Significantly more complex V-ATPase

From PLOS One.  Discussed in more detail here. 

Two very distinct mechanisms, which most likely evolved independently, are employed for ATP-driven H⁺ pumps: the rotary mechanism of the V-ATPase and the alternating access mechanism used by the P-ATPases (Fig 1). The significantly more complex V-ATPase consists of 25–39 protein chains compared to a monomeric or homodimeric polypeptide for the P-ATPase. The operating mechanism for the V-ATPase is also more elaborate consisting of an electric motor-like rotary mechanism. In contrast, the P-ATPase operates by switching between two (E1 and E2) conformations similar to most allosteric mechanisms.

How do we measure "significantly more complex" in terms of bits? 
Mechanism is discussed in more detail here. 

one classification of complexity measures

from page 5 of this document

Some interesting overview information in "Complexity Measures in Manufacturing Systems" by DeTony et al.  (Not sure how accurate some of the characterizations of the different complexity measures are.)

conversation with Jothee Shambit

I remember reading something of Barbara Forrest in which she talked about not being rattled by an attempt of the Discovery Institute to caricature her as Barking Forrest.  I had the impression that a Barking Forrest character (a dog joke?) was invented, and that this name was intended as some sort of attack on her.  I got curious.  No, it turned out to be an explicitly fictional interview with Barbara Forrest in which the fictional interviewer himself repeatedly gets her name wrong in ways that have no bearing on what she's saying.  The silliness of the interviewer is itself a parody of radio show hosts.  However, I sort of like the idea of a fictional stand-in for a caricature.

Jothee Shambit

So ... A fictional online interview  with Jothee Shambit, based on Jeffrey Shallit's writings on the Recursivity blog, some of his other writings & depositions, and similar sentiments by other ID-critics; so here in the alternate universe where Jothee rather than Jeffrey has taken part in the Dover trial and Shallit's various other exploits.  As a public figure, he is subject to satire and parody.  More than a bit of hyperbole in places here.  Nonetheless, you can decide for yourself whether it bears more than a superficial resemblance to Shallit's attitudes and arguments.  Any resemblances to people living or dead might not be a coincidence.  The reader can decide how much of it is caricature given how gracious and genteel Shallit is in actuality.

Adleman's K-potency and Kauffman's Atoms

One of the motivations for the sequence-space probability dispersion matrix is that as a model it might estimate the computational depth of nucleotide sequences, or the relative depth between two nucleotide sequences.   How deep is a given nucleotide sequence?

Kauffman writes in a recent foreword that the universe has produced every kind of atom it could produce (an ergodic process, whereas enumerating the realizable proteins is a non-ergodic process), but Leonard Adleman has elsewhere written in "The Rarest Things in the Universe" (among "entropy and information" here) that atoms with higher counts of protons than we've thus far encountered could be considered to have larger depth (and thus be somewhat analogous to Kauffman's sequences).

I am not a physicist, but I suppose it is possible to theorize about an atomic nucleus with a million protons. But what if I want to create one? It appears that producing transuranic elements takes huge amounts of time/energy and the greater the number of protons, the more time/energy it takes. It is even conceivable (to me at least) that there is not enough time/energy available (at least on earth) to actually produce one. Like the prime factorization of , it may exist in theory but not in reality. On the other hand, physicists from Russia and America, using lots of time/energy, have created an atomic nucleus with 118 protons called Ununoctium. Ununoctium is analogous to Childers’ prime factorization; both exist in reality; both were very costly to create.

In his 1979 paper "Time, Space, and Randomness," Adleman develops an idea about "K-potency" motivated by an analogy with thermodynamics, specifically chemical reactions that take much less time going in one direction than the other.

This approach to "one-way functions" is important to crytography, incidentally.  His definition for K-potency follows here:

The Whistling Herring

 "He eteþ no ffyssh But heryng red."

The dialogue's not perfectly accurate but is more or less the MirrorMask version of an old Yiddish joke:

What’s green, hangs on a wall, and whistles?

Gryphon: I give up, I think, no wait, wait… Fine. What’s the answer?
Helena: Okay. It’s a herring.
Gryphon: But a herring isn’t green.
Helena: You can paint it green.
Gryphon: But a herring doesn’t hang on a wall.
Helena: You can nail it to a wall.
Gryphon: But a herring doesn’t whistle!
Helena: Oh, come on. I just put that in to stop it from being too obvious.

This puts me in mind of Dembski's tractability condition.  (In more recent developments of algorithmic specified complexity, the tractability condition is built in as a feature of randomness deficiency.)  Even if it's true that a herring can be green because of paint, and even if it can be made to hang on a wall, the effect is to include anything that can be painted green and hung on a wall.  These increase the specificational resources of the target space so that it's no longer specific.  I can paint a squirrel green and stick it on the wall.  Or a carrot.  Or a handkerchief.  Which is precisely the joke. Once I include things that aren't normally, typically green, as well as things that don't typically, normally hang on a wall, the riddle becomes intractable.   

islands of functionality and the flash of genius

Hazen et al illustrate some alternative ideas of functional protein accessibility in protein space, where the plane represents the dimensionality of protein space (2 is much, much fewer than what would be needed) and the E axis represents the catalytic usefulness of the various points in protein space (in essence, a fitness landscape):

D has more of a needle-in-a-haystack problem than A, B, or C, due to its relatively small hypervolume (area in the figure) of protein space.  But it's not the only aspect that makes in inaccessible.  The relative distance from other islands of functionality.  Using the above figure a little out of its intended representation, we see that B and C are relatively close together, so that not as vast a distance of neutral mutation would have to be crossed to get from B to C as from A to D.

But vast oceans of neutral mutation to be crossed are not the only impediment to finding the points of especially high functionality ... There is also the fact that less optimal peaks might serve as attractors that divert computational resources away from the brass ring.  In this case, the good is the enemy of the best.

D of Hazen et al's figure above corresponds to this diagram from one of Douglas Axe's papers, where sequence/protein space is represented by only one dimension:

Here the white noise of suboptimal adaptations might be considered negligible, all of the being more or less neutral in that they don't change the survivability rate of the organisms enough to inhibit the traversal of sequence space.  It is possible that the neutral is filled with many little hills and valleys, a so-called rugged landscape.  In the Picasso-esque landscape below, it may be that the difficulty in finding the high peak of innovation is compunded, both by the volume of sequence/protein space to search but also by the attractive "force" of suboptimal solutions.

The size of the relevant space to be searched along with the distractive force of more accessible (more "obvious") solutions might contribute to the Non-Obviousness of the more optimal solution.

It would seem that both of these have relevance to Bennett's concept of "logical depth", as they both may drive up the necessary computational resources (or, the amount of brute force "tinkering") to realize the non-obvious solution -- where a flash of genius might render all that brute force tinkering unnecessary.  

In Shadows of the Mind, in which Roger Penrose argues for mathematical insight requiring something beyond computation, Penrose has a section on "Things that computers do well -- or badly":

Conscious understanding is a comparatively slow process, but it can cut down considerably the number of alternatives that need to be seriously consideredand thereby greatly increase the effective depth calculation.  

In other words, a flash of insight can cross large distances of "logical depth" a la Charles H. Bennett.  Insight is like a wormhole, a directed wormhole, through solution space.

macroscopically describable

If we repeat an experiment 2^k times, and define an event to be “simply describable” (macroscopically describable) if it can be described in m or fewer bits (so that there are 2m or fewer such events), and “extremely improbable” when it has probability 1/2n or less, then the probability that any extremely improbable, simply describable event will ever occur is less than (2^(k+m))/(2^n). Thus we just have to make sure to choose n to be much larger than k + m. If we flip a billion fair coins, any outcome we get can be said to be extremely improbable, but we only have cause for astonishment if something extremely improbable and simply describable happens, such as “all heads,” or “every third coin is tails,” or “only every third coin is tails.” Since there are 10^23 molecules in a mole of anything, for practical purposes anything that can be described without resorting to an atom-by-atom accounting (or coin-by-coin accounting, if there are enough coins) can be considered “macroscopically” describable.
   - Granville Sewell, Entropy, Ev and Open Systems, note 5

Coincidences

I've been thinking about how coincidences are evaluated in forensics.  How does one separate the pure coincidences from the can't-be-coincidences?

Supposedly, a detective does not believe in coincidence.

Let's say George has some mysterious charges on his bill.   Let's say that something comes to light that the exact digits of his credit card number appeared in a text from a Hans whom George doesn't know, but say George's daughter knows Hans.  There might be enough digits in that number to warrant thinking that it is not a coincidence -- that is, that those doesn't represent something else: a book-of-the-month club user registration, a hotel confirmation number, etc.  A long enough number is supposed to use up probabilitic resources.  After all, it should be the case that a random sequence of digits probably NOT be valid credit card number.  There could be a preponderance of evidence that Hans

To make this more interesting, let's say that these mysterious charges occurred shortly after George died under mysterious circumstances and his credit card went missing.  There is the proximity to George: two degrees of separation.  There is the time proximity: near the time of George's death.

To a detective, this forms a pattern, a specification.  Now the small probability is much larger than the Universal Plausibility Bound, but it is still compelling in many cases, once motive and alibi are dealt with.

The quality Dembski would call tractability, gets into the degrees of separation along with the specificational resources.  What are the chances that someone would have the credit card number and also that this person be only two degrees of separation from the credit card owner, close to the time the card went missing?   Too coincidental?   How to put some firm numbers behind that assessment?  I suspect that departments of justice deal with tractability all the time, and somehow it is not pseudoscience  despite the problem resisting really precise calculations in a lot of cases.

overview of the treatment of "information"

Excellent overview:
http://www.discovery.org/a/14251


I'm trying to remember which Dembski critique was claiming that genetic algorithms are a dark art.  And which was saying that genetic algorithms have a solid mathematical foundation in the work of Fisher.  Is the work of Fisher a red herring for the fact that genetic algorithms have to manipulated into provide specifc sorts of answers?

Update:  Ok... the first quote is actually Wein quoting Geoffrey Miller's "Techonological Evolution As Self-Fulfilling Prophecy" (intriguing title, n'est pas?):

The trick in genetic algorithms is to find schemes that do this mapping from a binary bit-string to an engineering design efficiently and elegantly, rather than by brute-force.... The genetic operators copy and modify the genotypes from one generation to the next.... Getting the right balance between mutation and selection is especially important.... Finally, the evolutionary parameters [such as population size and mutation rate] determine the general context for evolution and the quantitative details of how the genetic operators work.... Deciding the best values for these parameters in a given application remains a black art, driven more by blind intuition and communal tradition than by sound engineering principles.²⁴ 

which I quoted here on this blog.

The first quote, I must've been thinking of the "eandsdembski" paper.  Elsberry and Shallitt actually try to avoid the problematic claims about genetic algorithms and imply that we know that the amazing functional complexity we see in nature simply follows from the math:

Dembski asserts that \evolutionary algorithms" represent the mathematical underpinnings of Darwinian mechanisms of evolution [19, p. 180]. This claim is egregiously backward. A large body of scholarly work is completely ignored by Dembski in order to make this claim, including Ronald Fisher's 1930 book, The Genetical Theory of Natural Selection.[16]  It is evolutionary computation which takes its underpinnings from the robust mathematical formulations which were worked out in the literature of evolutionary biology.

They draw a distinction between genetic algorithms and artificial life.  They seem to be implying that none of the fine tuning done for genetic algorithms applies to evolutionary computing in artificial life, as it's general target of survival doesn't predispose it to solving particular, goal-directed problems (such as Schneider's ev program?).

Aside:
It occurs to me that the information going from the environment to the population in question should be represented as the logarithm of the decrease in probability of death before reproduction.  Given all the bits of information being absorbed by a population about property X, what would the signal to noise ratio be?