|
Author
|
Topic: Information. ID versus non-ID
|
Frances
Member
Member # 169
|
posted 11. January 2003 14:54
I started a separate thread to keep the original thread about survival of the least appetizing on track per suggestion of Irving.
Irving suggested that Noel was looking for a real demonstration but I believe the issue was
Those who want Darwinism to be true must demonstrate that random incremental change and selection can increase information.
And ev and others have shown that in principle mutation and natural selection are sufficient to increase information in the genome. Experiments in real life are more complicated but SELEX experiments are the next step to show that in 'real life' mutations and selection can be shown to increase information in the genome.
Hi Noel,
you state
quote:
Those who want Darwinism to be true must demonstrate that random incremental change and selection can increase information. Should they be able to do this, which is doubtful, it still does not follow that everyone would opt for their explanation.
That is a very simple demonstration and various experiments and simulations have shown that this is indeed the case
ev: Evolution of Biological Information
quote:
The ev model quantitatively addresses the question of how life gains information, a valid issue recently raised by creationists [32] (Truman, R. (1999), http://www.trueorigin.org/dawkinfo.htm) but only qualitatively addressed by biologists [33]. The mathematical form of uncertainty and entropy implies that neither can be negative, but a decrease in uncertainty or entropy can correspond to information gain, as measured here by Rsequenceand Rfrequency. The ev model shows explicitly how this information gain comes about from mutation and selection, without any other external influence, thereby completely answering the creationists
Evolution of Biological Complexity
quote:
In order to make a case for or against a trend in the evolution of complexity in biological evolution, complexity needs to be both rigorously defined and measurable. A recent information-theoretic (but intuitively evident) definition identifies genomic complexity with the amount of information a sequence stores about its environment. We investigate the evolution of genomic complexity in populations of digital organisms and monitor in detail the evolutionary transitions that increase complexity. We show that because natural selection forces genomes to behave as a natural ``Maxwell Demon'', within a fixed environment genomic complexity is forced to increase.
In fact as I have argued elsewhere, the fourth law of thermodynamics as proposed by Dembski is imho nothing more than a reformulation of the second law of thermodynamics for a closed system. When realizing that in open systems, entropy can decrease/information can increase and that in evolution it is the environment which infuses information into the genome through selection, and we realize that information increase is not that hard to realize.
Source
Others have comment on this
Victor J. Stenger and Here
found via this link
Adrian L. Melott
Common objections are that ev somehow smuggled in information and although noone really seems to have shown that there is any smuggling done, increase of information of course requires 'smuggling in information' just as selection 'smuggles in' information from the environment into the genome.
I would be interested in any evidence that the information was pre-coded.
A simple model
I am looking into these references
Lipps, J. H., Collins. A. G. and M. A. Fedonkin. 1998. Evolution of biological complexity: Evidence from geology, paleontology, and molecular biology. In R. B. Hoover (ed.) Instruments, Methods, and Missions for Astrobiology, Proceedings of SPIE, Vol. 3441.
C. L. Nehaniv & J. L. Rhodes, ``On the Manner in which Biological Complexity May Grow'', Mathematical Computational Biology, Lectures on Mathematics in the Life Sciences, Vol. 26, American Mathematical Society, pp. 93-102, 1999.
C. L. Nehaniv & J. L. Rhodes, The Evolution and Understanding of Hierarchical Complexity in Biology from an Algebraic Perspective, submitted to Artificial Life (MIT Press).
quote:
We develop the rigorous notion of a model for understanding state transition systems by hierarchical coordinate systems. Using this we motivate an algebraic definition of the complexity of biological systems, comparing it to other candidates such as genome size and number of cell types. We show that our complexity measure is the unique maximal complexity measure satisfying a natural set of axioms. This reveals a strong relationship between hierarchical complexity in biological systems and the area of algebra known as global semigroup theory. We then study the rate at which hierarchical complexity can evolve in biological systems assuming evolution is "as slow as possible" from the perspective of computational power of organisms. Explicit bounds on the evolution of complexity are derived showing that, although the evolutionary changes in hierarchical complexity are bounded, in some circumstances complexity may more than double in certain "genius jumps" of evolution. In fact, examples show that our bounds are sharp. We sketch the structure where such complexity jumps are known to occur and note some similarities to previously identified mechanisms in biological evolutionary transitions. We also address the question of, How fast can complexity evolve over longer periods of time? Although complexity may more than double in a single generation, we prove that in a smooth sequence of t "inclusion" steps, complexity may grow at most from N to (N + 1)t + N, a linear function of number of generations t, while for sequences of "mapping" steps it increases by at most t. Thus, despite the fact that there are major transitions in which complexity jumps are possible, over longer periods of time, the growth of complexity may be broken into maximal intervals on which it is bounded above in the manner described.
[ 13. January 2003, 07:40: Message edited by: Moderator ]
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 11. January 2003 16:30
quote:
And ev and others have shown that in principle mutation and natural selection are sufficient to increase information in the genome.
Actually, I don't think ev shows that at all. I think it shows that information can be programmed initially into an intelligent selector.
IP: Logged
|
|
rossum
Member
Member # 463
|
posted 11. January 2003 19:21
I disagree with Roger R. Dembski says that Evolutionary Algorithms cannot generate information (CSI to be precise). However Dembski agrees that EAs can transfer CSI from their fitness function into their output; the information is not generated, merely transferred. This means that, accepting what Dembski says, it is perfectly possible for evolution to transfer CSI from its fitness function into a species' genome. Hence the information in the genome will increase. The information has been transferred from the fitness function into the genome. There is supposedly no new information, but nevertheless there is more information in the genome that there was before the transfer.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 12. January 2003 08:40
quote:
Dembski says that Evolutionary Algorithms cannot generate information (CSI to be precise). However Dembski agrees that EAs can transfer CSI from their fitness function into their output; the information is not generated, merely transferred.
Agreed. And consistent with the point I made.
quote:
This means that, accepting what Dembski says, it is perfectly possible for evolution to transfer CSI from its fitness function into a species' genome. Hence the information in the genome will increase. The information has been transferred from the fitness function into the genome. There is supposedly no new information, but nevertheless there is more information in the genome that there was before the transfer.
Ah, but that isn't accepting what Dembski says. Dembski claims there is no known materialist source of CSI. Hence, there is no CSI that evolution can "transfer" to the genome w/o either an intelligent designer (the role of Tom in ev), unless there are currently unknown physical laws of the universe which contain or create CSI.
from Dembski quote:
All such focusing and constraining imparts information. Provided that information is both complex and specified, I show you never get more specified complexity out of such evolutionary processes than was programmed into them through such constraining.
IOW, an evolutionary process generically can create no CSI. It can only transfer CSI. In the case of ev, Tom writes a program to produce a specific output. It uses CSI and teleology to gain the specific result. Is that what evolution is doing? That seems to be what the dispute is about. Is there a teleological fitness function and intelligent selector in Darwinian evolution?
[ 12. January 2003, 08:44: Message edited by: Roger R ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 12. January 2003 13:20
Roger
quote:
In the case of ev, Tom writes a program to produce a specific output. It uses CSI and teleology to gain the specific result. Is that what evolution is doing? That seems to be what the dispute is about. Is there a teleological fitness function and intelligent selector in Darwinian evolution?
Please show that teleology was used in Tom's program to 'gain a specific result'. Your use of the terms intelligent selector and teleological fitness function require you to show that there is a difference between the program simulation and how evolution works in nature. All that is done in the program is 1. introduction of variation 2. selection based on fitness
Btw while Dembski may argue that materialistic forces cannot add CSI he has not shown this to be the case. In fact the similarity between his 'conservation of CSI' law and the Second Law of thermodynamics shows that for open systems CSI can increase. No need for intelligence.
I encourage Roger to show how the programs are coded for a certain outcome and how the program differs from nature. It seems that Roger is using some of the same irrelevant strawmen raised by Dembski namely that experiments by necessity involve some form of intelligence.
The main page for Ev
Rebuttal to William A. Dembski's Posting and to His Book "No Free Lunch"
CSI and the fourth law of thermodynamics
First I would like to point out, as has Sobel, that specification seems to be trivial and applicable to almost any event which suggests that we may have to infer design for almost any event.
quote:
From this second illustration can be gathered that Dembski's theory enables a moderately imaginative person, with a list of possible delimitations of an event, easily eliminate relevant chance-hypotheses for the event; if they all make more probable that not its non-occurrence, and avoiding 'false negatives' concerning relevant chance-hypotheses for this event is somewhat (it need not be very) important to him. From the two illustrations, one may
gather that by the lights of Dembski’s book, we are entitled, and will always be entitled to conclude, that not much happens by chance.
So now we have to deal with complex information and its conservation. In fact as others have pointed out the term conservation is misleading in that the law does not conserve. It states that in a closes system complexity/entropy can only decrease/increase. Sounds familiar? Well how does one propose to resolve the increase in complexity? Dembski provides for one method, intelligence which injects CSI from the outside. But why should only intelligence be capable of increasing/decreasing information/entropy? In fact extending the fourth/second law to include an open system we find that in an open system complexity/entropy can increase/decrease.
Matt Young has shown How to Evolve Specified Complexity by Natural Means
Peter Godfrey Smith in Information and the Argument from Design addresses some of these issues as well. He argues that Dembski does not allow even the tiniest amount of information to be added by nature but I believe he is wrong, Dembski allows information to be added as long as it is less than the univeral bound.
[ 12. January 2003, 13:47: Message edited by: Frances ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 12. January 2003 15:40
Several of us made interesting points with regard to exactly this subject in reply to Dembski's Evolution's Logic of Credulity: An Unfettered Response to Allen Orr thread. So far I have seen no answers to those questions and comments, and they seem to be almost exactly the same issues being discussed here.
Especially relevant is the point "EAs can transfer CSI from their fitness function into their output", which means that in the end the source of information in evolution is the envoronment, according to the specific definitions and terminology being used. As I point out, this is exactly what biologists were saying in the theory of evolution.
[ 12. January 2003, 15:46: Message edited by: gedanken ]
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 12. January 2003 15:48
quote:
Originally posted by Frances:
Please show that teleology was used in Tom's program to 'gain a specific result'.
Let us let Tom put it in his words:
quote:
After I finished the work on my thesis I had one week before my thesis defense. I sat down and wrote the ev program in about 2 days (this is only possible if you have a clear picture of where you are going!) and [b]discovered to my delight that in only 500 generations I could get the full evolution[b], and Rsequence did indeed approach the predicted Rfrequence and oscillate around that level.
Sounds like he had a teleological goal in mind. And that goal drives the "mistake" counting in his program. Run the program over again, and it goes toward the same goal. Yes, it varies slightly in how quickly it gets there, due to the randomizing function, but it always goes toward that specific goal, and not other goals that other GA's go towards. Why would that be, if it wasn't goal driven?
quote:
Your use of the terms intelligent selector and teleological fitness function require you to show that there is a difference between the program simulation and how evolution works in nature.
Now that's just plain silly. I've never claimed there is no teleology in "nature", so the program may indeed simulate nature. Isn't part of the general disagreement here about the nature of nature? My only point is that the program models teleology, not what it says about the nature of actual biological evolution. I don't think it can demonstrate what capabilities non-teleological non-intelligent selectors such as NS can perform.
quote:
All that is done in the program is 1. introduction of variation 2. selection based on fitness
Fitness defined in terms of a future goal. Read what Tom says. That's an intelligent selector, not the BWM we call NS. Unless you want to ascribe teleology to NS.
quote:
Btw while Dembski may argue that materialistic forces cannot add CSI he has not shown this to be the case.
While I'll admit that the issue is still in dispute, the same applies to your claim in the OP. I don't think there is much doubt that rossum's claim about what Dembski says is in error.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 12. January 2003 16:27
There are at least two distinct mathematical/scientific concepts of information. One concept is ‘Information as a definable abstract mathematical concept which is useful in modeling and analyzing real world phenomena’. This might be characterized as the ‘information as a type of mathematics such as calculus perspective.
The second concept is ‘Information as a type of physical particle or energy.’ The information as a physical phenomena concept is at best extremely speculative. There is no direct evidence for any such particle or form of energy. The traits attributed to this phenomena by various experts, including Dembski, are pure unsupported creative speculation.
Apparently, because information as a physical phenomena is compatible with the belief or desire to believe that ‘life force’ can be reduced to a classical force in physics, the speculation on information as a form of energy has been addressed extensively in the literature. Because the subject is discussed in the literature, individuals like Frances begin to believe it is true.
If you view information as a definable mathematical concept, then the answer to the question "Can biological systems generate information?" depends on how information and the creation of information is defined. On a common sense basis, one might be inclined to define ‘creation of information’ in such a way that ‘the change from ape to man’ represents and increase in information’ or ‘creating the Mona Lisa’ or ‘creating a Shakespearean play’ involving an increase in information. It is obviously not logically necessary to define information in such a manner, but it would IMO be reasonable and useful to do so.
To attempt to answer the creation of information question by a review of the literature is to assume or pretend that someone has actually found some type of physical phenomena called information and been able to analyze it. Frances has identified many references discussing characteristics of information. The discussions are all based on the ‘assumption’ that information exists as a physical phenomena. I wonder what if any evidence Frances can produce that this phenomena actually exists.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 12. January 2003 16:41
Roger
I notice that you still have not identified how and where Tom Schneider used teleology to gain a specific result. In fact you seem to confuse the hypothesis formulated by Tom with the experiment to support or disprove the hypothesis. His hypothesis was that random mutations and selection could explain the observed Rfreq. But did he code this into his program? Of course not.
That no matter what initial condition this goal is always reached is an example of robustness but was this goal programmed into the program? Roger has failed so far to show that this is indeed the case. In fact if Roger had read the pages I
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 12. January 2003 16:42
Roger
I notice that you still have not identified how and where Tom Schneider used teleology to gain a specific result. In fact you seem to confuse the hypothesis formulated by Tom with the experiment to support or disprove the hypothesis. His hypothesis was that random mutations and selection could explain the observed Rfreq. But did he code this into his program? Of course not.
That no matter what initial condition this goal is always reached is an example of robustness but was this goal programmed into the program? Roger has failed so far to show that this is indeed the case. In fact if Roger had read the pages I referenced he would have found that different initial conditions and runs lead to different outcomes.
quote:
The ev program was run repeatedly to 2000 generations starting with 100 different random seeds. The lowest observed final information content was 2.3 bits and the highest was 5.2 bits, with a mean of 3.8 0.5 bits. Duplicate runs occured 7% of the time. These duplicates do not affect any conclusions, but they do suggest that the random number generator is not the best. Despite this, the program invariably gave a significant information increase. From the observed values, we can determine that the probability of a return to zero information is 1.5 x 10-14 (7.6 standard deviations).
Fitness was not defined as a future goal, if you had read Ev's manual and the accompagnying papers you would have realized that the fitness function is dependent only on time t. Your claim that teleology or an intelligent selector is being used fails to show that 1. teleology is actually been used, in fact the opposite seems to be the case 2. that intelligence is a requirement for the selector.
As far as Rossum's claims, he seems to be correct. Dembski does allow algorithms to generate information and Dembski does allow that EA's can 'transfer' CSI. But similarly to EA's being probability amplifiers, human intelligences functions as probability amplifiers as well. Dembski's claim that in a Closed system CSI can only decrease seems to be correct after all his 4th law seems to be nothing more than a reformulation of the 2nd law of thermodynamics. But once he allows the system to be open such as through an intelligent designer OR through a natural designer, complex specified information can increase in both cases.
What part of Rossum's response do you doubt?
So far you have failed to show where in the program Schneider used teleology. Your argument so far confuses hypothesis with the program, suggests that contrary to fact the fitness is with respect to future and seems unaware of the fact that the outcome of the experiment is NOT the same always.
I suggest that you read the relevant papers and pages that deal with Schneider's program before you accuse him of something you cannot even support.
A great Resource
Additionally I would like to point out that the use of NFL theorems by Dembski may suffer from some setback now that it has been shown that NFL theorems likely not apply.
See On Classes of Functions for which No Free Lunch Results Hold
Some good papers addressing and refuting Dembski's claims can be found Here click on the link Intelligent design
Another Reason is that the environment and the organism co-evolve.
And then there is the problem of false positives
quote:
Another book, in which we find a more detailed and systematic criticism of Dembski’s work was published [7] by the professor of philosophy Del Ratzsch . The entirety of Ratzsch’s writing makes it clear that he himself belongs to the camp of “design theorists.” However, unlike most of his co-travelers, Ratzsch is usually logical and meticulous in his discourse. In an appendix to the mentioned book, Ratzsch subjects some parts of Dembski’s work to a strong critique. Ratzsch’s critical remarks relate almost exclusively to Dembski’s “explanatory filter.” In particular, Ratzsch convincingly illustrates the fallacy of Dembski’s assertion that his “filter” does not produce “false positives,” which is in itself sufficient to render the entire concept of that “filter” largely useless.
Source
[ 13. January 2003, 07:41: Message edited by: Moderator ]
IP: Logged
|
|
Noel Rude
Member
Member # 516
|
posted 12. January 2003 17:13
I can imagine constructing a machine that "learns" from experience, that adds data to its data banks for later use -- say a robotic rover sent to Mars. Even apart from sending information home to minds that can understand it, there is the likelihood that the machine can be programmed to improve how it reacts to its environment, maybe to map out the best route back home from wherever it had been sent. But is this really specified complexity? Is it any different than a leaf falling to the ground and being imprinted by stains from the soil?
Does any of it really make any sense without a mind at one or the other end of the information continuum?
IP: Logged
|
|
rossum
Member
Member # 463
|
posted 12. January 2003 18:47
Roger R. said:
quote:
I don't think there is much doubt that rossum's claim about what Dembski says is in error.
My apologies for not giving the reference: In "Why Natural Selection Can't Design Anything" Dembski said:
quote:
To sum up, there is no getting around the displacement problem. Any output of specified complexity requires a prior input of specified complexity. In the case of evolutionary algorithms, they can yield specified complexity only if they themselves are carefully front-loaded with the right information (typically via a fitness function) and thus carefully adapted to the problem at hand. In other words, all the specified complexity we get out of an evolutionary algorithm has first to be put into its construction and into the information that guides the algorithm. Evolutionary algorithms therefore do not generate or create specified complexity but merely harness already existing specified complexity.
(Section 4 - The Displacement Problem)
The paper is available here on this site.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 12. January 2003 19:13
rossum's initial claim:
quote:
However Dembski agrees that EAs can transfer CSI from their fitness function into their output; the information is not generated, merely transferred. This means that, accepting what Dembski says, it is perfectly possible for evolution to transfer CSI from its fitness function into a species' genome. Hence the information in the genome will increase.
Now, from his citation of Dembski supposedly backing his claim:
quote:
In the case of evolutionary algorithms, they can yield specified complexity only if they themselves are carefully front-loaded with the right information (typically via a fitness function) and thus carefully adapted to the problem at hand. In other words, all the specified complexity we get out of an evolutionary algorithm has first to be put into its construction and into the information that guides the algorithm.
IOW, only if we have a designer inputting CSI. CSI doesn't come from materialist processes, according to Dembski. He may be wrong in that opinion, but that clearly is his opinion, and clearly contradicts rossum's claim that what he proposes is consistent with what Dembski says.
It isn't.
[ 12. January 2003, 19:16: Message edited by: Roger R ]
IP: Logged
|
|
rossum
Member
Member # 463
|
posted 13. January 2003 17:47
I think that I can now agree with Roger R. Evolution can indeed increase the information in the genome. The question is where the extra information in the genome comes from. There seem to be at least three options:
- Evolution: the extra information in the genome is generated directly by evolution as part of a natural materialistic process of interaction with the environment.
- Front Loaded ID: the extra information in the genome was originally present in the fitness function, placed there by Intelligent Designer/s, and has merely been transferred into the genome by evolution.
- Interventionist ID: the extra information in the genome was placed there directly by the Intelligent Designer/s
I suspect that Roger and myself would disagree as to which of these applies. Unfortunately there is currently a lack of evidence as to the generation of information in the genome. Plenty of rhetoric, but not a lot of evidence. Information is present in the genome, and it seems to match information in the environment. Beyond that there is not a lot.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 14. January 2003 16:18
quote:
Originally posted by Frances:
I notice that you still have not identified how and where Tom Schneider used teleology to gain a specific result.
Not true. I told you the main place was in the mistake counting routine. All the logic for counting mistakes is programmed in advance. It is telelological, and meant to produce specific results. Such does not mean the program has no value, only that it doesn't show what non-teleological non-intelligent selection on random variation can do.
There are other fine tunings, such as not allowing extinction, and the "special rule", which was the subject of some controversy between Schneider and Dembski. But the mistake counting routine is the heart of the teleology.
quote:
In fact you seem to confuse the hypothesis formulated by Tom with the experiment to support or disprove the hypothesis. His hypothesis was that random mutations and selection could explain the observed Rfreq. But did he code this into his program? Of course not.
So, did the mistake counting routine get inserted by special creation, or did it evolve? But random mutations and selection isn't the issue. Dembski clearly knows that intelligent selection can produce an information rich genome. It is Natural Selection that is the subject of dispute.
quote:
That no matter what initial condition this goal is always reached is an example of robustness but was this goal programmed into the program? Roger has failed so far to show that this is indeed the case. In fact if Roger had read the pages I referenced he would have found that different initial conditions and runs lead to different outcomes.
And is the fact that Dawkins program always gets to METHINKS IT IS LIKE A WEASEL and not NOW IS THE TIME FOR ALL GOOD MEN an example of the robustness of some unknown physical law favoring the former? No, I would say it is teleology implemented. And if you had read my prior post, you would know I referenced the issue of the varying speed with which each run reached its goal, due to the randomizing function. The only difference in your citation is that they fixed the number of generations, and see how far towards the goal the run got. Now, you may claim that is a "different outcome", I don't. If we ran the Dawkins program numerous times for 25 generations each, we would also get "different outcomes". Does that mean there was no teleology involved? Hardly.
I know you and I won't agree on this point. But that is a central part of the dispute, and not something that both sides can easily stipulate to, as your OP implies.
IP: Logged
|
|
|
Printer-friendly view of this topic