|
Author
|
Topic: Information creation and transcendence
|
Frances
Member
Member # 169
|
posted 25. January 2003 13:30
Warren,
Your use and abuse of thinly veiled ad hominems such as
quote:
In the business world it is considered misleading to use assumptions which are known to be incorrect and which are known to distort analytical results. I have always found it curious that that standard does not appear to be applied in evolutionary biology.
suggests to me that you have no real arguments. May I also point out that such comments seem to collide with the goals and guidelines of this forum.
In his reply Warren still shows that he does not seem to realize that evolutionary biology deals with populations not individuals. Thus his claims about "distoring the results" and "only work when unrealistic and distorting assumptions are used" remain once again without much supporting evidence.
And lost for arguments Warren continues his series of ad hominems
quote:
Gedanken’s argument that biologists know that the environment is not stable supports my assertion. Biologists ‘know’ the assumption is unrealistic and they know that RM&NS processes can not find adaptive solutions is the environment changes more than once per lifetime. Despite the fact that it is known that realistic rates of environmental change are incompatible with Darwinian evolution, biologists continue to rely on demonstrations based on environmental assumptions that are known to be both unrealistic and distorting.
I suggest that Warren refrains from such silly accusations which he cannot even support and avoid creating an atmosphere which is not constructive nor scientific.
But then again, I am not a moderator.
[ 25. January 2003, 14:41: Message edited by: Frances ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 25. January 2003 14:19
Gedanken and RBH,
So as to avoid getting into forbidden territory I never used the term fraud. Rex presented demonstrations intended to show that evolution can be explained, modeled and simulated by what I call low information models. The demonstration presented is no original to Rex and represents a variations of arguments commonly presented in support of Darwinian low information type models. As I correctly pointed out, 1) the models or simulations do not work unless you make the listed assumptions, 2)biologists known the assumptions used are unrealistic and 3)biologist know the assumptions used distort the result presented.
There are legitimate uses in mathematical modeling for unrealistic simplifying assumptions. The use of simplifying assumptions is for example generally considered acceptable when it can be demonstrated that the assumption do not distort the conclusions. I do not know the rationalization for the practices followed in evolutionary biology.
Gedanken’s argument that biologists know that the environment is not stable supports my assertion. Biologists ‘know’ the assumption is unrealistic and they know that RM&NS processes can not find adaptive solutions is the environment changes more than once per lifetime. Despite the fact that it is known that realistic rates of environmental change are incompatible with Darwinian evolution, biologists continue to rely on demonstrations based on environmental assumptions that are known to be both unrealistic and distorting.
Implicit in the demonstrations presented by Rex is the assumption that change is steady and gradual. Related to this is the implicit assumption that there is no need to recognize time limits on change. Again this assumption is both known to be false and known to distort results.
Quote gedanken: Warren, the lack of a complete mapping does not negate that the physical characteristics of an organism are almost completely determined by genetic information.
As I have discussed elsewhere, the misconception that ‘genes determine or cause physical traits’ appears to be based on confusing a ‘descriptive relationship’ with a ‘causal relationship’. In most common instances, the same genes result in the same types of physical traits. It is easily demonstrated both that the observed descriptive relationships do not qualify as the type of mappings required by a low information theory. It would be more accurate to say that there is a complete lack of maps rather than an incomplete mapping.
Quote gedanken: The real world has so much complexity that no generalization is going to be a complete 100% fit. But the biological evolutionary models fit a vast grid of evidence within very reasonable experimental tolerances.
Biological evolutionary models only fit the data when unrealistic and distorting assumptions are used. The issue of fit or ‘scientifically acceptable fit’ is a complex technical issue which I will be glad to discuss under another heading.
RBH seems to be confused about the subject being discussed. The issue being discussed is the impact of unrealistic assumptions of demonstrations supporting Darwinian type theories.
Quote: Another place warren goes astray is his apparent assumption that organisms process every bit of information flowing in their environments.
I have no idea where you got this erroneous impression or what possible relevance it has to anything being discussed. If you think it is relevant maybe you could explain in a bit more detail.
To return to the subject of this thread, the issue here is the definition of information and its impact on scientific analysis of life forms. The definitions I proposed suggest the mathematical possibility of modeling biological information processing, including evolution with either ‘low information models or high information models’. I suggested that Darwinian low information models are inadequate. Rex presented demonstrations purporting to show the adequacy of low information models and I demonstrated the models presented were based on assumptions which are both unrealistic and which distort results.
[ 25. January 2003, 14:40: Message edited by: warren_bergerson ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 25. January 2003 15:52
warren wrote quote:
Gedanken's argument that biologists know that the environment is not stable supports my assertion. Biologists 'know' the assumption is unrealistic and they know that RM&NS processes can not find adaptive solutions is the environment changes more than once per lifetime. Despite the fact that it is known that realistic rates of environmental change are incompatible with Darwinian evolution, biologists continue to rely on demonstrations based on environmental assumptions that are known to be both unrealistic and distorting.
and quote:
RBH seems to be confused about the subject being discussed. The issue being discussed is the impact of unrealistic assumptions of demonstrations supporting Darwinian type theories.
Once again, I allowed myself to be drawn into the quagmire. When will I learn? I will say only that if I am confused, I am not the one who is confused. I will say this one more time then retire to the sidelines:
Evolutionary biologists and researchers in allied disciplines know that environments change on all time grains from moments to millenia, and they actually study (i.e., they do actual research, as distinguished from armchair speculating based on subjective impressions gleaned from pop literature and casual observation) the determinants of adaptation at all those time grains!
I can't say it any plainer than that.
warren claims that quote:
Biologists 'know' the assumption is unrealistic and they know that RM&NS processes can not find adaptive solutions is (sic) the environment changes more than once per lifetime.
True. But they also know that regular biological evolutionary processes operating over generations can produce short-term mechanisms (like, say, "learning") by means of which organisms do find adaptive solutions during their individual lifetimes. It is a mystery to me why warren would think that short-term adaptations must be directly and immediately produced by biological evolution by means of random genetic mutations, recombination, and natural selection occurring during an individual organism's lifetime in order to enable that particular organism to adapt to changing environmental circumstances within its lifetime. As far as I know, he is unique in the world in claiming that's what evolutionary biology should account for.
"Modeling biological information processing," as warren asserts he wants to do, requires actually knowing something about the processing that occurs in biological systems and about the processing mechanisms in real biological systems that are to be modeled. I see no evidence that's the case in warren's posts.
RBH
Added in edit: Perhaps the problem is that the discussion is at a level of abstraction well removed from actual biological phenomena, and thus it is difficult to connect the abstractions with biological reality. Would warren be good enough to give us several examples of real adaptations in real biological organisms that exemplify the kinds of things evolutionary theory cannot account for and that therefore require some other kind of account?
RBH
[ 25. January 2003, 16:58: Message edited by: RBH ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 26. January 2003 00:19
Let's get the irrelevant yet distracting stuff out of the way first.
quote:
In the business world it is considered misleading to use assumptions which are known to be incorrect and which are known to distort analytical results. I have always found it curious that that standard does not appear to be applied in evolutionary biology.
In the debate world is is considered improper to use attacks on the character of the opposing side instead of addressing their arguments. So let's not, hm?
Okay, on to the interesting parts.
quote:
First, you assume the environmental conditions or the fitness evaluation function as you call it, is stable over the life of the organisms.
No, I don't. In fact, I said:
quote:
In terms of evolution, you can think of life as a fitness evaluation function, f(G,P) where G is the genome of the organism and P is the physiology of its parent. (The dependence on the environment is implicit.)
(Emphasis added.)
I say nothing about stability over time, and explicitly acknowledge the role of the environment. If you want me to make it explicit, we can use f(G,P,E), where E is the environmental conditions over the life of the organism.
The mutations I talked about have the convenient aspect of having f(G,P,E)=0 for essentially all relevant E, so my point about the relative information content in genotype vs. physiology stands.
quote:
Second your analysis is based on the assumption that both genetic and evolutionary change are gradual and steady.
You mean the analysis where I used an average rate of information gain of one bit per 55 generations? You used that rate to claim that there was insufficient time to generate enough information. I won't vouch for the number, but as I showed, it is sufficient to account for the information in the genome.
If you want to make different assumptions in your criticism, I will address those. I did not in principle see a problem with using average information gain rates and total information in the genome, especially since you were using average information gain rates and total information processed by the organism.
quote:
Finally, your analysis assumes that there is an identifiable functional relationship or mapping from genotype to phenotype and from phenotype to genotype.
Why do I assume that it is identifiable? I only assume that it exists.
The last 140 years of genetic research have generated a fair few examples of genotypes having a major impact on phenotype.
I hope you are not contesting that fact.
quote:
If development increases the volume of information, then the existence of a genotype to phenotype functional mapping is a mathematical impossibility.
Fair enough; the classical description is usually rather sloppy mathematically. This doesn't make the conclusions invalid (it's a decent approximation for the most part), but if you want something more precise:
Genotype and phenotype are related by a development function d that maps genotype, parental phenotype, and environment to a probability distribution on the set of all possible phenotypes:
d : (g,P,E) -> (k:{p}->x)
It turns out that our observations are that for a wide range of E, and for most phenotypes p, and for the common set of parental phenotypes P, that k:{p}->x is well approximated by a delta function about a central exemplar phenotype p0 (at least in terms of features we usually care about). Thus, approximately,
d : g -> p0
Is this better?
In any case, there is no problem with development increasing the volume of information because--as you implied in your first point--the result of development (and the fitness of the organism) is depentent on the parent and on the (rest of the) environment.
Can you show that this mapping is a mathematical impossibility?
quote:
In terms of information processing, this cell contains 1)stored genetic material, 2)a physical processing environment(hardware) and 3)a processing logic or program (software). Standard Darwinian genetic theory suggests that all the information needed to operate a complex organism is contained in the genetic material.
I won't claim to speak for "standard Darwinian genetic theory", but developmental biologists, especially developmental neurobiologists, recognize that environment plays a major role in the development of an organism; given abnormal visual input, for example, genetically normal animals can fail to develop functional sight.
These developmental biologists are hard at work identifying genes that are responsible for the genes that are involved in this environmental coupling. A few have been found--for example the daf (dauer formation) pathway in C. elegans is moderately well understood at a genetic level.
Anyway, the bottom line is that there doesn't seem to be any incompatability with having the genetic material specify (loosely--see probability distribution mapping above) the ways in which development will be impacted by the environment.
quote:
Based on our existing knowledge of information processing in computers, it is easily demonstrated that the functionality observed in a complex organism such as a human or a mouse could not be defined by 30,000 relatively simple pieces of information.
Could you sketch how this demonstration would work? It's not your 10^10000*10^8*10^13 (or whatever) calculation above, is it?
If it is, I'll try again to explain why that calculation is irrelevant. If not, it would help if you could sketch the demonstration using mice instead of humans, as it may be instructive to look up relevant experiments, and experiments on mice tend to be more ethical than those on humans. (Plus, mice develop faster, so we get our answer sooner.)
quote:
As I have discussed elsewhere, the misconception that 'genes determine or cause physical traits' appears to be based on confusing a 'descriptive relationship' with a 'causal relationship'.
If I knock out the white gene in flies, and the corresponding flies have white eyes, in what sense is the relationship between the gene and the eye color non-causal? (Note: it is possible to re-express the gene in only one eye and get flies with one white eye and one (normal) red eye. Is that causal enough yet?)
quote:
It is easily demonstrated both that the observed descriptive relationships do not qualify as the type of mappings required by a low information theory.
Maybe it would help if you gave the demonstration. I haven't any idea what it would look like.
[ 26. January 2003, 02:25: Message edited by: Rex Kerr ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 26. January 2003 14:26
Rex,
It may be useful to go back to your post of the 24th. You comments referred to a set of studies or observations on when in the life cycle non-beneficial mutations produced their impact. You suggested the results of these studies are more compatible with a ‘low volume Darwinian type theory or model’ than with the interpretation you offered for my ‘high volume theories or model’. I suggested your arguments were flawed because they were based on assumptions that were known to be false and to produce misleading results. I could also have argued that your presentation is based on inconsistent use of assumptions.
As RBH concedes, low volumes processes can not find an adaptive genetic solution unless it is assumed the environment as it impacts survival is stable. You now are claiming that the analysis being offered is based on an environmental assumption which directly contradicts the assumption necessary to produce genetic change. Needless to say, inconsistency in assumptions is also misleading.
The slow and steady evolutionary change assumption is also implicit in all low information theories, because there is no process or mechanisms to slow or speed up the process other than environmental changes and we have already conceded that a low volume theory or model requires that such change must be infrequent.
As you concede, your argument is based on the assumption that there exists a genotype to phenotype map. The existence of a functional mappings from genotype to phenotype and from phenotype to genotype is a requirement of a low volume information type theory. Your argument, I assume, is that there is a causal genotype to phenotype map as opposed to a descriptive or incidental relationship. The genotype to phenotype causal relationship can not be a simple ‘immediate’ relationship because the same set of genes can produce cells with a very wide range of different characteristics(phenotypes). Also, the phenotypes associated with individual cells change over time-sometimes, as with neurons, the phenotype changes rapidly and often.
There are several types of functional genotype to phenotype causal relationships which could potentially account for the transformations. Transformations involving clocks and/or chemical triggers for example. The validity of such causal maps is, however, easily tested and disproved. The assumption of a functional genotype to phenotype mapping is thus the assumption of a phenomena which it is known does not and can not exist.
Your demonstration suggests that a low information type theory would predict a certain pattern of lethal mutations. But as shown, you can not even get to the point of making a prediction without using either assumptions that are known to be false and misleading or without using inconsistent assumptions in different parts of your analysis
Quote: In the debate world is is considered improper to use attacks on the character of the opposing side instead of addressing their arguments. So let's not, hm?
I have not commented on your character, which I agree would be inappropriate, but on arguments you are presenting. You presented arguments suggesting ‘Darwinian low volume information theories’ would make certain predictions. I am simply demonstrating that those arguments are based on inappropriate use of assumptions.
You made a number of additional points which I will try to address latter.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 26. January 2003 22:16
quote:
Gedanken’s argument that biologists know that the environment is not stable supports my assertion. Biologists ‘know’ the assumption is unrealistic and they know that RM&NS processes can not find adaptive solutions is the environment changes more than once per lifetime. Despite the fact that it is known that realistic rates of environmental change are incompatible with Darwinian evolution, biologists continue to rely on demonstrations based on environmental assumptions that are known to be both unrealistic and distorting.
Implicit in the demonstrations presented by Rex is the assumption that change is steady and gradual. Related to this is the implicit assumption that there is no need to recognize time limits on change. Again this assumption is both known to be false and known to distort results.
quote:
As RBH concedes, low volumes processes can not find an adaptive genetic solution unless it is assumed the environment as it impacts survival is stable. You now are claiming that the analysis being offered is based on an environmental assumption which directly contradicts the assumption necessary to produce genetic change. Needless to say, inconsistency in assumptions is also misleading.
And not only these, but there are many other examples of scientists and statisticians misleading the public:
Population trends: Population trends are meaningless because individuals are born in erratic intervals. In fact a given mother may not have a baby for months or even years, then all the sudden she has a baby. This is entirely variable, with sudden punctuated increases in the population in a family. The Census Bureau is misleading the public.
Baseball averages: No hitter could be judged by his hitting average -- in some games he may hit a lot, in others only a little. This is varying all over the map from game to game. In fact, there are large periods of the game in which the hitter is not even at bat (environmental variation moment to moment). And more than that, there are much longer periods (in relative terms) between games in which the hitter does not even get an opportunity to hit. So hitting averages are meaningless. Scorekeepers are misleading the public with their statistics.
Weather forecasting: Individual rain drops hit the ground with entire fractions of a second between the drops. There is great irregularity on any given square inch of ground. On a second by second basis a square inch of ground may be struck by one rain drop, or several rain drops or no rain drop at all. Forecasters are misleading the public to give rain measurements. And temperature and other aspects are all varying second by second, minute by minute, hour by hour. How can trends be forecast for a day or two in advance? Again weather forecasters are misleading the public.
In the business world it is considered misleading to use assumptions which are known to be incorrect and which are known to distort analytical results. I have always found it curious that that standard does not appear to be applied to the Census Bureau, to baseball score keepers, and to weather forecasters.
[ 26. January 2003, 22:34: Message edited by: gedanken ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 26. January 2003 22:17
quote:
As RBH concedes, low volumes processes can not find an adaptive genetic solution unless it is assumed the environment as it impacts survival is stable. You now are claiming that the analysis being offered is based on an environmental assumption which directly contradicts the assumption necessary to produce genetic change.
I think you may be confusing an evolutionary story with my test for information content. I was only addressing the question of whether the "starting point program" as you call it was more or less information-rich than the "starting values".
I did not address whether the information could be generated--that's a separate issue--but only whether the quantity was sufficient, since I was under the impression that you had claimed it was insufficient. (I agree that for information to accumulate via an evolutionary algorithm, the fitness function cannot be reduced to a random variable by the environment; genetic algorithms have no advantage over picking at random when mapping is essentially random. The world is manifestly nonrandom, though. I will address this point in greater depth if you wish.)
To figure out where the information was, I suggested looking for flaws in maintaining the information, and concluded that defects in coding regions of genes were responsible for most of the observed lethal defects. Thus we can conclude either
* The genome contains most of the heritable information, or
* The genome is by far the least reliable method of storing heritable information or
* The last 60 years of work on molecular genetics is irrelevant.
You seem to be adopting the last point:
quote:
There are several types of functional genotype to phenotype causal relationships which could potentially account for the transformations. Transformations involving clocks and/or chemical triggers for example. The validity of such causal maps is, however, easily tested and disproved. The assumption of a functional genotype to phenotype mapping is thus the assumption of a phenomena which it is known does not and can not exist.
That is an extraordinary claim in light of modern genetics. Can you please either give an example yourself, or cite work consistent with this claim?
It would be good also if you could account for existing experiments in light of your claim that there is no causal map.
Here are a few examples you might wish to consider: color blindness, which co-segregates with the loss of X-linked genes that code for color-specific opsins; Down's Syndrome, which co-segregates with trisomy for Chromosome 21; expression of GFP under the control of the mec-7 promoter in C. elegans, which results in worms that have GFP in their mechanosensory neurons (those neurons in which the wild-type MEC-7 protein is found); the white (w) gene in flies, where w+/w+ (red eyed) crossed with w/w (white eyed) gives entirely red-eyed progeny (w+/w, as w+ is the dominant allele), but that selfing the F1s gives a 3:1 red-eyed to white-eyed phenotype. (Note that the molecular defect in the w gene is known for many alleles.))
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 27. January 2003 08:35
It may be useful to explain that this discussion is not as far off topic as it might at first appear. The subject here is information and the importance of a precise definition of information and information processing for the proper scientific analysis of biological systems. I proposed a mathematical definition of information and information processing based on computer concepts. Based on the definition proposed, it is possible to identify both ‘low volume of information’ models, simulations, and theories and ‘’high volume of information’ models, simulations, and theories.
I claim that while it is possible to construct predictive models, simulations or theories using ‘low volume of information’ as suggested by Darwinian and neo-Darwinian concepts. I also claim it is not possible to VALID predictive models, simulations or theories using low information models (and I suggest this conclusion is relatively obvious). Rex purported to offer an argument to show that ‘low volume theory’ produced a better ‘prediction’ than an alternative model. I demonstrated that his ‘low volume model or theory’ could not produce the purported prediction without inappropriate manipulation of assumptions.
Arguments based on inappropriate manipulation of assumptions arise frequently in evolutionary because of the desire/attempt of some individuals to treat ‘descriptive, non-causal, or pre-scientific’ Darwinian and neo-Darwinian theories or concepts as ‘predictive, testable, mathematical, hard science, causal’ theories and models. The dubious, but widely used, technique for demonstrating that Darwinian concepts can be translated into causal theories involves breaking complex evolutionary processes into a series of simple operations. It is then demonstrated that the each simple operation could be modeled by low volume information models. The flaw in this approach, is that in order to make it work, that different components of the demonstration must be based on different sets of assumptions. As I pointed out, the assumptions needed for these demonstrations are often either known to be unrealistic and misleading, and/or inconsistent from sub-demonstration to sub-demonstration.
Serious scientists have, I believe, always recognized Darwinian and neo-Darwinian theories as descriptive rather than causal or predictive. However, it at least appears that a significant portion of individuals in the scientific/academic community at least tolerate the misleading and misguided efforts to describe and demonstrate Darwinian theories as causal and predictive. Until the ‘assumption manipulation’ issue is recognized and addressed, it will continue to be difficult to have a serious discussions of information and information processing in biological systems.
The interesting and positive aspect of the definition I proposed, is that it makes it possible/practical to define and analyze biological information processing and biological ‘intelligent design processes’ in terms of ‘high volume information’ concepts. The low information volume concepts associated with Darwinian and genetic models and theories appear to be a scientific dead end. The high volume information concept/approach, I suggest provides an opportunity for interesting and productive scientific analysis.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 27. January 2003 14:51
Warren, I agree that this discussion that would seem to many to be quite sidelined actually has some relevance to the main topic theme -- and will expand on this.
quote:
Arguments based on inappropriate manipulation of assumptions arise frequently in evolutionary because of the desire/attempt of some individuals to treat ‘descriptive, non-causal, or pre-scientific’ Darwinian and neo-Darwinian theories or concepts as ‘predictive, testable, mathematical, hard science, causal’ theories and models.
If there were a grain of truth to this, it would indeed impact on issues of whether an AI could also create information by techniques such as genetic algorithms -- and as such would be relevant. So in that we are agreed.
But Warren,
Please explain to us why the techniques used by the Census Bureau, baseball score keepers, and weather forecasters are not “manipulation of assumptions” that mislead the public?
Without understanding of how momentary locally time varying effects can have an overall trend effect on a system, we cannot reach any further agreement or even have a discussion that will make sense to the majority of readers, in my opinion.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 27. January 2003 16:50
I posted (a version of) this earlier, but apparently the changeover of ISCID blew it away. I'll try to reconstruct it.
warren wrote quote:
As RBH concedes, low volumes (sic) processes can not find an adaptive genetic solution unless it is assumed the environment as it impacts survival is stable. You now are claiming that the analysis being offered is based on an environmental assumption which directly contradicts the assumption necessary to produce genetic change. Needless to say, inconsistency in assumptions is also misleading.
That's false. RBH conceded no such thing. What I actually said was: quote:
Evolutionary biologists and researchers in allied disciplines know that environments change on all time grains from moments to millenia, and they actually study (i.e., they do actual research, as distinguished from armchair speculating based on subjective impressions gleaned from pop literature and casual observation) the determinants of adaptation at all those time grains!
That is precisely the opposite of what warren attributed to me. Please do not misrepresent what I say, warren.
As I read my previous posts, this is what I see me as saying:
1. Environments display change on a wide range of time scales ranging from moments to millenia. "Adaptation" as a general concept occurs on all those time scales. Sometimes that adaptation is the direct result of evolutionary processes, sometimes it is the indirect result. See below.
2. Populations and organisms have a variety of adaptive mechanisms for dealing with varying environments, some of them produced directly by biological evolution over generations and millenia, some of them in the form of evolved wired-in automated mechanisms that adapt to moment-to-moment changes, and everything in between. Some of them are the direct product of biological evolution in populations; some are the indirect result of evolution. Some exemplify learning during the lifetime of individual organisms, some represent genetic changes occurring in populations over many generations. None are foreign to science! Those blind, stupid, bullheaded scientists actually do study them, contrary to warren's misrepresentations. In contrast to warren's distorted caricature, scientists know, for example, that some aspects of the physical environment are relatively stable - gravity and the percentage of oxygen in the atmosphere are two examples. They also know that some properties of environments change on multiple time grains - climate (centuries and millenia), weather (days and weeks and seasons), and snow or not this afternoon (hours to days). And by golly, their evolutionary models can actually handle that kind of variation! It seems they're not as stupid as warren implies they are.
3. A huge amount of "high volume" information processing capacity is wired into the sensory apparatus of organisms, enabling them to process high volumes of information in real time. To give but one example, the neural wiring of retinas in mammals performs an enormous amount of information processing from second to second. To give one instance, all mammalian visual systems make heavy use of lateral inhibition in the neural wiring of the retina. The effect of lateral inhibition is to make discontinuities - rapid changes in time and/or space - in the visual field more salient while rendering static aspects of the visual field less salient. That represents a whole lot of information processing before the visual signal even leaves the periphery.
4. Warren continues to conflate the "organism" unit of analysis with the "population" unit of analysis. That's an important difference.
5. Gedanken's remarks on averaging are on the mark. I recommend that warren read a good population genetics text for the math dealing with how small reproductive advantages can come to dominate a population over generations in the face of variability.
6. Finally, I routinely evolve populations of artificial organisms in EAs to adapt to complex changing environments that are piped into my computers from the outside world. The environments vary on a relatively fine time grain (sometimes substantially, sometimes less so; sometimes rapidly, sometimes slowly). Some environmental changes are often faster than the population can evolve to track the rapid changes directly.
The populations in my computers typically evolve two basic approaches to adaptation in these circumstances, both approaches generally operative at the same time. First, they evolve to extract whatever invariant information there is that underlies the short-term variations if sensing those invariants contributes to the reproductive advantage of the artificial organisms carrying the genes that encode for the ability to detect that information.
Second, they often also evolve short-term tactics for dealing with too-rapid or too-extreme environmental variation that the population as a whole cannot evolve to track directly via mutations, recombination, and selection. For example, one tactic I often see them evolve is their equivalent of "hiding" until the environment settles down some.
Thus in a population of artificial organisms one sees both adaptation to stabilities in the environment and the evolution of tactics to deal with - adapt to - short-term environmental variation. And the population acquires both sets of adaptive mechanisms using regular old random mutation, recombination, and natural selection.
Those adaptive traits do not come cheap. Over the tens of thousands of generations that our EAs run, millions of artificial organisms die. They are eliminated from the breeding population. The EA "chooses" the surviving population, discarding millions of alternatives over thousands of generations. In information theoretic terms those discarded alternatives are not negligible and must be taken into account in measuring the informational "throughput" of the system. And the EA's "choices" are made over generations of evolution using good old random mutation, recombination, and selection.
RBH
[ 27. January 2003, 16:59: Message edited by: RBH ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 28. January 2003 20:57
Well, so much for my last message. The glitch was indeed as bad as advertised. I had hoped that this was just pessimism.
Anyway, the basic points were
- This is an overly specific yet on-topic discussion as far as I can tell, since if Warren is right, I see no way to get his teleological mappings to work without invoking a transcendent entity.
- I continued to complain, more insistently this time, that Warren pay attention to my examples showing extensive support for a causal relationship between genes in that a specific gene product is often necessary and sufficient for a given function or developmental process, in the context of an otherwise largely normal organism.
- Some other stuff which I forget.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 29. January 2003 08:38
Rex,
Quote: I continued to complain, more insistently this time, that Warren pay attention to my examples showing extensive support for a causal relationship between genes in that a specific gene product is often necessary and sufficient for a given function or developmental process, in the context of an otherwise largely normal organism.
As I stated in the lost response, genes and/or the initial egg cell may be necessary for the successful development of the full grown organism, but clearly and demonstrably they are not sufficient. Based on what is currently known of genetics, the number of possible developmental pathways for a given set of genes is very, very large. Only a very, very small portion of those pathways will produce a full grown organism capable of reproduction. Again based on current knowledge, it can be demonstrated that the probability of a full grown living organism developing from an initial egg cell is far less than 1 in 10^1000 or even 1 in 10^10,000. In the absence of powerful, within lifetime information generation processes, the probability of producing a living, functioning, reproducing organism from an initial fertilized egg containing a full set of genetic material is essentially nil.
This can be demonstrated either in terms of genotypes or in terms of phenotypes. In terms of phenotypes, we know that the for an organism to successfully develop and reproduce it must avoid dying. To avoid dying, the organisms must at different points in its life achieve ‘adaptive states’ or phenotypes. An adaptive state or adaptive phenotype is a member of the adaptive subset of the set of possible forms the organism can take at a point in time. if we look at cells making up a multi-cellular organism at any point in time, we known there are a very large number of different types of cells all developed from a single set of genes and single initial fertilized egg cell. The ‘adaptive set of cell types’ at any point in time is a very, very small subset of the set of possible cell types at a point in time.
In terms of genetics, we know that the assumption that genetic material does not change or ‘evolve’ during the organisms lifetime is false. In fact, we know that the state or form or regulatory genes are continually changing (evolving) during the organisms lifetime. If you look at the number of possible paths defined by possible changes in regulatory genes during an organisms lifetime, and if we look at the portion of paths that are likely to result in the development of living reproducing organism, we know there is not and can not be a causal relationship between genes and fully developed organisms.
The claimed functional relationship or mapping between genotype and phenotype is descriptive not causal. The mappings are a ‘simplifying fiction’ used as a starting point in analysis. The claim that the relationship is actually causal or functional is based on intentionally ignoring known facts.
Quote: This is an overly specific yet on-topic discussion as far as I can tell, since if Warren is right, I see no way to get his teleological mappings to work without invoking a transcendent entity
People seem to have no difficulty accepting that mechanical computers can perform huge volumes of information processing and that the computing capacity of computers can be increase geometrically. People seem to have no difficulty accepting that humans have the capacity to produce designs which require the equivalent of massive amounts of information processing/information generation. At the same time, people seem to have a hard time accepting the possibility that biological systems could have ‘evolved’ powerful information processing/information generating capabilities.
I have proposed definitions that makes it practical to quantify the volume of point in time information. I have also proposed definitions that make it possible to quantify the power, force or intelligence, needed to generate point in time information. I can also offer definitions that will make it possible to quantify creative intelligence. All my definitions are based on the mechanical or scientific concept of teleology that has been around for at least 2000 years.
Based on the definitions provided, I claim that living organisms such as humans have the ability to perform the ‘mathematical’ equivalent of 10^10,000 and probably 10^100,000 cycles of teleological information processing per lifetime. It is important to note that with biological systems, as with electronic computers, there are both hardware and software ‘gimmicks’ that account for the very high ‘equivalent’ volumes of information processing. The measure of processing power is defined in terms of a very simplistic and inefficient random search process.
One of the major roadblocks to understanding information processing in biological systems is the strong and widely held belief that ‘low volume information’ models can explain biological design processes. An even more serious roadblock to understanding information processing in biological systems is the ‘inappropriate manipulation of assumptions (and facts) which are used to lend credibility to the low volume information models and theories.
Back to your original point. Not only is the claim of a causal or functional relationship between genotype and phenotype false, it is clearly known to be false.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 29. January 2003 08:46
This is a reposting of my response to gedanken
This response to gedankan should probably be under a heading ‘The Nature of Scientific Causation’. I will be glad to move it.
Science or hard science can be characterized as an iterative three stage analysis of causal relationships as follows:
STAGE 1 OBSERVATION: Scientist evaluates observed/recorded/verifiable occurrences of a relationship, develops a predictive mathematical/logical model to fit the set of observed occurrences and formally expresses the model as a theory in the general form "Under ideal conditions A causes B’. or "Under ideal conditions there is a function F such that F(A)=B
STAGE 2 TESTING AND VERIFICATION: Scientists independently develop perform formal analysis to test and validate the predictions generated by the theory, and to more precisely define where and how the theory is applied. If the theory is falsified then return to stage 1.
STAGE 3 APPLICATION: The tested and validated predictions of the theory are used to reliably anticipate or predict future events. These predictions are useful in solving practical problems. If the theory does not provide the reliable predictions needed for problem solving then return to stage 1.
The hard science, or predictive science, or ‘real science’ process or paradigm only produces useful, productive results for causal relationships or natural laws which satisfy or which are reasonably expected to satisfy the principle or assumption of scientific determinism. Scientific determinism is the requirement that a causal relationship must have the same logical form applicable to past observations, current testing and validation, and future predictive applications.
Now to gedanken’s question about statistical models and theories. We can collect observations that indicates B1 follows A 10% of the time. Based on this observation we can formulate a statistical theory of the form "Under ideal conditions, A causes B1 10% of the time".
Using the ASSUMPTION that the observed relationship will repeat in the future, we can perform various tests to validate or falsify the proposed theory. The proposed theory can produce reliable predictions and there are numerous uses for this type of statistical theory. A statistical predictive theory is, however, intentionally misleading if we have reason to know or suspect that the assumptions used are false and misleading. A statistical theory is intentionally false and misleading if we know or have reason to believe that the observed pattern is unlikely to repeat in the future.
As we have been discussing, a similar problem of false and misleading assumptions seems to arise with respect to evolutionary and genetic theories. Darwinian and neo-Darwinian concepts, taken at face value, suggest that evolutionary change can be modeled, simulated, and explained by ‘low volume information’ type theories( theories based on 1 cycle per lifetime of information processing). As was discussed, it is not possible to fit such models to known data with out inappropriate manipulation of assumptions.
The problem here, IMO, does not seem to be as much with ‘evolutionary science’ as with ‘academic interpretations of evolutionary science’. I would think that most scientists would be reluctant to claim that evolutionary science currently has valid predictive hard science causal theories of evolutionary change processes. Anything in the area that might qualify as a predictive theory has a very limited scope.
While scientific theorists may be reluctant to claim the existence of valid predictive theories, academics who claim to be authorities on evolutionary theory, quite clearly are not unwilling to make such claims. If these claims take the form of low volume information models, then it can be readily demonstrated that the claims are based on inappropriate manipulation of assumptions. If the claims for a valid predictive theory are based on something other than low volume information, then the theories have not been explicitly formulated.
Demonstrating that you can not construct a valid low volume information model or theory of evolutionary change is, IMO, fairly simple and fairly obvious. It is not even possible to construct such models without inappropriate manipulation of assumptions. However, as is apparent from the discussion here, there exists a considerable difference of opinion as to what constitutes rigorous and/or acceptable standards of mathematical modeling. The issues raised here will not be resolved without explicitly defining and testing mathematical models and concepts. While such formal mathematical analysis may be interesting, I can’t see it being performed effectively in this forum. Given the bias involved, it is doubtful if most evolutionary biology academics would be qualified to even participate in such formal analysis.
DYNAMIC AND TELEOLOGICAL CAUSATION
As I have discussed elsewhere, a scientific theory based on a dynamic and teleological causal relationship has the general form "Under ideal conditions, the process A causes or produces the teleological process or relationship B which increases the likelihood of achieving goal G". It is possible to have multiple levels of such theories. For example, associated with the theory described above, you could have a theory "Under ideal conditions, process A1 causes or produces teleological process or relationship A which increases the likelihood of goal G".
As I define it A1, A and B are all ‘dynamic and teleological causal relationships’ and ‘natural laws’. This type of theory can be expressed in terms of sets of mathematical functions(solution spaces ) H and H1 containing functions F and F1 respectively where F(A)=B and F1(A1)=A. Typically F and F1 would be expressed as computer or logic machine programs.
The question of whether such theories are acceptable in conventional science is not entirely clear. It can be argued that engineering involves models and theories which fit this format. As gedanken points out, theories constructed from genetic algorithm models might be expressed in this form. Whether or not such theories are acceptable in conventional science, it seems no such theories of evolutionary change are available for discussion (accept the theories I proposed).
A major part of the problem with discussing teleological causation is the tendency to intermix ‘scientific teleology causation’ with ‘metaphysical teleology causation’. In today’s terminology, scientific teleology is a complex form of causation which can be modeled or expressed in terms of the operations of logic machines and information processing.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 29. January 2003 11:49
[A version of this was also lost, I assume due to the transition. I am assuming it was transition loss and will keep reposting versions of my posts. And Warren’s post has changed (or a different set is now up), so my response is modified.]
quote:
The hard science, or predictive science, or ‘real science’ process or paradigm only produces useful, productive results for causal relationships or natural laws which satisfy or which are reasonably expected to satisfy the principle or assumption of scientific determinism. Scientific determinism is the requirement that a causal relationship must have the same logical form applicable to past observations, current testing and validation, and future predictive applications.
I notice that Warren chooses to no longer dispute the essential nature of averaging and analysis of trends, rather now is attacking the causal relationship of genetics. He has removed the “dynamic” changing conditions aspect from his claim of “manipulation of assumptions”. The reason, of course, is that there is no answer that would make that former argument seem reasonable.
I’ll repeat (to counter the old claims) that scientific and statistical techniques of study of dynamically changing conditions (including those used by the Census Bureau, baseball score keepers, and weather forecasters) can indeed effectively find trends of different time granularity. These are how slow processes (like population change) can change in a particular direction showing a trend, even though the individual events that effect or cause that trend are dynamically changing, sometimes radically variable. Warren has appropriately dropped the existence of dynamically changing conditions as part of his charge of “manipulation of assumptions”.
An aside note, does increase in population depend on a causal relationship in the prediction? And similarly for baseball scorekeeping? Where is the causal relationship in batting average? Such causal relationships exist, but they are not explicitly necessary in finding the trends useful. Weather forecasting is of course more causal in its assumptions.
Then as to the causal nature of genetic materials, Clearly the causal nature of genetic material has been demonstrated in “real science” (as Warren calls it) experiment after experiment. Insertion of new genes causes changes in the new organism, and this is demonstrated by extensive testing not only of the effects, but also including studies of the mechanisms of such causality. I’ll leave the rest of this to those more expert than myself.
quote:
Demonstrating that you can not construct a valid low volume information model or theory of evolutionary change is, IMO, fairly simple and fairly obvious. It is not even possible to construct such models without inappropriate manipulation of assumptions. However, as is apparent from the discussion here, there exists a considerable difference of opinion as to what constitutes rigorous and/or acceptable standards of mathematical modeling. The issues raised here will not be resolved without explicitly defining and testing mathematical models and concepts. While such formal mathematical analysis may be interesting, I can’t see it being performed effectively in this forum. Given the bias involved, it is doubtful if most evolutionary biology academics would be qualified to even participate in such formal analysis.
I find it rather laughable that Warren claims that ISCID forum is a place with such bias against his theory of teleological causation. (Or did Warren mean that his theory contains such inherent bias that it could not be analyzed by normal mathematical and scientific techniques of requirements that it have mathematical logic and conform to observed data.)
Most significant is that the argument is now reduced to simply the claim that genetic material is not “causal” in the development of organisms.
This is significant to this thread in that genetic algorithms study “causal” relationships in that the function of a given trial is tested for performance. Genetic algorithms test randomly generated structures to see if they are high in performance by measures that exist in the GA environment, including both dynamically developed competition and externally defined relationships. Of course they would not be relevant to real-world evolution if in the real world individuals did not inherit characteristics. But how Warren can deny the importance of inherited characteristics, or of the high degree of causal relationships found in physical nature and analyzed by science is beyond my comprehension.
[Edited:]
The post by RBH is significant in understanding the difficulties here, and I highly recommend reading that post above.
I withdraw my point about Warren removing his claim about the dynamic changing environment not having a trend effect for the processes of evolution to work in the population level. Warren simply removed it in response to my points (where it was obviously demonstrating failure), and placed all these claims in response to Rex Kerr where the subject was on causal nature of genetic material. In other words the responses were divided so that the arguments to me were of items in Rex Kerr’s area of expertise, and the arguments on Rex Kerr’s points included the issues of dynamic change that I showed to be irrelevant.
Warren’s response to Rex:
quote:
People seem to have no difficulty accepting that mechanical computers can perform huge volumes of information processing and that the computing capacity of computers can be increase geometrically. People seem to have no difficulty accepting that humans have the capacity to produce designs which require the equivalent of massive amounts of information processing/information generation. At the same time, people seem to have a hard time accepting the possibility that biological systems could have ‘evolved’ powerful information processing/information generating capabilities.
If this is the case, I don’t understand how Warren argues against evolutionary processes. This view of what he is saying is simply that biological systems in their physical embodiment process tremendous amounts of “information”, and that better ways of studying this could be useful. I’m sure biologists would agree with this -- I don’t see this version as having any basis for a claim against evolutionary theories as they stand.
In other words, there may be areas of “information processing” that have not been expanded on as much as would be useful in understanding biological systems. That this is the case is not an argument that the theories of evolution are flawed, rather it is an argument that more can be learned. But Warren wants to insist that the “more that can be learned” is inherently an argument against the theories of evolution. The argument doesn’t work.
[ 29. January 2003, 12:24: Message edited by: gedanken ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 30. January 2003 11:44
Gedanken comment raise four general issues which need to be addressed- 1)the nature of statistical theories, 2)the difference between a ‘descriptive’ and ‘causal theories’, 3) consensus versus scientific validation, and 4) the process of scientific validation and number
STATISTICS AND THEORY CONSTRUCTION
Quote: I’ll repeat (to counter the old claims) that scientific and statistical techniques of study of dynamically changing conditions (including those used by the Census Bureau, baseball score keepers, and weather forecasters) can indeed effectively find trends of different time granularity.
As I stated, there are many valid uses for predictive statistical theories and for theories that incorporate statistical concepts. However, the development of a predictive theory based on statistical relationships requires that certain knowledge be available and that certain assumptions be explicitly defined.
The claim or suggestion that ‘probability distributions’ can be used to explain dynamic environmental conditions in evolutionary theory is patently ridiculous. You can not claim that survival and reproduction depends on adaptation to average phenotype conditions. An organism adapted to an average environmental temperature will not necessarily survive the range of temperatures that may occur.
In formulating vague, non-scientific, non-causal, non-predictive, descriptive theories it may be acceptable to make vague and undefined uses of concepts like statistical fluctuations. However if statistical fluctuations are to be used appropriately in theory construction, a great deal of information is needed to justify the assumptions used.
DESCRIPTIVE VERSUS CAUSAL THEORIES
To begin, it is my understanding that scientists recognize a clear distinction between a descriptive and a causal relationship. It is my understanding that scientists clearly recognize that ‘modeling or enumerating the stages of a change process’ is fundamentally different that ‘modeling a causal process or causal chain’. If as you suggest, scientists today claim that the ‘recognized stages of development’ constitute a ‘scientific causal chain’, then I think you need to demonstrate that scientists are actually making such a claim.
The claim in evolutionary biology for a causal or functional relationship between genes and mature organisms is not for a known or identified causal or functional relationship but for a speculative, possible, as yet undefined relationship. Since the causal relationship has not been defined or identified, it is not possible there to be evidence supporting such a causal relationship.
My counter claim with respect to developmental processes is easily expressed mathematically. I claim that for any given fertilized egg, there are a very, very large number N of possible developmental paths that could follow for a given fertilized egg. I claim that of the possible developmental paths there is a much smaller number Nf of the N possible developmental paths which will result in a mature organism capable of reproducing. I am suggesting that for a complex multi-cellular organisms the ratio of N to Nf is at least 10^1,000 to 1. An organism, I suggest, would not successfully develop and reproduce without the existence of processes and mechanisms capable of generating the information needed to find one of the highly improbably adaptive developmental pathways.
Note that the issue here is not the existence of a causal chain of events from the single cell to the grown organisms. The issue is what type of causal chain is involved. Specifically, the question is whether the simplistic type of causal chain required by Darwinian concepts exists.
CONSENSUS VERSUS SCIENTIFIC EVIDENCE
Quote: I find it rather laughable that Warren claims that ISCID forum is a place with such bias against his theory of teleological causation.
Because almost all scientists, including ID scientists, agree that some type of low volume information theory based on Darwinian concepts can explain at least some simple types of evolutionary change, does not mean the belief is valid. Because almost everyone agrees there is a simple functional/causal relationship between genes and full grown organisms does not mean that the belief is valid.
Three broad classes of teleological causation can be identified- 1)low volume information teleological causation, 2)high volume information-rapid processing teleological causation, and 3)intent or consciousness based teleological causation. The first type of teleological process involves finding adaptive solutions to the problems of survival using the equivalent of less than a few billion processing cycles per lifetime. The second type of teleological causation involves finding adaptive solutions with the equivalent of more than 10^1000 cycles per lifetime. The third type is what I would label non-scientific or metaphysical teleological causation. High volume-rapid processing teleological causation is not generally recognized or understood by either most ID supporters or most supporters of Darwinian evolution.
It is important to note that the consensus in evolutionary biology is not that -"Evolutionary change can be explained by a known, predictive Darwinian theory’. The consensus is that "It is likely that we will someday be able to formulate a predictive Darwinian type theory which can explain evolutionary change". As should be obvious, there is an major difference between those two statements. Many ID proponents would accept the second statement if it was modified to "which can explain materialistic evolutionary change".
The general acceptance of Darwin, I suggest, is due in large part to the lack of a viable alternative. High volume-rapid processing teleological causation, I suggest, is the alternative that can lead to the development of explicit predictive, scientific theories of biological design processes including evolutionary change.
VALIDATING HIGH VOLUME-RAPID PROCESSING TELEOLOGY
It is fairly easy to demonstrate that current formulations of low volume-Darwinian type theories are based on intentional misuse of assumptions. It is much more difficult, and much more mathematically complex to demonstrate the validity of models and theories based on high volume-rapid processing teleology.
When I stated that ISCID was not an appropriate forum, I simply meant it is difficult to address complex mathematics where the rules of presentation are not compatible with the format of such analysis. When I suggested that some individuals with scientific credentials are too ‘biased’ to engage productively in such a discussion, I meant that some individuals insist on using personal subjective authoritative opinions rather than arguments and analysis to support their positions.
Validating models and theories based on high volume-rapid processing concepts of teleology is a lengthy and complex process. The first step in the process involves defining and analyzing the mathematical concept of high volume teleology. This is accomplished by creating abstract mathematical logic machines with the capacity to find ‘adaptive solutions’ to complex problems by interacting with an abstract mathematical external environment. These logic machines are controlled by programs which are in some respects similar to genetic algorithm programs.
Using these high power teleological logic machines, it is then possible to define/explain how such logic machines can exhibit both intelligence and creative intelligence. In addition, it can be demonstrated how such machines can evolve increases in intelligence and creative intelligence.
Once the mathematics of high speed teleology has been demonstrated, the next step is to show how these complex mathematical machines can be used to model the teleological behavior of biological systems. Finally, it needs to be demonstrated how these models can be transformed into testable, predictive causal, scientific theories.
There is a clear if somewhat complicated path to validating high volume-rapid processing concepts of teleological causation.
IP: Logged
|
|
|
Printer-friendly view of this topic