Past Posts of Interest

I've noticed the problem of having to cross-post your significant ideas is pretty common. One solution that I've found to be very nifty is Planet, which uses RSS feeds to create a kind of "joint blog" with many contributors who maintain separate blogs. See http://planet-z.philisha.net for an example that I set up with my friends. That might be worth looking into; you can find more at http://www.planetplanet.org. I was at your lecture tonight, and the first question really made me think. ("Is looking for patterns that indicate design 'science' in itself, or is it 'forensic science'?") If you don't mind my saying, your answer disappointed me somewhat. I believe the question presupposes the idea that "true science" should only be subject to things that are also science. This presupposition seems to me a somewhat "fixed deck" if you will--you are granting reductionist naturalism a lot of ground if you grant it. You said that the definition of science is not fixed, and it's difficult to say whether or not a thing (such as "forensic science") qualifies as science. (Forgive me if I misrepresent your remarks; I could be wrong about this.) My reaction is that one should be eager to reject the notion that scientific studies should only be subject to criticism that itself is scientific. Denying that certain types of evolutionary research (ahem--big bang) require a good deal of philosophy is one of the oldest tricks in the book. Where would science be if it were not subject to the Law of Non-Contradiction, a principle that cannot be said to be strictly within the realm of science. If you'll forgive the absurdity, spelling is not science, yet hopefully everyone holds that scientific research should be subject to its demands. Pardon me if my remarks are too obvious or if I sound like I am simply parroting Johnson's _The_Right_Questions_. I just want to emphasize the subtle danger of considering natural philosophy too high to be subject to the claims of other disciplines.philh

April 18, 2005

April

04

Apr

18

18

2005

10:04 PM

10

10

04

PM

PST

Copy Comment Link

This is an interesting exchange concerning a paper you wrote (linked below) between Sweetness (an obvious pseudonym) and me concerning the limits of evolutionary potential. Sweetness is a molecular biophysicist. His comments are marked by a single carat (>) and my unmarked responses follow. Double carat (> >) are also my comments from an earlier exchange to which Sweetness is responding. http://www.designinference.com/documents/2005.03.Searching_Large_Spaces.pdf > >He mentions a very > >low ratio 1 in "10^130". In order to get such a target > >ratio in sequence space, you would need a protein sequence > >with an equivalent of 100 fully specified residues. > >So, you are wrong. He does mention the degree of > >complexity he is talking about. > > Oh, for crying out loud. Can you > READ? He gives this example in his > discussion of a blind search. Then > in the very next chapter he explains > what is an assisted search. Dembski says, "For a search procedure to be more effective than blind search, it therefore needs to provide additional information." He basically does the basic math for blind searches and concludes that blind searches cannot find target sequences with very low ratios in sequence space. He then looks at the evolutionary assumption of "assisted search." Dembski defines assisted search as "any search procedure that provides more information about candidate solutions than a blind search." - like an "Easter egg hunt" where different sequences are not either "right or wrong" but "warmer or colder". In an assisted search, there is a gradient in sequence space where nothing is ever completely right or wrong. Rather, everything is in shades of gray. This means that from every position in sequence space there is an indicator of direction, an arrow if you will, pointing toward a target island. The searching sequence can somehow recognize which way this arrow is pointing and follow it, step-by-step, to the target. Dembski rightly concludes that information for such an assisted search, if available, would dramatically improve the odds of finding a very rare target sequence. Such assistance is very "effective" if present. Dembski then asks the all-important question. That is, "What is the source of u [assisted search] that makes a successful search possible?" With me he notes that all known sources of successful searches that actually work in finding rare targets on a consistent basis are based in intelligent or higher-level outside informational input that goes beyond the system in question. But, he asks, "Could they not simply be brute givens, gifts that a profligate environment bestows on certain searches without any need of explanation? In particular, might not the measure μ simply be a free lunch?" Dembski approaches this most important question by re-summarizing the problem and the fundamental need for assisted search if evolution is to be true. He writes: "Let’s examine these questions more closely. First off, it should be clear that an assisted search needs to input a substantial amount of novel information to make the search successful. The problem with blind search is that as the search proceeds, no signal is coming back to the search to ensure that it is getting closer to the target. Blind search offers extremely limited feedback and no sense of approximation to a target. Assisted search, by contrast, attempts to provide the feedback necessary for such approximation." In the rest of the paper Dembski argues that only intelligent input can provide the necessary novel information - that non-intelligent processes simply do not have access to the type of information required. Personally, I agree that more information, than is contained by the system in question, is required for the search to be successful. This information need not reach the level of what we may call "intelligence". It just has to be higher-level information that affects the search in a way that gives directional information to the search sequence in regard to the target sequence. In other words, the one that provides the information must know something about the target sequence ahead of time with respect to the location of the search sequence and the search sequence must be able to interpret/understand the information that is being provided. In short, there must be a way to base selection between two or more options on higher information than simple random chance. There must be some relatively improved advantage in picking one option over another. The question is, can anything in mindless nature provide such a biased directional bias when it comes to sorting out random mutational events? Natural selection has been proposed as the most likely mindless biasing agent that drives evolution. The problem here is that natural selection can only select based on functional differences between sequences in sequence space. Why is this a problem? - Because we know that two or more different yet equally non-beneficial sequences will be neutral with respect to selectable function. Again, it is like picking quiziligook over quiziliguck when neither one means anything more or less beneficial in a given setting/environment. That's the problem in a nutshell. An assisted search is vital to evolution every step of the way and yet there is no adequate mindless candidate for such assistance beyond very low levels of functional complexity where target sequence become extremely rare and entirely and extensively surrounded by equally non-beneficial sequence options. > He then proposes something he calls > "No Free Lunch Regress", where he > says that blind searches for small > targets in a large space are > unlikely to succeed; that therefore > the search needs to be assisted; > that such a search, however, resides > in a higher space of its own, and > needs to be searched for either blindly > or by another assisted search > that is even higher level. Exactly . . . Finding the right directional biasing agent for a given target in a given search space requires a blind search. > This is a completely different type > of mathematics then the basic > algebra you have been using. Not at all. Dembski is simply trying to show the futility of actually coming across a mindless biasing agent that actually works for finding a particular target sequence in a particular sequence space via mindless processes alone. The logic and the math are actually quite intuitive. > In fact, according to it, what you > consider to be a problem (search > in sequence space) isn't really a > problem; the problem, according > to Dembski's paper, is finding the > assisted search in the larger > environment. The reason why finding the larger assisted search algorithm is even necessary is because finding a rare target sequence via any sort of random search is impossible. That is what makes looking for some sort of non-random assistance so vital - - which is exactly what I've been saying. > His contention is, of course, wrong > on several points. One is > particular to his paper: > > - The regress stops when it > hits the ultimate arbiters, the > well-defined natural laws. These are > as they are, and nobody can claim > anything (so far) as to their origins. > Dembski can, if he wishes, argue > that they are designed by an > Intelligent Designer, but he will be hard > pressed to provide a probabilistic > analysis of how probable our > universe's set of laws is as compared > to some other set of laws. Actually, Dembski argues that the regress never stops short of intelligent design. That's just it. No rational "mindless" cause can be found to assist the search at such higher-levels, which have very low target ratios, short of ID. Certainly the mindless processes of natural selection are not informationally complex enough since they cannot preferentially select between equally non-beneficial sequences (like quiziligook and quiziliguck). Only an agent with a great deal of pre-established knowledge/information concerning the target sequence can judge between completely non-functional and/or non-beneficial sequences with regard to the needed functionally beneficial target sequence. A mindless nature simply does not have this pre-established knowledge. That's the problem for the ToE. > One overlaps with your hypothesis: > > - The regress is significantly > constrained at the first step by the > structural considerations. Pretty > basic math can show that even > random-walk searches of sequence > space within a single subset of > structure space (say, all beta-barell > proteins) is much faster then the > random search through unconstrained > sequence space of the same size. We > know from observation that all > existing proteins are clustered around a > few dozen basic overall folds, > each defining one such constrained > subset. Therefore, there are NO KNOWN > PROTEINS that would have EVER > required long random walks to some > poorly defined "target sequence". Again, you forget that protein based functions are not made up of a single type of fold. At higher and higher levels of minimum functional complexity (i.e., minimum sequence size and specificity), these basic folds must be specifically arranged in a specific 3D structure. The greater the minimum size and specificity of arrangement, the more and more isolated that function becomes, structurally, from its neighbors. You think that because subparts of a protein are highly or even completely homologous to all other proteins that the entire protein structure is closely clustered to all other protein structures. This is simply not true. The greater the minimum requirements, the exponentially less and less clustered protein-based functions become until they are truly isolated like tiny separate islands on a vast ocean. > And one is quite ironic: > > - This paper, actually, provides an > interesting view of ID. In terms of > this paper, the whole premise of ID > can be presented as a claim that > there is an intelligent designer, who > serves to stop this No Free Lunch > Regress by acting as the ultimate > "assisted search" mechanism; which > then "projects informatino" downward > onto all other search levels. > However, AFAIK, something no > IDer has ever done is try to present the > probability of existence of such designer. That is because the existence of such a designer is incomprehensible. This does not mean that belief that such a designer exists in illogical or non-supportable. It is just that understanding the ultimate origin of the designer is impossible - beyond human understanding. Even from your evolutionary perspective, there are equally impossible questions that go beyond human comprehension - such as the origin of the "Big Bang" or how higher informational complexity can come from complete chaos (a reversal of informational entropy, if you will, without outside input of information into the system). > I mean, if probabilities of No Free > Lunch Regress apply to everything, > they should apply to the designer > himself. If not, why not? If yes, > what is the probability that an ultimate > assisted search exists that is > also intelligent, as opposed to the > probability of an assisted search > that works, but is unintelligent? This could be, but is it less likely. If the ultimate end to the regress was truly an unintelligent process, this would not only be unexplainable as far as the source of this highly unusual unintelligent process is concerned, but it would have no other parallel in the known universe. At least we know of other intelligent agents that are in fact capable of ending the No Free Lunch Regress. We know of no non-intelligent agents that are capable of doing this. Sean Pitman www.DetectingDesign.com Seanpit@gmail.comseanpit

April 18, 2005

April

04

Apr

18

18

2005

03:45 PM

3

03

45

PM

PST

Copy Comment Link