ratio functional space/search space. But, if we at present have no satisfying way to approximate the functional space (but we can find ways), the sequence length can be an approzimation of dFSCI, and therefore be very useful. That is very simnple. What is your problem? It also seems to me that dFSCI does not measure anything about the text itself. dFSCI measures only one thing, in all cases: the probability to get a digital sequence that satisfies the objectively defined function by a random search. Can actually do that, not “in principle” but actually do what you claim is possible here? Step by step? The only thing I cannot do is c), but as already said I am not a text scholar. I believe that reasonable ideas about how to do that can be developed. In the meantime, I am sure that a threshold of, say, 100000 bits of pure length, conveying a specific meaning in the simplest way, will be found to be a criterion that empirically gives no false positives in designed text recognition. Alternatively, if you could provide a previously worked example of a specific piece of text and a specific value for the dFSCI in it that would also suffice. I am rather positive that your summary of Hamlet already exhibits dFSCI, with its 1500 bits of condensed length. I have raised the threshold to 100000, just to be on the safe side. You say those values are not useful. Prove it! Show me any example of text of that length and functionality that is generated by a computer from random variation. And again, I have aways discussed proteins here. Proteins are more relevant the evolution debate than text. So I have focused on them. But if you want to focus on text, yio will find dFSCI a wonderful tool for research.gpuccio
January 18, 2012
January
01
Jan
18
18
2012
03:37 AM
3
03
37
AM
PDT
gpuccio,
That is simply wrong. dFSCI, as I have defined it, is an empirical marker of design (that implies consciousness), but its absence does not mean that an object is not designed.
Sure, so we we can turn that round then and say "if this example output has dFSCI it *must* be from a conscious entity"? So while we cannot use it to determine the lack of consciousness we can absolutely use it to determine the presence of consciousness, correct?
But a function can be defined also for a piece of language: for example, the function can be defined as the ability to convey some specifically defined meanings, and it can be tested by using that piece of language with a sample of readers, and verifying that the connected meanings have been conveyed.
What is the function of Hamlet then?
a) I would try to express the meanings in the most simple way that is complete. That would be the startting functional object.
So the starting functional object for Hamlet could be this then?:
The play, set in the Kingdom of Denmark, recounts how Prince Hamlet exacts revenge on his uncle Claudius, firstly for murdering the old King Hamlet (Claudius's brother and Prince Hamlet's father) and secondly for then succeeding to the throne and marrying Gertrude (King Hamlet's widow and mother of Prince Hamlet).
Agreed?
b) I would define the search space as the complexity in bits of that piece of language.
Please do that for that summary then, that would be useful.
c) I woull try to measure empirically the functional space, IOWs I would try a mathemathical analysis of the possible variations of that text that still can convey the full meanings defined. I understand that is not easy, but in principle it can be done, and some large approximation could not be so difficult to achieve.
Earlier you said
I strongly believe that the failure of computers in generating true creative language output is an empirical demonstration that they cannot generate new dFSCI.
Yet now you fall back to "in principle it's possible"? That's a very different claim to the ones you have made earlier in this thread, as just quoted. If you can only "in principle" measure empirically the functional space how can you possible know that computers cannot generate new dFSCI if you can't measure it in practice?
But how much is the complexity? Well, if anybody here knows the length of the Hamlet text in ASCII, it is easy: starting from the byte value, it is 8^number of letters (including punctuation marks, spaces, etc.). Anybody wants to try to give the result?
So dFSCI is simply a function of how long the original text is? That does not seem very useful to me.
Now, Hamlet is about 150000 characters, with a search space complexity of about 700000 bits. Now, even if we define the function as giving a clear idea of the story, the characters, and the meanings in the play, whatever our evaluation of the target space, it seems not credible that the target space can be even near to the order of magnitude of 700000 bits.
How does that help you determine if a given example text is produced by a conscious entity or not?
As Dembski’s UPB is only 500 bits, I would safely affirm that Hamlet does exhibit dFSCI. And a lot of it.
So it seems that dFSCI is like obscenity - you know know it when you see it. I'm afraid "a lot of it" would not pass muster at kindergarten as an answer. If you can't calculate the specific value for dFSCI for any text example can you really claim that it has *any* utility as a metric? If dFSCI is simply a measure of the length of the text then it is useless for what your are proposing. It also seems to me that dFSCI does not measure anything about the text itself. Could you perhaps take the first act of Hamlet and
a) I would try to express the meanings in the most simple way that is complete. That would be the startting functional object. b) I would define the search space as the complexity in bits of that piece of language. c) I woull try to measure empirically the functional space, IOWs I would try a mathemathical analysis of the possible variations of that text that still can convey the full meanings defined. I understand that is not easy, but in principle it can be done, and some large approximation could not be so difficult to achieve.
Can actually do that, not "in principle" but actually do what you claim is possible here? Step by step? Alternatively, if you could provide a previously worked example of a specific piece of text and a specific value for the dFSCI in it that would also suffice.Peter Griffin
January 18, 2012
January
01
Jan
18
18
2012
02:26 AM
2
02
26
AM
PDT
Peter Griffin: OK, here I am. First of all a basci correction. You say: This method would presumably also have relevance to future Turing tests, if you can reliably determine that a given output has no dFSCI then by (your) definition it is not coming from a conscious entity. That is simply wrong. dFSCI, as I have defined it, is an empirical marker of design (that implies consciousness), but its absence does not mean that an object is not designed. dFSCI is defined so that it has no false positives. But it does have a lot of false negatives. But perhaps I should give you my explicit definition of dFSCI. As it is late now, I will just paste here form an old post of mine. Maybe not everything is here, but just to begin the discussion:
Of course consciousness in central in ID theory. ID is about detecting design in things. Design is a process which originates in conscious intelligent beings (the designers). ID affirms that designed objects are recognizable with certainty as such if they exhibit a specific property, CSI. CSI is the main idea in ID. It is objectively recognizable, and in the known world it is always the product of design by an intelligent conscious being (leaving apart biological information, which is the object of the discussion). A special subset of CSI, digital functionally specified complex information (or, if you want, dFSCI), is specially useful for the discussion. It is easily definable as any string of digital information with the following properties: complexity higher than 10^150 (that is, length of about 500 bits); non significant compressibility (it cannot be generated through laws of necessity from a simpler string); and a recognizable, objectively definable function. That definition is very strong and useful. According to that definition, dFSCI includes language, software and practically all relevant biological information (in particular, the sequences of protein coding genes and the primary sequences of proteins). It is easy to show that no example is known of dFSCI (apart from biological information, which is the object of the debate) whic does not originate from a cosncious intelligent being (humans). And our common experience is that consciousness and intelligence are exactly the faculties used by humans in producing dFSCI. Biological information is dFSCI (any functional protein is). That’s why ID, with very sound inference based on analogy, assumes that some conscious and intelligent designer is the origin of biological information. That is, very quickly, the main idea in ID. Neo-darwinism cannot explain the emergence of dFSCI in living beings. The work of a designer can. I would like to mention that dFSCI originates from conscious intelligent beings directly; or indirectly, through some non conscious machine which has received from an intelligent conscious being the pertinent dFSCI. In other words, Hamlet is dFSCI. Hamlet can be outputted by a PC, but only if someone has inputted it in the software. No computing machine can create Hamlet (or anything equivalent). Specification, function and purpose are definable only in relation to consciousness. Only consciousness recognizes them actively. So, consciousness is central to ID. Without consciousness, no function can be recognized. With consciousness, function can be defined, recognized and measured. And function is the only relevant form of specification in biological information.
Now, you ask about language. I usually use dFSCI for protein domains, but it is possible to use it quite well for language. The main difference between the two cases can be explained using Abel's terminology: a protein is a cse of prescriptive information, while language is usually descriptive information. As dFSCI is based on functional specification, it is more direct to use it for proteins, that have an easily definable biochemical function. But a function can be defined also for a piece of language: for example, the function can be defined as the ability to convey some specifically defined meanings, and it can be tested by using that piece of language with a sample of readers, and verifying that the connected meanings have been conveyed. The procedure is the same, at that point. The piece of language for which we defined the function is the object about which we want to measure dFSCI (for that function). Empirically, I would proceed in this way: a) I would try to express the meanings in the most simple way that is complete. That would be the startting functional object. b) I would define the search space as the complexity in bits of that piece of language. c) I woull try to measure empirically the functional space, IOWs I would try a mathemathical analysis of the possible variations of that text that still can convey the full meanings defined. I understand that is not easy, but in principle it can be done, and some large approximation could not be so difficult to achieve. For long pieces of text, it is rather onvious that the search space is hugely bigger than the target space, and I believe that above a certain number of bits the ration between the target space and the search space tends to be very similar, in order of magnitude, to 1/search space. d)For language, beyond a minimal complexity, I believe that no known simpler algorithm can generate the output. Therefore, the problem of compressibility is not particularly relevant. Again I paste from an old post:
Hamlet is what it is because it was thought, felt, represented and intuited by Shakespeare, and then and only then, converted to digital information which, when appraised by other conscious intelligent beings, can recreate in them, at least in part, the conscious experiences of the author’s mind. This point is fundamental, and no real understanding of ID can be achieved, IMO, without recognizing the central role of consciousness. Consciousness and conscious processes have to be considered as empirical realities, as facts which must contribute to our description of reality, because they are part of reality. Subjective experiences do exist, and nobody in the world can explain them in a reductive, materialistic way. It is no surprise that many materialists, probably aware in some way of the impossibility to explain consciousness in materialistic ways, just try to deny it. Take, for instance, the concept of function, which is a special form of purpose. It originates in consciousness. Matter has no purposes, as far as we can understand. The atoms in a machine are not aware in any way of the function they help perform. But it was a man who built the machine to realize his conscious, purposeful representations, and another man, seeing the machin performing its function, can usually understand what it is doing, and the purpose behind it, exactly as a man reading Hamlet can feel the perspective, the feelings, the genius which was in Shakespeare. Now, just follow me. Here comes the importance of CSI, or of the related concepts, like FSCI or digital FSCI. Design is always specified, but it can be simple, and it often is. I can design just a simple arrow. It is certainly designed, and in my consciousness it has a purpose, that is to point in one direction. But, in the absence of any context, an observer could doubt that the attow was designed, because it could easily be the result of casual forces. In other words, my arrow has a specification (it points in one direction), but it is not complex. It is designed, but it is not recognizable as such with any certainty. The problem is more clear if we shift the discussion to digital symbolic information. That’s why I always prefer to discuss digital FSCI, instead of CSI in general. So, let’s pretend we happen to observe a string of digital symbols (recognizable on a medium of some kind). Let’s say we observe a string of two symbols which could be interpreted as a sequence of 0s and 1s, in other words as a binary string. Let’s say we observe 8 “digits”, and that their sequence, if interpreted as 0s and 1s, corresponds to the byte which, in the ASCII convention, represents the letter “a”. Would that demonstrate that the string is designed? No. It could have been designed, but we don’t know for certain (unless we saw the author writing it down). The fact is, that string is specified (in a bibary code and in the ASCII interpretation of binary numbers), but it is not very complex. It is just 8 bits. The search space is only 2^8, that is 256 possible outputs. If our two symbols, which we interpret as 0 and 1, whatever their physical support may be, have been mixed up by some random non intelligent force, like the wind on a beach, or electromagnetic noise acting on a magnetic tape, there would still be one chance in 256 that a string of eight of those symbols represented the letter “a”. But let’s pretend that our string is much longer, and that its interpretation in ASCII corresponds exactly to the full text of Hamlet. Any serious inquirer would no more attribute it to random forces. Any serious inquirer would be certain that some conscious agent, a human being, wrote that string, either directly, or through some machine (computer or else), where anyway the string had been previously inputted. Whatever we can say, Hamlet does not come out by chance. Not by the wind in the beach, not by random electromagnetic noice. It does not originate by chance and never will, not in 15 billion years, not in 15 billion universes. Because Hamlet is the product of intelligent (and, let me say, wonderful, deep and exciting) conscious representations. And why are we so sure that the Hamlet string did not originate by chance? If we don’t understand it immediately, probably because in some way we have lost our common sense and our feelings, then reason and science may come to help us: we just have to compute the complexity of the string. The specification is already there: it is immediately recognizable to any human being who understands english (and, obviously, the intermediate ASCII code). Hamlet definitely “means” something: a lot of things, indeed. But how much is the complexity? Well, if anybody here knows the length of the Hamlet text in ASCII, it is easy: starting from the byte value, it is 8^number of letters (including punctuation marks, spaces, etc.). Anybody wants to try to give the result? So, no Hamlet without consciousness. And, obviously, you know very well how my discourse would go on from here…
Now, Hamlet is about 150000 characters, with a search space complexity of about 700000 bits. Now, even if we define the function as giving a clear idea of the story, the characters, and the meanings in the play, whatever our evaluation of the target space, it seems not credible that the target space can be even near to the order of magnitude of 700000 bits. As Dembski's UPB is only 500 bits, I would safely affirm that Hamlet does exhibit dFSCI. And a lot of it.gpuccio
January 17, 2012
January
01
Jan
17
17
2012
05:12 PM
5
05
12
PM
PDT
gpuccio,
One of the most outstanding behaviours of conscious intelligent beings that cannot be simulated by computation is the generation of new dFSCI.
Could you share with me your method of examining a section of text and determining the value of dFSCI present in it? Would it be possible, say, to take the first few posts on this thread and determine the value of dFSCI present in each, if it's not too much trouble?
b) Consciousness is the best explanation of why humans can output new dFSCI, and computers cannot.
This method would presumably also have relevance to future Turing tests, if you can reliably determine that a given output has no dFSCI then by (your) definition it is not coming from a conscious entity.
I strongly believe that the failure of computers in generating true creative language output is an empirical demonstration that they cannot generate new dFSCI.
I can't actually find any resources on the calculation of this metric so please provide a demonstration or link to future resources on how to specifically calculate dFSCI for distinct bodies of text. I also have some further questions about dFSCI but I'll save them for the moment. Most will be no doubt answered once I fully understand how to apply it to language output, as you seem to be able to do gpuccio.Peter Griffin
January 17, 2012
January
01
Jan
17
17
2012
03:34 PM
3
03
34
PM
PDT
No, you haven't been "too hard". I just think you've missed the mark completely :) Certainly if you are using "awareness" as synonymous with "experience" you are making my job easier. Clearly ASIMO is "aware" of the objects in its path - how else would it be able to avoid them? And that awareness is achieved in exactly the way we understand awareness is achieved in human beings. Awareness - attention - is an extremely well-studied field in neurscience (indeed, it's the field I work in). But clearly we are not going to agree on this, and to reach even the beginnings of an understanding will take more time than is possible over sporadic posts. I've been meaning to write a blog post about this for sometime, and probably will. I've also got the beginnings of a book! But there are a lot of prelimaries that are necessary to any discussion, not least being a clear definition of terms. But it's been nice to talk to you anyway. And you must admit, that video of ASIMO avoiding the pink bits of paper and trying to stand on the blue piece is rather cool :) But before I sign off, I should make one thing clear: I do not think that ASIMO has anything comparable to human experience, if we could even call what it has experience at all. I think what it has and does is simply far too impoverished to have any resemblance to the way we are and experience being. But I do think what it has is a first step on that road, which, IMO, requires a link between perception and action. I think that is the key to consciousness. Cheers :) LizzieElizabeth Liddle
January 17, 2012
January
01
Jan
17
17
2012
03:14 PM
3
03
14
PM
PDT
Elizabeth: No, I am not proposing a similar inference by analogy. I am merely demonstrating that your criterion for awareness is useless in determining whether or not a non-biological entity has experience. Why? The only way we concieve awareness and consciousness is iether by direct percetion in us or by inference in other beings. I have only tried to elucidate what the inference is based on. You cannot assume conscious experiences in other entities in other ways: only by an inference by analogy. That's exactly the point you try to bypass, without succeeding. What I am doing instead is to consider how what we call awareness arises in human beings (in whom we know they do, because we are one, as it were) not by simple observational correlations but by testing specific hypothesis derived from good theoretical models in actual experiments in which one variable is experimentally manipulated, and using the resulting information as to how awareness of the world arises in humans to build a model of it in the form of a robot who exhibits the predicted awareness of its environment. I am tempted to say that the clarity of this paragraph shows well the clarity of your reasoning, but perhaps that would not be very kind :) . I will just say that "using the resulting information as to how awareness of the world arises in humans" is a masterpiece of hidden and explicit unwarranted assumptions in one short sentence. Now, you’ve brought in another word: experience. I am not sure how you are distinguishing “experience” from “awareness” but I would regard “experience” as reflecting some kind of learning process. Well, I suppose I was using the two words as synonims (indeed, I was anyway referring to "conscious experiences"). Probably, there is a difference however: we could say that conscious expereinces are organized representations in an aware I. I am not specially interested in the concept of "learning process" because, as you have demonstrated, the words are often applied to non conscious processes in even trivial algorithms, and therefore they are in no way, if used so, characteristic of a conscious experience. I might be “aware” of an aversive stimulus, but once I have “experience” of that aversive stimulus, I can take evasive action. Whatever, provided that you are describing conscious processes here. Which, again, is what ASIMO does. No. Here is the nonsense in your reasoning. ASIMO does nothing lilke that. It is not aware of any aversive stimulus (it just processes in a certain way information from the outer world that it has beeen programmed to process that way. Nothing is aversive to it, because it does not feel pain). It learns how to navigate a constantly changing environment. Maybe, in the non conscious sense of "learning". But not because it wants to avoid the pain deriving from an aversive stimulus, or because it understands the meaning of the environment, or because it wants to achieve anything that will give it joy or satisfaction. In other words, it learns from experience, and becomes aware of any hazards and goals that enter that environment. Absolutely not. It can "learn" in the non conscious sense from inputted inforamtion (not conscious experience). And it never becomes aware of anything. Or, if not, something with exactly the same functions No. Just something that can grossly imitate an outer behaviour, without any of the inner phenomena that take place in conscious beings. If a robot does just what an animal does, It just imitates some gross behaviours, it certainly does not "do just what an animal does". I don't believe that a Turing test, even if really passed, would prove consciousness (as far as I know that was not even Turing's idea). But please remember that a true Turing test has never been passed. with exactly the same kind of circuitry as we think an animal has, Well, that's a very strong statement. Are you sure we know the "circuitry" of an animal? And are you already assuming that the "circuitry" explains or generates consciousness? And if not, how can you say that the "circuitry" works on the same way in a robot as in an animal? After all, in an animal the circuitry constantly interacts with conscious states. If the robot has not conscious states (and in no way you have demonstarted that it has them), then the circuitry cannot work in the same way. and exactly the same kind of learning processews, Here you are assuming that non cosncious learning processes are the same as conscious learning processes. A very strong and unwarranted assumption. Or, even worse, that learning processes that appear similar to conscious learning processes must necessarily be conscious. On what grounds do you withhold those words? Strange question indeed. I use words with a very specific and correct meaning, their menaing. You try continuosly to change that meaning, as though giving a different meaning to the same word could prove that the meanigns are the same. And you do that in good faith! Strange indeed. why would we doubt that it is doing the same thing? For a lot of good reasons, as shown above. Your reasoning is a bad version of a bad application of the Turing test to elementary animal behaviour. It takes in no serious consideration the empirical nature of conscious representations, and conflates arbitrarily and in multiple ways behaviour and awareness, with devastating cognitive results. Ehm, have I been too hard? :)gpuccio
January 17, 2012
January
01
Jan
17
17
2012
02:13 PM
2
02
13
PM
PDT
Now you are suggesting a similar inference by analogy for an entity that: a) Is not biological (the analogy of the body is lacking) b) Has a behaviour that very grossly imitates some simple features of our behaviour c) For which imitiation we have a very simple explanation: it was programmed by us to imitate those gross behaviours. Therefore, those “imitated” behaviours can be easily explained without any need for conscious subjective experiences. That said, I definitly refuse to make the inference by analogy of cosncious subjective experiences for your robot. I feel quite safe in assuming that it has none. So, as far as I can judge, your robot feels no pain, has no thoughts, has no purposes, has no fears, and understands no meaning. Your robot, very simply, is not a perceiving and acting subject.
No, I am not proposing a similar inference by analogy. I am merely demonstrating that your criterion for awareness is useless in determining whether or not a non-biological entity has experience. What I am doing instead is to consider how what we call awareness arises in human beings (in whom we know they do, because we are one, as it were) not by simple observational correlations but by testing specific hypothesis derived from good theoretical models in actual experiments in which one variable is experimentally manipulated, and using the resulting information as to how awareness of the world arises in humans to build a model of it in the form of a robot who exhibits the predicted awareness of its environment. Now, you've brought in another word: experience. I am not sure how you are distinguishing "experience" from "awareness" but I would regard "experience" as reflecting some kind of learning process. I might be "aware" of an aversive stimulus, but once I have "experience" of that aversive stimulus, I can take evasive action. Which, again, is what ASIMO does. It learns how to navigate a constantly changing environment. In other words, it learns from experience, and becomes aware of any hazards and goals that enter that environment. Or, if not, something with exactly the same functions :) On what grounds do you withhold those words? If a robot does just what an animal does, with exactly the same kind of circuitry as we think an animal has, and exactly the same kind of learning processews, why would we doubt that it is doing the same thing?Elizabeth Liddle
January 17, 2012
January
01
Jan
17
17
2012
09:49 AM
9
09
49
AM
PDT
Champignon: As far as I can tell, the only explanatory functions you have assigned to the immaterial soul/mind are 1) to explain consciousness and 2) to explain free will. Wherever have I used the words "immaterial soul" or "immaterial mind"? You are doing all by yourself. I have to quote myself: "First of all, I have never used the concept of “soul” in this discussion. You are bringing it in, I don’t know why. I have discussed, consciousness, because it is an empirical fact. If consciousness is a property of an immaterial souls is a philosophical question, and I have not discussed it." The same can be said for "mind". I have few clear ideas about those words: a) "Soul" is a philosophical and religious concept, related to the concept of substance substance, and cannot be approached scientifically. b) "Mind" is an ill defined and ambiguous concept. I rarely use it, and if I use it it's to indicate some functions operated by consciousness. c) "Immaterial" is ill defined and ambiguous, exactly like "material". I usually try to avoid those ambiguous concepts, like other even more dangerous, such as "natural". But, as I have alredy said, I have been discussing consciousness. The existence of consciuous subjective experiences, referred to a perceiving I. Consciousness is a fact. It is empirically certain. Therefore, it should be part of any scientific map of reality. But as I pointed out earlier, postulating an immaterial mind doesn’t solve the consciousness problem at all. Why must an immaterial mind give rise to consciousness? I have never, never tried to explain consciousness. Least of all by using the concepts of an immaterial mind. My point is that consciousness has not been explained, and IMO cannot be explained, as the product of some configuration of matter. Those who affirm nthat are wrong. Nothing in our knowledge support an explanation of consciousness based on configurations of matter. That said, I am perfectly fine with the statement that cosnciousness cannot be explained. Indeed, I believe that it needs not be explained. But it exists, and must be part of our map of reality. Even if we need not explain consciousness, we can certainly describe it and some of its formal properties, and investigate its connections with other expressions of reality, such as matter. because consciousness does interact with matter. Moreover, consciousness allows us to define many concepts that hyave sense only if referred to a conscious perceiver: meaning and purpose are among them, and they have special importance for ID theory, and for the concept of dFSCI. The concept itself of "function" cannot be defined without the use of a conscious perceiver. Those concepts are certainly of great scientific utility. If you can’t answer that question — and I haven’t seen any indication so far that anyone can — then an immaterial mind doesn’t explain consciousness. I perfectly agree with that. On the contrary, it's consciousness that can give meaning to some definition of "mind". But, as I said, I am not so interested in defining mind. I have never said that consciousness is immaterial (although I do believe that, at least in a very general meaning of "immaterial"). I have not even said that we can scientifically prove that consciousness can exist independently of matter, although I do believe that, and I believe that there are at least some scientific reasons to think that way (NDEs for example). What I have said is that nobody has ever scientifically proved, or even reasonably supported, the theory that consciousness can be explained as a configuratin of matter, and that consciousness cannot exist independently of matter. That is my point, and I am ready to defend it. As for free will, I have the impression from earlier comments of yours that you accept the idea of libertarian free will. That's absolutely correct. How do you know that libertarian free will exists? I don't know that scientifically. I have always stated very clearly that the problem of free will, IMO, cannot be solved scientifically (in that, it is completely different from the problem of design detection). Free will is a problem that can be approached only philosophically. I have debated it many times, and it has been explicitly a philosophical debate, not a scientific one (although there are obviously some important consequences at the scientific level, as it always happens). I have recently summarized some of my reasons to believe in free will in my answer to Mark here: https://uncommondescent.com/intelligent-design/have-materialists-lost-their-minds/ At post # 20.1.1.1.2 To be clear: a)I do believe that a belief in strict determinism can be held with some philosophical consistency, but that it implies extreme consequences that almost nobody would be really ready to accept entirely, and that it explains very badly reality, and especially the reality of consciousness. b) I absolutely believe that a belief in libertarian free will, if correctly conceived, can be held with wonderful philosophical consistency, and that it explains a lot of reality, especially of the reality of consciousness. c) I do believe, as I have already said, that compatibilism is an intellectual fraud, and complete nonsense. Finally, you say: I should add that the will itself can be disrupted by brain injury, so a putative immaterial mind can’t even be solely responsible for that. OK, I could agree with that (with all my reservations about the "immaterial mind".) And so? Does the immaterial mind do anything at all? If not, it’s superfluous, and Occam’s Razor slices it right off. I wilol slice off the immaterial mind myself with pleasure :). But I will not slice off consciousness, and neither will Occam's razor. Consciousness does a lot of things. a) Even if we do not believe in free will, consciousness is the best explanation of why human cognition is not completely algorithmic (see Penrose's argument about Godel's theorem). b) Consciousness is the best explanation of why humans can output new dFSCI, and computers cannot. c) Finally, if we believe in free will, consciousness is the best explanation of how we can change our personal destiny, of moral responsibility, and of all the sense of human existence. d) Consciousness is the only generator of meaning. Without meaning, no science, no cognition indeed, is possible. e) Consciousness is the only generator of purpose. Without purpose, no science, no oriented activity is possible. f) Consciousness is the seat of joy, that is the final and universal purpose of human existence. Is that enough?gpuccio
January 17, 2012
January
01
Jan
17
17
2012
07:11 AM
7
07
11
AM
PDT
Elizabeth: Elizabeth: The inference of conscious experiences in other humans is an inference by analogy. Three kinds of evidences are the basis for such inference: a) We have conscious experiences, that we know intuitively in ourselves. Therefore we know they exist, and we have some idea of what they look like, even if we can't explain them. So we create words to describe them: "conscious expereinces", "pain", "fear", "meaning", "purpose", and so on. b) We can observe (always thorugh our consciousness) that we have a body and an outer behaviour, and we can constantly observe a series of connections between our conscious representations and our behaviour. In particular, we can observe the strict connection between inputs coming through the body (sensations) and the corresponding conscious representations, and between conscious representations and the corrisponding behaviour of the body (actions). c) We can observe that other humans have similar bodies and similar behaviour. Therefore, we infer that they have similar subjective experiences. That inference is very stromg, and automatically shared by practially all. As you can see, the inference of subjective experiences is not simple, and any way is based on our subjectuive experiences. The point is: how similar to ours are the body and the behavious of the entity we observe? That is the second important point in the inference by analogy: the analogy. There, the judgement can be subjective: I fully agree in inferring consciousness of some kind in higher animals. Most people would probably agree. For lower animals, plants, bacteria, it really becomes very subjective. Now you are suggesting a similar inference by analogy for an entity that: a) Is not biological (the analogy of the body is lacking) b) Has a behaviour that very grossly imitates some simple features of our behaviour c) For which imitiation we have a very simple explanation: it was programmed by us to imitate those gross behaviours. Therefore, those "imitated" behaviours can be easily explained without any need for conscious subjective experiences. That said, I definitly refuse to make the inference by analogy of cosncious subjective experiences for your robot. I feel quite safe in assuming that it has none. So, as far as I can judge, your robot feels no pain, has no thoughts, has no purposes, has no fears, and understands no meaning. Your robot, very simply, is not a perceiving and acting subject.gpuccio
January 17, 2012
January
01
Jan
17
17
2012
05:22 AM
5
05
22
AM
PDT
Elizabeth is trying toi argue that only the object and the process exist, and that the subject is not necessary.
No. Clearly this is false. Why would I be suggesting a subjectless verb? I am not. I'm saying that something is "conscious" if, and when, it conches something. The verb has a perfectly clear subject, as I'm sure I made clear. Please address the post I did make, not the post you wish I'd made :)Elizabeth Liddle
January 16, 2012
January
01
Jan
16
16
2012
02:49 PM
2
02
49
PM
PDT
I have nothing more to add to this discussion. I have no solutions to the problem of consciousness. I will merely agree that declaring some aspect of mind or consciousness to be immaterial adds no new knowledge and creates new problems requiring tortuous explanations.Petrushka
January 16, 2012
January
01
Jan
16
16
2012
01:01 PM
1
01
01
PM
PDT
gpuccio, I should add that the will itself can be disrupted by brain injury, so a putative immaterial mind can't even be solely responsible for that. Does the immaterial mind do anything at all? If not, it's superfluous, and Occam's Razor slices it right off.champignon
January 16, 2012
January
01
Jan
16
16
2012
12:44 PM
12
12
44
PM
PDT
1 2 3 4

Leave a Reply