|
Author
|
Topic: Demonology and Biology
|
Janitor@MIT
Member
Member # 125
|
posted 27. March 2002 15:00
While some of the participants here wonder about the “surreptitious” intervention of an “Intelligent Designer” in natural history, I thought I would relate the subject (while noting the rather ironic confluence of ideas) to the surreptitious intervention of a “thermodynamic demon” in evolutionary biology. The intent is to make more concrete this shared idea, mostly because I’m a little uneasy with either “improbable coincidences” or “convenient interventions” as explanations either of evolution or design respectively.
For proprieties sake, evolutionary biologists don’t refer to the demon out loud. Nevertheless he figures prominently in their thinking (sometimes in disguise as “Darwin’s demon”). Many of the participants here are more forthcoming, although the “demon” becomes a sort of “demiurge.” Nonetheless, whether “demon” or “designer,” the problem is shared and central to both “worldviews.”
I don’t mean to pick on Tom Schneider, only because his name was prominent in a number of posts, but his ev program is illustrative of the problem:
“The main question arising from Maxwell's demon is whether the Second Law of thermodynamics can be violated by some kind of subtle sorting operation. The new form of the Second Law given by equation (26) shows that for every bit of sorting that the demon performs, he must pay at least emin joules of energy dissipation away from himself. He cannot win because that separation costs at least as much as it would be worth, as shown above. Thus the demon cannot violate the Second Law by doing sorting operations. But sorting is the only trick that the demon has available to him! The demon violates the Second Law by decreasing the entropy of a system without compensatory dissipation. Maxwell's demon is dead.”—From Thomas D. Schneider’s web site presentation, “Sequence Logos, Machine/Channel Capacity, Maxwell’s Demon, and Molecular Computers: a Review of the Theory of Molecular Machines,” 2000 Oct. 21.
So, Schneider is well aware that “Maxwell’s demon is dead.” But this demon has more lives than Schrodinger’s cat! Having performed the rites over the demon, Schneider resurrects him (in the form of the ev program) to perform the necessary entropy reducing, “subtle sorting operation,” he indicates the demon cannot do!
Without going into any detail, the usual procedure is faithfully reproduced in ev: a target is specified and the information discarded or lost in attaining the target is dropped from the equations. I.e., information is definitively a reciprocal. Information is “added” in these exercises by dropping the denominator. If some other form of “information” is intended, then it behooves the experimenter to make this explicit in his formulations. Otherwise, the whole exercise is self-falsifying by being self-fulfilling, so to speak.
(As I indicated, I’m not picking on Tom Schneider, his theory of molecular machines is an important contribution to bioenergetics, sequence logos are a clever representational device, and his derivation of the Second Law from the channel capacity theorem is utterly ingenious in its simplicity… Wish I’d thought of it! Its only that his ev approach is exemplary, i.e., symptomatic. This demon is really a gremlin; wrecking a lot of hard work by his devious insinuations.)
Why is it that the demon cannot perform the task assigned to him? Usually, Leo Szilard is credited first with having returned the demon to the abyss of bad ideas, but I think Szilard has subtly begged the question. Szilard insightfully relates the problem to the demon’s “state of mind.” (Szilard replaces the demon with an idealized Carnot engine, but the translation is straightforward.) In effect, his diabolical brain was in a state of thermal equilibrium with the system he was supposed to perform work on. According to the most elementary formulation of the Second Law, no work is possible under these conditions. (One might say that the demon lacked sufficient specified information about the system to do his job.)
I believe that Szilard has correctly characterized the problem as one of the demon’s “knowledge” of the system; but then Szilard begs the question by making the demon so dumb that its still impossible for him to perform his task. So the question is, how much and what kind of (prior?) information must the demon be possessed of to do the trick?
I’ll leave it as an open question. Provide your own scenarios if you will. I considered it a problem of the minimum information sufficient to reduce the entropy of a mole of idealized molecular gas. I wasn’t very imaginative—didn’t think of any ways to make the demon cleverer than he needs to be to achieve the reduction, or change his diminutive stature, or imbue him with any special powers, or “exotic” implementations, etc. I realize that the solution of the problem, as I’ve stated it, is simple and straightforward, although I can't recall ever seeing an actual solution, but maybe the exercise will encourage discussion about some of the hidden implications (which I realize I haven’t “hidden” at all).
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 27. March 2002 15:54
I’m informed that Dr. Dembski has also performed the rites over the demon. Sorry, Doc, I haven’t read your book yet. (How embarrassing!) If I’m simply covering ground already well covered then I’ll leave it to the discretion of Mr. Moderator to dispense with the topic as he sees fit.
(I see there's no shortage of topics.)
By the way, congrats to the ISCID! Looks like quite a success.
IP: Logged
|
|
kyle7
Member
Member # 191
|
posted 30. March 2002 16:53
Janitor at MIT,
I think your post is excellent! Maybe we can expand on the idea beyond what Dembski presented. Here are some of my generalized thoughts on the subject:
- We must keep in mind the distinction between informational and thermodynamic entropy.
- We must show how the thermodynamic mechanism of the cell does not constitute a valid evolutionary "mechanism". In other words, we must show there exists a distinction between the biological mechanism that constructs the life form in regular procreation, and the evolutionary "mechanism" that is suppose to construct new complex biological systems for the life form.
- We must think of ways to test our informational entropy constructs and probability constructs to verify that we are not making a "Mind Projection Fallacy". Please note the thread "Signal agreement, and the role of information in the organization of complex systems" by Micah Sparacio. There is a link to "Clearing Up Mysteries" on the post written by Mika Vallittu on the 27th of February, 12:36. The paper presents the concept of the "Mind Projection Fallacy". I make a rebuttal to the "Mind Projection Fallacy" argument at the following link (Same thread, March 30 07:09)
kyle7 Rebuttal of the "Mind Projection Fallacy"
by using the "Ignorance Projection Fallacy" argument -- a term that I "coined" (at least I have not heard anyone else use it).
These are some thoughts that come to mind. I also have not read Dembski's other book that talks about Maxwell's Demon. Which one is it? I have his more theological book, which I was reading on a flight I recently went on. The day was clear and I had a laboratory just looking out the airplane window. "What is designed and what is natural?" I thought. I strongly suggest you try using this laboratory on your next flight!
[ 30 March 2002, 17:26: Message edited by: kyle7 ]
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 30. March 2002 18:53
quote:
Janitor@MIT: So, Schneider is well aware that “Maxwell’s demon is dead.” But this demon has more lives than Schrodinger’s cat! Having performed the rites over the demon, Schneider resurrects him (in the form of the ev program) to perform the necessary entropy reducing, “subtle sorting operation,” he indicates the demon cannot do!
(i) Schneider did not study thermodynamic entropy with his ev program.
(ii) Even if he would have done that, a biological population is not an isolated system so there's no reason to accuse him of relying on Maxwell's demon.
quote:
Janitor@MIT: Without going into any detail, the usual procedure is faithfully reproduced in ev: a target is specified and the information discarded or lost in attaining the target is dropped from the equations. I.e., information is definitively a reciprocal. Information is “added” in these exercises by dropping the denominator. If some other form of “information” is intended, then it behooves the experimenter to make this explicit in his formulations. Otherwise, the whole exercise is self-falsifying by being self-fulfilling, so to speak.
In his paper, Schneider was very clear on what he meant by "information" and how it is measured. He provided an explicit formula and even a computer program to compute it, so your comment that "If some other form of 'information' is intended, then it behooves the experimenter to make this explicit in his formulations." makes little sense. He could hardly have been more explicit. You, on the other hand, are very unclear about what you mean. In what sense was a target specified, how was some information "dropped from the equations", etc.? Please be specific.
Erik
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 31. March 2002 18:48
I thought this topic might have died a merciful death.
Erik: “(i) Schneider did not study thermodynamic entropy with his ev program.”
“This solves the problem of relating information theory to thermodynamics, which many writers have complained is not possible.”—Tom Schneider, see the presentation cited above for details. Your argument is with Schneider and not myself, since my attitude to the whole subject of the formal relation or identity of thermodynamic and informational entropies is indifference.
I believe this also addresses a question you had kyle7 and I’d appreciate your feedback. I also thoroughly enjoyed the Jaynes article, especially his hilarious comments on the state of quantum theory. The MPF has an interesting twist: In the “pure” sciences it is an “occupational hazard” to be avoided. But the object of engineering science is to project the mind onto matter. Think about it.
Erik: “(ii) Even if he would have done that, a biological population is not an isolated system so there's no reason to accuse him of relying on Maxwell's demon.”
It’s not apparent that Schneider considered relevant that Maxwell’s gedankenexperiment involves an isolated system. And I, like Schneider, don’t consider the demon biological. The demon’s purely metaphysical. However, Schneider treats the demon as if he were biological, Erik, to show that he can’t violate the Second Law. Again, your argument is with Schneider. Also, I didn’t “accuse” Schneider of anything other than making a mistake. He’s simply made a mistake. I mistakes all the time, so I’m not unsympathetic.
Erik: “In his paper, Schneider was very clear on what he meant by "information" and how it is measured. He provided an explicit formula and even a computer program to compute it, so your comment that "If some other form of 'information' is intended, then it behooves the experimenter to make this explicit in his formulations." makes little sense. He could hardly have been more explicit. You, on the other hand, are very unclear about what you mean. In what sense was a target specified, how was some information "dropped from the equations", etc.? Please be specific.”
Obviously, it’s not as clear to me. E.g., I thought Rseq was the target. But Rseq is not the target. Rather Rseq and Rfreq are related as a “distance,” as it is used in information theory. (Information measures are always distances.) The target is actually this distance. Schneider indicates that this distance measure is determinative. The problem is that setting either Rseq or Rfreq to zero is the equivalent of dividing by zero. What’s the distance between nowhere and the other side of nowhere? And how do I get from nowhere to nowhere? (Use a computer program, silly!)
Which raises additional questions: How do genes and their regulatory elements evolve independently? The program doesn’t even address this question. It’s petitio principii in silico. That’s the problem with computers. They only do what they are told to do. Which is what prompted my remark about self-fulfilling prophecies being scientifically self-falsifying. A computer is a self-fulfilling prophecy machine! Like any tool of science, we have to be careful how we use it.
I should point out that Schneider makes an important contribution to our understanding of bioinformation and how it works by indicating that the “meaningful” information is related as a “distance.” This is similar to my own thinking. Schneider shows that the relevant information is contained in an operation (finding a binding site) performed over the statistical information. I have nothing but words of encouragement for him to continue and extend this line of thought. I think it’s a very promising direction.
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 02. April 2002 14:03
quote:
Erik: (i) Schneider did not study thermodynamic entropy with his ev program.
Janitor@MIT: “This solves the problem of relating information theory to thermodynamics, which many writers have complained is not possible.”—Tom
Schneider, see the presentation cited above for details. Your argument is with Schneider and not myself, since my attitude to the whole subject of the formal relation or identity of thermodynamic and informational entropies is indifference. I believe this also addresses a question you had kyle7 and I’d appreciate your feedback. I also thoroughly enjoyed the Jaynes article, especially his hilarious comments on the state of quantum theory. The MPF has an interesting twist: In the “pure” sciences it is an “occupational hazard” to be avoided. But the object of engineering science is to project the mind onto matter. Think about it.
How can Schneider have invoked Maxwell's demon with his ev program, if he wasn't studying and modelling thermodynamic entropy? All I objected to was your implication that Schneider invoked Maxwell's demon with his ev program.
quote:
Erik: (ii) Even if he would have done that, a biological population is not an isolated system so there's no reason to accuse him of relying on Maxwell's demon.
Janitor@MIT: It’s not apparent that Schneider considered relevant that Maxwell’s gedankenexperiment involves an isolated system. And I, like Schneider, don’t consider the demon biological. The demon’s purely metaphysical. However, Schneider treats the demon as if he were biological, Erik, to show that he can’t violate the Second Law. Again, your argument is with Schneider. Also, I didn’t “accuse” Schneider of anything other than making a mistake. He’s simply made a mistake. I mistakes all the time, so I’m not unsympathetic.
Maxwell's demon is an imagined being located inside an isolated system. The remarkable thing about this being is that it appears to be able decrease the total entropy of the system. This would not be remarkable if the system weren't isolated, so I'd say that isolation is pretty central.
Schneider's nucleotide-frequency-entropy is neither an estimate of the population's thermodynamic entropy, nor is an evolving system isolated in a thermodynamic sense (or even in a nucleotide-frequency sense). Thus, it isn't really appropriate to bring up Maxwell's demon as anything but an attempt to enhance the presentation pedagogically. It does not work to try to argue against Schneider's conclusions from the ev program by saying that he relies on Maxwell's demon, because he doesn't.
quote:
Erik: In his paper, Schneider was very clear on what he meant by "information" and how it is measured. He provided an explicit formula and even a computer program to compute it, so your comment that "If some other form of 'information' is intended, then it behooves the experimenter to make this explicit in his formulations." makes little sense. He could hardly have been more explicit. You, on the other hand, are very unclear about what you mean. In what sense was a target specified, how was some information "dropped from the equations", etc.? Please be specific.
Janitor@MIT: Obviously, it’s not as clear to me. E.g., I thought Rseq was the target. But Rseq is not the target. Rather Rseq and Rfreq are related as a “distance,” as it is used in information theory. (Information measures are always distances.) The target is actually this distance. Schneider indicates that this distance measure is determinative.
I believe you have confused the definition of Rseq and Rfreq and the fitness determinition*. Rseq may be thought of as kind of "distance" between two probability distributions describing nucleotide frequencies (although, mathematically, it is not a proper distance, since it is anti-symmetric among other things). Rfreq is, according to Schneider's understanding of the molecular recognition process, the minimal number of yes-no choices that need to be made to locate a binding site. I don't see how Rfreq can be thought of as a distance in any natural way.
One of the main points of Schneider's ev paper is that Rseq and Rfreq are not related out of any mathematical necessity. They could take on values independently of each other without any mathematical contradiction. Schneider therefore concludes that there must be a biological explanation and proposes that Rseq has evolved to match Rfreq in order to provide the necessary data to base the yes-no choices on (otherwise the recognition process must operate with incomplete data, making it less reliable).
Having introduced his model for binding site recognition, he decided to model evolution using the eminently reasonable rule that organisms making fewer mistakes in recognizing binding sites should tend to reproduce more often (this is the only reasonable thing if you want to model what goes on in nature). It is here that the "distance" between Rseq and Rfreq comes in. As long as Rseq is lower than Rfreq, some mistakes will invariably be made and there is therefore a selection pressure to increase Rseq up to Rfreq.
quote:
Janitor@MIT: The problem is that setting either Rseq or Rfreq to zero is the equivalent of dividing by zero. What’s the distance between nowhere and the other side of nowhere? And how do I get from nowhere to nowhere? (Use a computer program, silly!)
No, it is not. Rfreq is equal to zero when there are as many binding sites as there are sites and Rseq is equal to zero whenever the nucleotides are independent and identically distributed. Everything is well-defined in these cases and at no point is there any division by either Rfreq or Rseq. The only relation between Rseq and Rfreq in Schneider's simulation is the one that implicitly arises because organisms making fewer mistakes (relative to the other individuals in the population) will reproduce more often.
quote:
Janitor@MIT: Which raises additional questions: How do genes and their regulatory elements evolve independently? The program doesn’t even address this question. It’s petitio principii in silico.
The program together with his paper addresses nothing more or less than precisely the question that Schneider wanted to address. Namely, the evolution of binding site recognition.
Erik
* I am aware that Schneider doesn't think that he uses a fitness function in the simulation, but he is simply unaware of how general the concept of a fitness function is.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 02. April 2002 16:48
Erik: “How can Schneider have invoked Maxwell's demon with his ev program, if he wasn't studying and modelling thermodynamic entropy?”
I can’t make it any clearer than Schneider did in the paper cited, Erik.
Erik: “The remarkable thing about this being is that it appears to be able decrease the total entropy of the system. This would not be remarkable if the system weren't isolated…”
Technically, the system is not isolated, as I indicated above.
Erik: “I believe you have confused the definition of Rseq and Rfreq and the fitness determinition…”
Never entered my mind. Schneider identifies Rseq/Rfreq as an information theoretic “distance” and that this distance is the target of the program.
Erik: “No, it is not…”
Maybe dividing by zero and getting something more than zero is a new form of evolutionary math! (LOL) Rseq/0 = 0. 0/Rfreq = 0. I just can’t seem to get this to work! (Of course Rseq/0 = zero positive infinity!)
Erik: “The program together with his paper addresses nothing more or less than precisely the question that Schneider wanted to address. Namely, the evolution of binding site recognition.”
Sorry, I thought this was a problem in the independent evolution of genes and their regulatory elements.
(My apologies to Mr. Moderator, I didn’t mean to make this a tit-for-tat. I’ll let Erik have the last word on this tangent.)
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 03. April 2002 05:02
quote:
Erik: How can Schneider have invoked Maxwell's demon with his ev program, if he wasn't studying and modelling thermodynamic entropy?
Janitor@MIT: I can’t make it any clearer than Schneider did in the paper cited, Erik.
Unfortunately, you have only quoted a general comment on Maxwell's demon from p. 6 of the article:
Schneider T. (1994) "Sequence Logos, Machine/Channel Capacity, Maxwell's Demon, and Molecular Computers: a Review of the Theory of Molecular Machines", Nanotechnology, 5 : 1-18
This paper does not discuss the ev program and the idea that the ev program invokes Maxwell's demon is purely your own. It is therefore not enough for you to refer to Schneider's paper in Nanotechnology. You will need to supply your own arguments, because nothing in the paper supports your position.
quote:
Erik: I believe you have confused the definition of Rseq and Rfreq and the fitness determinition…
Janitor@MIT: Never entered my mind. Schneider identifies Rseq/Rfreq as an information theoretic “distance” and that this distance is the target of the program.
Can you provide a full reference? I cannot find this identification in either the ev paper or the Nanotechnology paper.
Schneider T. (2000) "Evolution of biological information", Nucleic Acids Research, 28 : 2794-2799
Since Rfreq is constant in the ev simulation, and approximately constant in nature, it is pretty natural to normalize Rseq by Rfreq. However, it is not necessary and I can't see that Schneider did so in his ev paper.
quote:
Erik: No, it is not…
Janitor@MIT: Maybe dividing by zero and getting something more than zero is a new form of evolutionary math! (LOL) Rseq/0 = 0. 0/Rfreq = 0. I just can’t seem to get this to work! (Of course Rseq/0 = zero positive infinity!)
You claimed that "setting either Rseq or Rfreq to zero is the equivalent of dividing by zero". While it is true that Rseq cannot be normalized by division by Rfreq if the latter is zero, this normalization is not necessary. Besides, Rfreq is equal to 4 bits/site during the entire ev simulation. Because the nucleotides are independent and identically distributed in the initial state, Rseq does start out equal to zero. However, the normalization works just fine in this case and even if it didn't it wouldn't be a problem to refrain from this normalization (it's merely a matter of presentation).
quote:
Erik: The program together with his paper addresses nothing more or less than precisely the question that Schneider wanted to address. Namely, the evolution of binding site recognition.
Janitor@MIT: Sorry, I thought this was a problem in the independent evolution of genes and their regulatory elements.
The program does simulate the parallel evolution of binding site recognition and binding site, both of which are coded by the evolving genome. To this extent, it does simulate the parallel (not independent) evolution of genes and their regulatory elements. To the extent that gene regulation involves other processes than just binding site recognition, it does not. I am not a molecular biologist and don't know how many additional processes that are involved.
Erik
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 03. April 2002 09:11
Janitor and Erik:
I'll allow you one more round of posts on Schneider's ev program. It seems that you are going around in circles on this one.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 03. April 2002 12:12
Schneider’s papers are available online:
Sequence Logos, Machine/Channel Capacity, Maxwell's Demon, and Molecular Computers: a Review of the Theory of Molecular Machines
Evolution of Biological Information
[This .tex article explains the thermodynamics of computation, which is how Szilard redefined the problem. (Most stuff on the Web about Maxwell’s demon is bunk!)]
PhilSci Archive: Maxwell's Demon and the Thermodynamics of Computation
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 04. April 2002 12:24
Maybe I could open the topic up, and forget the obfuscating thoughtexperiment. (There seems to be some confusion, and in any case little interest. But thank you Erik and kyle7 for your input.)
Is there any basis in information theory for an “information added” function? As I indicated above, it seems that information is added only by switching between information as formally defined and information as “colloquialism.” The “common wisdom” is that biological evolution adds information to the genome, somehow drives evolution overall to higher levels of “complexity,” etc. The idea that information is added in the course of evolution is intuitively powerful, but appears to be just as powerfully misleading. There appears to be no basis for this idea in information theory.
But that information is added in the course of evolution seems to be trivially demonstrable. (And many such “demonstrations” are proffered.) Unfortunately, that’s usually what we do end up doing--demonstrating something trivial. Usually we end up demonstrating, trivially, that we’ve failed to apprehend or consistently apply information theoretic principles. But maybe there is some very fundamental principle that I’m missing here…
Is there a solution to this problem?—or do we end up invoking some sort of metaphysical being, demon, designer, leprechaun (I want to be politically correct to my Gaelic heritage, you know)?
Any thoughts?
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 04. April 2002 17:59
Janitor:
The fact that information theory gives us no way to understand the difference between a bacterium and a human being is one more excellent reason to abandon the whole information-theoretic approach to complexity, it seems to me.
By the way, it was announced today on the news that the rice genome has been sequenced, and found to be considerably larger than the human genome. Another excellent reason to reject the genocentric view of the organism. The two questions are not unrelated, since the genocentric viewpoint is tacitly modeled on the computer program metaphor, which also underlies the information-theoretic or algorithmic complexity approach.
Human beings are evidently more complex than either bacteria or rice, under any reasonable definition of "complex." Therefore, we need to move beyond algorithmic oversimplifications by expanding our notion of physics to encompass nonintegrable, nonlinear, and ultimately (I believe) intrinsically teleological emergent dynamics.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 05. April 2002 12:59
My view is far less expansive and far less ambitious than yours, James Barham, I have to admit. I believe the code-program paradigm is salvable—if biologists ever can be convinced to take it seriously, explore its potential, and stop looking at the evolution of the genetic code-program “as a savage looks at a ship.”
I think I’ve also explored tentatively some aspects of a synthesis, but you probably wouldn’t find it very appealing, because I approach it from the perspective of engineering. How do engineers design complex, non-linear systems? How do they control emergent properties and the immanence of teleology? They certainly do so rather clumsily compared to how these feats are accomplished in a grain of rice!
Touchette, Hugo & Seth Lloyd. 2000. “Information theoretic limits of control,” Phys. Rev. Lett. Vol. 84, pp. 1156-1159, and IEEE Transactions on Information Theory, “Information-Theoretic Approach to the Study of Control Systems.” (Pre-print accessed online 4/2/01.)
“The idea of reducing the entropy of a [non-linear] system using information gathered from estimating its state is not novel by itself. It has, in fact, been treated abundantly in the physics literature in the context of thermodynamics, particularly in connection with the so-called Maxwell’s demon paradox… However, it is an unfortunate fact that familiarity with the Maxwell’s demon paradox is not widespread among engineers working in control theory [or biologists], and therefore discussions of this subject in physics had a very limited impact, if none, on the field of control.”
The authors compare and contrast linear and non-linear control systems modeled as communication systems. If you are interested in how complex non-linear systems “evolve” or “emerge” teleologically as systems of control (as in biology) these articles are certainly a contribution to our understanding.
I’ve always been impressed by how “loopy” (non-linear) biology is. I know how an engineer handles these loops analytically (via decomposition and simultaneous closure), but this represents a significant problem for biological systems that have no “end in mind” as they evolve. (Traditional biological theory is absolutely irreconcilable with this “sum-over-histories” approach.) For their purposes the authors effect closure with a little “Bayesian trickery.” I.e., they exploit prior information that is not given in the system they are considering to get it to do what they want it to do.
Now some people entertain the idea that this “prior information” existed at the origin of the universe. To me it’s curious then why life isn’t quite common. (No one would ever wonder about us being alone. (We’d be complaining about over-crowding!) Or why biologists can’t “create” life in a flash, but seem to be obsessing on an extremely narrow and highly improbable set of necessary and sufficient conditions.
My question (at the top) was prompted by a desire to demonstrate that a recoding over inputs is an “information added” function. This requires that I eliminate all the “prior information.” But that’s absurd! (And the reason for my criticism of Tom Schneider, because he states that he does exactly that.) I suspect that what is really required is an extension of information theory (which you may have noticed has been one of my consistent themes); such as a synthesis with the mathematical theory of algorithms. (In computer science the application of an algorithm to information is logically irreversible, which results in algorithms having some “bizarre” information theoretic properties.)
Such a synthesis, even if it is the key, is far beyond a mathematical hack like myself. Nor is it obvious that measuring the information content of an algorithm (if it can be done) is a solution to the problem!—we simply take one more step in the infinite regress. The shade of Hume looms over all! The SOB!
The problem of “prior information” is the skeleton in biology’s closet. The interesting thing is that nearly everyone who treats biological evolution consistently in information theoretic terms discovers this skeleton. I think Dr. Dembski has some ideas whose skeleton that is… But he might think that’s not the proper subject matter here. Still… I have to wonder. (LOL)
But now I’m rambling. I better stop before Mr. Moderator tells me to start making sense.
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 06. April 2002 08:11
Janitor:
Thanks for the recent Seth Lloyd citation, which I had missed. I found his 1988 paper (with the late Heinz Pagels) on "thermodynamic depth" to be one of the most interesting discussions of complexity, precisely because it gets away from a purely formalistic, information-theoretic approach, and asks some of the essential physical questions.
I still don't think that the notion of thermodynamic depth is adequate for understanding biological complexity, however, which seems to me to depend on an understanding goal-directed action (or functionality). If you are interested, I compare and contrast both Bennett's "logical depth" and Pagels & Lloyd's "thermodynamic depth" with my own view of "epistemic depth" based on my dynamical model of the meaning of information in a little paper called "On the Objectivity of the Scala Naturae," Evolution and Cognition, 1999, 5: 2--11.
The basic idea is simply that our intuitive idea of biological complexity is based on the perception of functional capabilities, which can be defined as ways of interacting with the world. And using this idea as a metric, a perfectly objective "arrow" of increasing complexity can be discerned over the course of evolution from prokaryotes to eukaryotes to metazoans to encephalized metazoans to us. I believe this is what most people really mean when they talk about biological complexity, and information theory sheds no light whatsoever on this phenomenon.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 15. April 2002 13:39
Apparently Richard Dawkins (http://www.skeptics.com.au/journal/dawkins1.htm) has answered my question, by an analogy of “information” which is definitively true! (LOL)
Of course, this is exactly the issue I noted above: When “theorists” such as Richard Dawkins’ claim that there is an “information added” function to evolution they switch between formal definitions of information (which Dawkins’ just can’t seem to get right in the first place) and “information” informally given, or in this case given “by definition.” Frankly, I prefer Shannon’s information because it is measurable, whereas Dawkins’ “information” is “definitively” unmeasurable.
I think my question/problem remains. But I wonder if some of the addle brained, bamboozling theorists here can begin to count the technical and conceptual errors Dawkins makes in this brief article?
We might consider a number that is tossed around a lot, “Can we measure the information capacity of that portion of the genome which is actually used? We can at least estimate it. In the case of the human genome it is about 2%…” There is no basis in information science for this measure. Where biologists got it, I have no idea. Since I consider biological theory “folklorish science” maybe this is just a bit of folklore? But I would be interested if anyone knows of the source of this little bit of folklore?
We might want to consider Shannon’s comments on redundancy, which are a lot clearer than Dawkins’. And we might also want to consider Charles Darwin’s comments on whether natural selection reduces a “prior uncertainty” or a “posterior uncertainty.”
My impression is that Dawkins’ works are a poor reflection of his primary sources. (E.g., his popularization of William Hamilton’s “selfish gene” concept… well, never mind, that’s pretty much a dead issue, except in “popular” or “folklorish” science.)
Any creationist propagandists out there are invited to participate. (LOL)
[Thank you, j1! Did I ever tell you you’re wonderful?]
IP: Logged
|
|
|
Printer-friendly view of this topic