|
Author
|
Topic: The Origin of Biological Information
|
Doug Wedel
Member
Member # 1901
|
posted 08. March 2006 16:57
quote:
CSI is still a controversial concept
What sort of a mathematical "thing" is CSI? Is it a way of defining something? Is it a proof of something? I thought CSI was a way of defining information such that deductions could be made about whether some specific information was random, of unknown source, or the product of intelligent design.
quote:
Identifying the specification is not always easy because it is, in virtue of being partly constituted by an intelligent cause, not amenable to observation.
I foresee a problem here for my endeavor. I propose to use automata which are _unintelligent_ at least at the beginning of the evolutionary run. Does this mean somehow my proof will not be about CSI?
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 08. March 2006 17:18
Doug,
Your proof could be about CSI, so long as the information output (i.e. some string) matches an independent informational pattern that was not somehow jerryrigged into the automata. If you succeed, what you will have shown is that the CSI concept is bogus...not that CSI can be produced by blind automata.
What you would want to do is show that something like the series of prime numbers could be encoded in physical matter by pure mechanism, unaided by the programmer.
I should mention that I've been focusing mostly on specification in my comments. The information on both ends in the specification relation always has to meet certain rigorous standards of complexity.
[ 08. March 2006, 17:26: Message edited by: Micah Sparacio ]
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 08. March 2006 17:34
David,
I will also take a direct crack at this:
quote:
Yes information can be stored. We popularly think of information as "massless". However:
Information & Matter: Storing, copying, and transmiting information in the real world requires a physical change of the spatial configuration of matter, or of the transmission of matter or energy. e.g. magnetic domains or photons.
Information & Energy: Storing, copying or transmitting information further requires energy. e.g. there is a minimum amount or number of quanta of energy required to store a bit of information, depending on the storage format.
I don't want to wax too philosophical but I will describe how I view this abstraction that you're getting at. I view the two-legged stool of reality -- with matter and energy being the two stools -- as being one leg short of a stable stool ;-) The third leg of the stool that I sit on is called data.
MATTER - ENERGY - DATA
You say: what the heck is "data"? I defined it earlier in this thread when I thought people might let me get away with defining some terms. Data is a physical signal that's been transduced into a pattern in an information processing system.
In my "metaphysics", just as matter has a fixed and describable relationship with energy, so data and the information that is constructed from it has a fixed and describable relationship with mass and energy -- albeit one that is yet to be fully deciphered. Just as energy comes in many forms, so does data. Data, like energy, can be highly organized, or, also like energy, it can be highly unorganized. With both data and energy, if they are in a very unorganized state they are useless to living beings. It is only organized energy and organized data that can be tapped for life's purposes by organisms.
When organisms use stored energy they do what is called "work", energy applied over time. When organisms use information, they use it most crucially -- as Professor Dembski points out -- to make choices.
I in my own small way am trying to do the same thing that Professor Dembski is trying to do, which is to figure out the interrelationship of information to life, and how to correctly describe and symbolize and discuss the various elementary components of the system.
So yes, energy is required for data to exist. At a bare minimum, for example, one photon is required to trigger a rhodopsin molecule to transduce a signal that a brain can then process as information. Furthermore, information processing consumes energy, as you point out.
What interests me is that while energy and information are in some ways very much alike -- i.e. come in different forms, can be degraded from useful to useless, et cetera -- there are some ways in which the two domains are very very different.
The most important way in which the data processing domain differs from the energy processing domain is that in the later, the order is always forever on average declining, while in the former, the order is usually increasing. How can this possibly be? The answer is that they are different domains, and governed by different albeit compatible and interactive laws. It does take energy to alphabetize a list of names or to run an evolutionary algorithm. But while we may be able to measure exactly the amount of energy required to run a computer for the time necessary to compute an algorithm, what we lack are the tools and intellectual apparatus with which to measure and evaluate the order that is created within the computer.
We transduce energy into order. It sounds impossible, but it's not, because the kind of order we're talking about is informational order that exists with at least one foot outside of the energy and matter world.
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 08. March 2006 17:56
quote:
Your proof could be about CSI, so long as the information output (i.e. some string) matches an independent informational pattern that was not somehow jerryrigged into the automata. If you succeed, what you will have shown is that the CSI concept is bogus...not that CSI can be produced by blind automata.
What you would want to do is show that something like the series of prime numbers could be encoded in physical matter by pure mechanism, unaided by the programmer.
Here are some things I will demonstrate:
(1) a coevolutionary protocol
(2) a precisely-defined problem space
(3) automata that generate templates for searching dataspaces
(4) a way of using selection and pseudorandom number generation along with the templates to identify and evaluate relevant patterns in the data space, using this information to solve a 10-variable problem in nonlinear dynamics better than a human being can solve it.
What impressed me about this little toy model of mine is that it solved a problem that I personally had no idea how to solve, and could not even begin to guess at the answer.
But will this demonstration matter? Will it even be _relevant_ to the CSI issues?
IP: Logged
|
|
Irving
Member
Member # 535
|
posted 08. March 2006 21:06
Doug,
quote:
a way of using selection and pseudorandom number generation along with the templates to identify and evaluate relevant patterns in the data space, using this information to solve a 10-variable problem in nonlinear dynamics better than a human being can solve it.
Your problem will be with the use of selection. The template must be independent of the selection mechanism. Or in other words, the template cannot be used in any way by the selection mechanism. If the template is related to the 10-variable problem, and if selection is referencing the template against the pseudrandom generation...then the template isn't detached.
IP: Logged
|
|
David L. Hagen
Member
Member # 323
|
posted 08. March 2006 21:21
Thanks Doug for noting the importance of seeking order:
quote:
“virtually all of the greatest science that has been done in the last 100 years has been done with a fixed vision focused on uncovering mechanism and natural law”
Why? Look back to the origins of modern science in the century prior to that. 60% of the founders of modern science were devout Christians, only 4% skeptics.
Christianity Is a Science-Starter, Not a Science-Stopper, Nancy Pearcey
They believed in Intelligence behind the universe that established order or laws in nature. Consequently they searched for them. Thus their common assumption of looking for natural causes when searching for the laws of nature. However, we see the appearance of design in the universe and especially in biotic life. Consequently we make the assumption that intelligence may have been its cause. As with the previous assumption that there is laws of nature, now we seek to find principles, rules or laws of Intelligent Design to see if they explain information better than “necessity and chance”. Then let us compare the explanatory power of the different models.
I most heartily agree that we need to have all three of “data” (information), energy and matter. It’s the interaction of those and the origin of information that is the key debate.
quote:
I'm really only attempt to prove that automata can evolve and become more intelligent.
That may be an interesting exercise in modeling evolution. Micah explains one challenge:
quote:
Your proof could be about CSI, so long as the information output (i.e. some string) matches an independent informational pattern that was not somehow jerryrigged into the automata.
Doug
quote:
So I'm just building a proof that begins: ASSUME A TURING MACHINE
I would like to focus on the origin of your Turing Machine or automata itself, or equivalently the first cell. My difficulty with this is that you appear to be beginning by assuming a system that conventionally embodies Complex Specified Information originating from an intelligent designer.
If we are exploring the ORIGIN of biochemical information, then it appears from my engineering design perspective, that you are a priori assuming what we are debating, or begging the question. I.e., you assume a system that engineers commonly design, and thus expect that it embodies intelligent design. E.g. complex specified information.
My challenge to you is to explain HOW THAT TURING MACHINE was formed from natural causes of necessity and chance without the involvement of intelligence. I believe this is the Achilles heel of Darwinian evolution. Evolution requires the self reproducing cell to begin with, as prior to that there is no "natural selection."
My understanding of both computers and biotic cells indicate that all physical information processing machines require energy for every step in the process. It also requires energy to construct those processing machines. Detectives say "follow the money". As an engineer I say: "Follow the Information, energy and material flows."
The definition of the Universal Turing Machine appears to ignore energy. Some have examined actual energy requirements in computers and the thermodynamic limits of computing. E.g., Physical Limitations to Computations
Practically, our challenge appears to be to explain the origin of the first reproducing cell with its information, energy and material processing. While reversible computing in the quantum limit is an interesting challenge, how are we to explain the origin of the cell where it appears that discrete energy and matter is required for every bit of information processed.
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 08. March 2006 22:25
quote:
Your problem will be with the use of selection. The template must be independent of the selection mechanism. Or in other words, the template cannot be used in any way by the selection mechanism. If the template is related to the 10-variable problem, and if selection is referencing the template against the pseudrandom generation...then the template isn't detached.
Allow me describe to you more in a little more detail how templates are used in my model, and then perhaps you will be able tell me whether they pass muster.
In my model automata compete to solve a problem in "combinatorial algebra". The problem space is formally defined and represented on a tessellated plane (just a grid of equal sized cells or you could view it as a specialized type of cellular automata (CA). The cells in the automata can be in any one of three states. All possible legal transitions on the cellular automata are fully specified by formal transition rules.
The automata compete against one another in tournaments to see who can best solve the 10-variable optimization problem. The way they do this is by taking turns in a game format selecting from among currently available transitions in the CA.
Each automata, when its turn to move comes, is presented with an automically generated list of legal transitions for the current cellular automata's state. The automata then computes which transition best "fits" its current evolved values for the states that it finds in the CA (this is the formal decision model).
What kind of patterns and pattern specifications and evaluation are we talking about?
The automata perform their decision-making in two steps: (1) tabulation; and (2) evaluation.
The "tabulation" step identifies and counts patterns found in the current Cellular Automata.
What kind of pattern are we talking about? The "pattern gadget" that I use in my model is a geometrical curiosity called a "polyomino", after the word domino, which in fact is one of the polyominoes.
A polyomino is two or more cells attached along an edge -- in its essence, a very, very simple object. I use the polyomino as a kind of scaffolding for making masks to detect patterns. In particular, in my model, I use the three-celled linear triomino as a mask for searching the spaces of the CA.
There are ten possible combinations of self-opponent-empty in a three-cell triomino in my model. Each one is assigned a random value at the beginning of the test. The automata use these values to compute their choices. The winners mate with recombination using crossover as well as some mutation.
So the "masks" that I use to build my pattern-detection equipment are constructed out of pure geometry, i.e. polyominoes.
The patterns that are "masked" are standard combinations of states.
The values correlated with these patterns (combinations) are first set at random but evolve over time.
Can you tell enough yet to tell me whether I am in the right venue?
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 08. March 2006 23:03
quote:
If we are exploring the ORIGIN of biochemical information, then it appears from my engineering design perspective, that you are a priori assuming what we are debating, or begging the question. I.e., you assume a system that engineers commonly design, and thus expect that it embodies intelligent design. E.g. complex specified information.
My challenge to you is to explain HOW THAT TURING MACHINE was formed from natural causes of necessity and chance without the involvement of intelligence. I believe this is the Achilles heel of Darwinian evolution. Evolution requires the self reproducing cell to begin with, as prior to that there is no "natural selection."
I sympathize with your yearning for "ultimate knowledge", I don't think I can meet your challenge. I feel like you are asking a physicist to explain Planck's Constant. As an engineer you know the formal definition of a mechanism. I am seeking to describe the simplest possible mechanism that can evolve and get smarter. In other words, I am seeking a "designing mechanism". I expect this "designing mechanism" to be very clever and elegant, but not without content. It will contain CSI in its most condensed and potent form, I suppose. So I am just wondering whether what I am doing won't be viewed as a tautology as you suggest. I am strictly following up on John von Neumann's work in "The Theory of Self Reproducing Automata." I am seeking to extend his purely logical model of evolution to where his self-reproducing creatures start getting smarter and start speciating.
IP: Logged
|
|
David L. Hagen
Member
Member # 323
|
posted 09. March 2006 09:33
Doug
quote:
I expect this "designing mechanism" to be very clever and elegant, but not without content. It will contain CSI in its most condensed and potent form, I suppose.
Thank you Doug. This is the heart of the issue.
If you start with a functioning mechanism containing CSI, you can show output that contains CSI.
Challenges for:
Neo-Darwinian Evolution: The critical question for neo-Darwinian evolution is how do you obtain that first functioning mechanism from natural causes that DO NOT contain CSI?
How do you discover laws of "apparent design" from the a priori assumption of chance mutations and "natural selection" - especially when you have no selection before you start reproducing the first cell?
Intelligent Design: Conversely, the challenge for ID is to quantify that CSI, and demonstrate that it cannot have arisen within a closed system of "necessity" or stochastic "chance" natural causes. ergo, this implies an Intelligent Design as its origin.
The corresponding ID challenge is to "reverse engineer" life, and discover the design rules, principles and laws used in formulating the intelligent design. Yes this is a "challenge,"
but I think at least some of it is very doable, even if we do not understand the means of achieving them.
1) Your system embodies algorithms that appear to be crafted by an intelligent designer rather than by chance or necessity;
2) It runs the algorithms on a mechanism that embodies CSI;
3) The mechanism is supported by a power sytem embodying CSI;
4) The method uses a material processing system that embodies CSI.
I would thus consider each of these separately, and together to be examples of such design rules that do not arise from natural causes of necessity (law) or chance. e.g. within the universe from the four natural forces.
You can start taking Micah's challenge of showing that the algorithm starts with chance and necessity and INDEPENDENTLY arrives at CSI - WITHOUT embodying that CSI into your algorithm.
However, I think you will have the challenge of showing that the method by which you evaluate and reward the automas originated from and relies on chance/necessity and does not embody the CSI you claim to find.
If you can do that, you still have the challenge of demonstrating the origin of a practical Turing Machine system and AND associated power system, AND the associated material processing system, that each separately derive from chance and necessity without CSI. They further must simulataneously assemble and come together by necessity and/or chance without CSI.
I look forward to your laying out your system. However, from what you have shown so far, I do not see how you can avoid embodied CSI. If any of this appears to be a tautology, please point it out and I will try to correct/explain it.
[ 09. March 2006, 09:36: Message edited by: David L. Hagen ]
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 09. March 2006 10:53
"I expect this "designing mechanism" to be very clever and elegant, but not without content. It will contain CSI in its most condensed and potent form, I suppose."
quote:
Thank you Doug. This is the heart of the issue. If you start with a functioning mechanism containing CSI, you can show output that contains CSI.
I'm curious -- is Planck's constant considered CSI? How about e=mc^2?
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 09. March 2006 18:07
I'll post my problem, though I must confess to a growing sense of frustration. I began the thread talking about algorithms and information. I promised to present an algorithm that would generate information. But before I did so I wanted to make sure we had a working definition of terms. Thousands and thousands of words later, with extensive and (often) kind help from several of you in here, I feel no closer to understanding what you call CSI than I was at the start. Sometimes I think I've got it -- then it once again slips from my grasp. For example, while riding the MUNI today, reading a book by the mathematician John Barrow, I came to a chapter on logic. And I asked myself: does human logic -- the symbols and the rules -- contain CSI? I found I was _absolutely_ unable to answer this question. I did not know how you in here would answer it. I even speculated, based on the variety of posts I've gotten in here, that some of you would say, yes, human logic OBVIOUSLY contains CSI, while others would perhaps not agree. Of course if logic itself contains CSI then we have the problem pointed out by Pierre Boutroux
quote:
Logic is invincible because in order to combat logic it is necessary to use logic
Nevertheless, I will at least generally describe the problem since I said I would, and if anyone is interested in pursuing it further, I'd be glad to oblige. However, I will not take up your time developing any _proof_ unless, once the problem is stated, we can finally agree whether it _does_ or _does not_ contain CSI in the very statement of the problem or in the proposed method of solving it such that the information that will be generated by the algorithm can be said to have been present in "seed" form in the statement of the problem and description of the mechanism.
STATEMENT OF THE PROBLEM
The problem is from game theory. Stated most generaly, the question is this: How can we derive a formula for playing board games (tic-tac-toe, checkers, chess, go or any other formally describable board games of this general type ) such that from any point in the game tree one can calculate the best move?
Note that this question _assumes_ that such a theory exists somewhere in the noosphere. In fact there is a mathematical proof from early in the last century that such formulas do exist, at least in theory, for the type of board games I've mentioned. The problem is, of course, deriving them ;-)
So let's take the simplest of these games to start with: tic-tac-toe. How would one go about deriving an algebraic formula for calculating the best move from any node in the gamespace?
Notice that I am NOT seeking a restatement of the algorithmic rules we learned when we were five, such as, IF the opponent plays first in the corner then play next in the center. I am looking for a method of deriving a FORMULA, an algebraic formula, to _compute_ the best move. And I am looking for an approach to solving this problem that will be generalizable to checkers, chess and go, which, of course, we did not master at five.
Would anyone have _any_ idea how to go about deriving such a formula?
If anyone would care to respond, I will then take the discussion the next step, describe a hypothetical solution to the problem, introduce the 10 variables I mentioned, and then ask how to go about optimizing the variables to solve the problem.
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 09. March 2006 18:30
One thing I forgot to make explicit: brute recursion won't work. Yes, it works fine on tic-tac-toe, but for the other games the gamespace is far too large for "mere" recursion to work -- solutions such as: expand the game tree to its terminal nodes and select the path that leads to victory won't work in the real world. Would that it were that simple ;-)
IP: Logged
|
|
Irving
Member
Member # 535
|
posted 09. March 2006 20:59
quote:
I'm curious -- is Planck's constant considered CSI? How about e=mc^2?
quote:
I'll post my problem, though I must confess to a growing sense of frustration. I began the thread talking about algorithms and information. I promised to present an algorithm that would generate information. But before I did so I wanted to make sure we had a working definition of terms. Thousands and thousands of words later, with extensive and (often) kind help from several of you in here, I feel no closer to understanding what you call CSI than I was at the start.
Perhaps because you focus on CSI as an exclusively mathematical issue. Let me try it this way...
-----------------
There are currently two scientific rovers exploring Mars. They were built on Earth. They were built FOR Martian specifications. They were built on Earth. The optimal design specifications for the Martian environment is NOT the same as for exploration rovers on Earth. The Martian environment is independently specified from a Terran environment.
Un-guided processes using natural selection would use selection pressure based upon the Terran environment to design to the Terran environment. It takes a guided (i.e. Intelligence) process to build to a specification that is independent and detached from the local environment. i.e. Building a rover in the Terran environment FOR the Martian environment.
A mathematical specification especially as expressed in decimal notation, is independent of the physical artifacts, resultant from natural, physical properties. In other words, there's no "natural selection" reason why physical processes would build to a physical manifestation of the decimal expression of a mathematical specification. i.e. An actual, physical, tangible emitter, emitting the decimal expression of Pi to 500 decimal places...
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 09. March 2006 22:06
quote:
Un-guided processes using natural selection would use selection pressure based upon the Terran environment to design to the Terran environment. It takes a guided (i.e. Intelligen[t]) process to build to a specification that is independent and detached from the local environment. i.e. Building a rover in the Terran environment FOR the Martian environment.
Thanks for staying with me on this ;-) Your post helps, a remark Dembski makes in one of his papers (about Kolmagorov complexity) helps, David Hagen's posts and other posts have helped. I think I now have a new way of stating what I'm doing in terms of CSI ;-). What I want to demonstrate is a new form of "super CSI" -- the densest, most "ordered" CSI on earth, the "formula of formulas," CSI that will be able use information and logic to _construct_ information from "pieces and parts" and use that information to solve hard problems, including problems far beyond the ability of humans beings. Would that be interesting?
IP: Logged
|
|
Doug Wedel
Member
Member # 1901
|
posted 09. March 2006 22:10
SORRY. I screwed up one word in that last post:
It should have said: CSI that will be able use data and logic to _construct_ information from "pieces and parts" and use that information to solve hard problems, including problems far beyond the ability of humans beings. Would that be interesting?
IP: Logged
|
|
|
Printer-friendly view of this topic