|
Author
|
Topic: Does randomness ride on ignorance?
|
sanc olde
Member
Member # 1939
|
posted 07. April 2006 05:19
Among heavyweights (you guys), I won't be able to give the right replies. So for my first post I'd rather ask the right question (is it?).
Is it possible for an intelligent agent to come up with an algorithm which can produce randomness?
“To be perfectly random nothing should be left to chance.” I forgot who said this but I think I got it from reading in your archive. The bottom line of producing an algorithm which can produce randomness is, intelligence played a role. So that makes the randomness it will produce suspect.
The most common example for producing random outcomes is coin tossing. My problem with this setup is, to me, it can’t totally produce random outcome. The suppose random outcome rides on our ignorance of the initial conditions. Suppose we know the initial conditions, eg., force on the coin, angle of application, weight of the coin, etc. We often take the outcome to be random because we don’t know or choose not to know the initial conditions which can predict the outcome. If these factors were known the randomness of the outcome vanishes.
We can extrapolate the coin tossing to producing a computer program which can churn out random numbers. There will always be the possibility that somebody can “crack” the program such that the outcome can be had in advance.
So again from the title of this thread, does randomness rides on ignorance?
IP: Logged
|
|
Albert Voie
Member
Member # 1941
|
posted 07. April 2006 16:48
It depends upon how one defines randomness. A random string of bits is normally defined as a string that can't be compressed in a lesser program. It follows that an algorithm that puts out a random string of bits must be greater in bits than the string itself. This is randomness by design.
There are other ways to produce randomness in mathematics. Since whether a Turing machine halts or not (finds a solution or goes into an infinite loop) by any input (e.g. a diophantine equation created by coin tossing) is undecidable, the result can be looked upon as random. The halting probability of a Turing machine (Chaitin's omega) is looked upon as worse than pi - it is uncomputable (see Greg Chaitin for more). I think that this is also randomness by design, but here the design is the abstract Turing machine (diophantine equation) and the self referring syntactic system.
I also think I remember that Stephen Wolfram has claimed that one of the cellular automata (I think #31), puts out real complexity. I am not sure about this claim.
Best
Albert
[ 07. April 2006, 16:51: Message edited by: Albert Voie ]
IP: Logged
|
|
Zachriel
Member
Member # 1793
|
posted 08. April 2006 23:17
sanc olde: "Is it possible for an intelligent agent to come up with an algorithm which can produce randomness?"
No discrete algorithm running within finite memory can create *true* randomness. However, there are a variety of quite good pseudo-random number generators that will work for most applications and are virtually impossible to distinguish from true randomness without pre-knowledge of the algorithm and seed.
To design a *true* random number generator it is only necessary to harness random aspects of physical processes, usually from sampling a source of entropy. You can also use quantum decay.
Random.org offers true random numbers to anyone on the internet.
IP: Logged
|
|
Bruce Fast
Member
Member # 924
|
posted 10. April 2006 18:56
Sanc Olde, I am dissappointed at how little air your topic is producing. You ask a good question.
As I understand it, Dembski recently suggested that randomness itself must have been designed. I so far have found his case unconvincing.
Let me suggest a few things about randomness:
1 - True randomness is not all that common in nature. What we see in nature, however, is a lot of normal distribution (bell) curves. This type of data does not allow for the precision analysis of the next entry in the sequence. For instance, you can't say exactly how many photons of sunlight will hit your eye in the next second. Because the exact value of each unit of data is really quite inimportant, this type of data is not CSI. The amount of information in a bell curve "randomly" populated is really quite small, being limited to the shape of the curve itself.
2 - When the RM in RM+NS is discussed, it is sometime defined as Random with reguards to function. The reason such a statement is made is that the "random" bit is often not truly random, but is that very common normal distribution. In any case, independant of the source of a mutation, if the mutation generator does not have a stated purpose involving function, then the effect is functionally the same as if the mutation did meet some standard of 'truly random'.
3 - Though it is correct that a descreet algorithm will never produce true randomness, it can very easily produce "random with reguards to function." Pseud-random generators, therefore, are fully adequate to simulate the RM of RM+NS simulations.
4 - If my understanding is correct, the famous "Shannon information algorithm" will detect information in a normal distribution because a normal distribution is compressable (the Huffmann compression algorithm will achieve compression on any normal distribution.) Please, you guys, correct me if I am wrong.
IP: Logged
|
|
Albert Voie
Member
Member # 1941
|
posted 11. April 2006 04:05
Bruce and Sanc olde,
Maybe we should decide on what to discuss in this thread. Is it whether true randomness exists or not as phenomena in nature ,or is it whether randomness has explanatory power?
I have a brief discussion on randomness in the following paper. http://home.online.no/~albvoie/index.cfm
True randomness would be something that can't be compressed into the laws of nature. Life certainly has such properties. It appears random, but so does a computer program. However, no rational or empirical justification exists for attributing linear, digital, encrypted, genetic recipes to stochastic ensembles in any amount of time.
IP: Logged
|
|
Bruce Fast
Member
Member # 924
|
posted 11. April 2006 12:15
Dr. Voie, very interesting paper. I have not fully read it, only purused it, but I will read it more fully. Have you signed the "descenting scientists" document?
I think that something Dembski said has inspired sank olde to initiate this thread. Dembski has suggested that true randomness in itself is of necessity the product of design. His argument seems to be, we can write a computer program that produces a good approximation of randomness. Therefore the guy who wrote the incredible computer of the original living organism could also write a true random generator. Therefore he did, and that's where randomness comes from. Its an idea worth hashing around in a philosophy class, but certainly not one I'd stake my life on.
My personal schtick is that data that conforms to a normal distribution is, by definition, compressible. It therefore should be seen by Shannon's equation as information. Yet the true information content is miniscule compared to the information encased systems of DNA and of computer programs. It is an asside, but might merit its own thread.
Lastly, I may just be a commoner but I am having trouble getting my mind around "However, no rational or empirical justification exists for attributing linear, digital, encrypted, genetic recipes to stochastic ensembles in any amount of time." Are you saying that you don't believe that random stuff (stochastic ensembles) is caused by a great algorithm in the sky?
IP: Logged
|
|
Albert Voie
Member
Member # 1941
|
posted 11. April 2006 14:45
quote:
Dr. Voie, very interesting paper. I have not fully read it, only purused it, but I will read it more fully. Have you signed the "descenting scientists" document?
Yes, I did not resist.
quote:
I think that something Dembski said has inspired sank olde to initiate this thread. Dembski has suggested that true randomness in itself is of necessity the product of design. His argument seems to be, we can write a computer program that produces a good approximation of randomness. Therefore the guy who wrote the incredible computer of the original living organism could also write a true random generator. Therefore he did, and that's where randomness comes from. Its an idea worth hashing around in a philosophy class, but certainly not one I'd stake my life on.
I see! Scientists like Anton Zeilinger suggests that the objective randomness of the individual quantum event is a necessity of a description of the world in view of the significant influence the observer in quantum mechanics has. Maybe you could comment on this short essay by him in Nature:
http://www.quantum.univie.ac.at/publications/pdffiles/The%20message%20of%20the%20quantum.pdf
From the essay:
The randomness of the individual event is then a direct consequence of the fact that not enough information is available to pre-define the outcomes of all possible measurements. The same holds for complementarity, implying that the information available only suffices to define the outcomes of one of a number of mutually complementary measurements. Finally, entanglement is the observation that the finite information available to characterize two (or more) systems can either be used to define the properties of the individual systems, as in classical physics, or to define the results of joint observations of both or all systems together. So the experimentalist, by choosing the apparatus, can define which quality of a number of possibilities will become reality in the measurement. But the individual measurement result remains objectively random because of the finiteness of information. I suggest that this randomness of the individual event is the strongest indication we have of a reality ‘out there’ existing independently of us.
Could you also tell us what you mean by "true" randomness and tell us how we can observe it?
quote:
My personal schtick is that data that conforms to a normal distribution is, by definition, compressible. It therefore should be seen by Shannon's equation as information. Yet the true information content is miniscule compared to the information encased systems of DNA and of computer programs. It is an asside, but might merit its own thread.
I agree
quote:
Lastly, I may just be a commoner but I am having trouble getting my mind around "However, no rational or empirical justification exists for attributing linear, digital, encrypted, genetic recipes to stochastic ensembles in any amount of time." Are you saying that you don't believe that random stuff (stochastic ensembles) is caused by a great algorithm in the sky?
No, this is not my point.
My point is that pseudo randomness or randomness does not explain things like the origin of life, molecular machines or the mind. Neither does the laws of nature. This is what I discuss in my paper.
From an observer's point of view the origin of life would yet appear totally random, since the event can't be explained by the laws of nature - the laws of nature are in this respect incomplete.
IP: Logged
|
|
Bruce Fast
Member
Member # 924
|
posted 11. April 2006 16:21
Albert Voie: "No, this is not my point..." Oh, I see, and wholeheartedly agree. I am a software developer. I do not believe that randomness plus selection could write the code that I write, "not in a million years."
You seem quite clear about the origin of life, what is your view of the rest of life. Is it reasonable, as far as you know, to believe that RM+NS (in all of its glory including HGT etc.) can realistically account for the variety and complexity of life as we see it, or is the current explanation just fundimentally inadequate.
The Shannon thing does call for its own thread so it will get it.
IP: Logged
|
|
Albert Voie
Member
Member # 1941
|
posted 12. April 2006 13:51
quote:
You seem quite clear about the origin of life, what is your view of the rest of life. Is it reasonable, as far as you know, to believe that RM+NS (in all of its glory including HGT etc.) can realistically account for the variety and complexity of life as we see it, or is the current explanation just fundimentally inadequate.
I think that selection is important - natural selection works at the phenotypic level, therefore natural selection can't account for the genetic programs running inside the organism. A genetic algorithm can increase the complexity of a functional object, but the distance between the selection points is always low (a few bits). Life contains solutions to problems that have the size of hundred or thousand bits . Thus, I believe that selection must occur at the genetic level.
[ 12. April 2006, 18:19: Message edited by: Albert Voie ]
IP: Logged
|
|
sanc olde
Member
Member # 1939
|
posted 17. April 2006 01:40
quote:
by Zachriel
To design a *true* random number generator it is only necessary to harness random aspects of physical processes, usually from sampling a source of entropy. You can also use quantum decay.
It's quite interesting that man can't produce true randomness without the aid of the random aspect of the physical world. By choosing not to know (ex. employing the unpredictability of quantum events) he has pegged himself in a situation where he can't predict the outcome, but also he can't reproduce it.
So if a truly random event has unknowable initial conditions which produced it, it would be safe to say that no algorithm can be had to reproduce it.
IP: Logged
|
|
GSchultz
Member
Member # 1953
|
posted 18. April 2006 00:00
A definition of randomness I have found useful is, "information that is completely uncorrelated--both within itself, and with anything outside of it." This seems like a good fundamental starting point for me because it quickly leads to ruling out as random the things we know are not (repeated information, compressible information, etc.)
Thus I think that randomness must ride on ignorance. If you don't have total ignorance, then the information in question is correlated with something outside of that information, namely your (meta-)knowledge of it.
quote:
I also think I remember that Stephen Wolfram has claimed that one of the cellular automata (I think #31), puts out real complexity.
I believe he stated that the only way to know the nth state of the automaton was to run it from the start to the nth state; i.e., there is no way to simple compute the nth state by a shorter route. This kind of information would not qualify by the above definition, as correlations would exist within the data, and it is compressible since it came from a small deterministic algorithm with a known starting state.
Randomness defined this way is also relative, as we can intentionally remain ignorant of possible correlations. A coin toss is "good enough" for deciding who gets to start a game because everyone's knowledge of the initial conditions, etc., is so small, and the coin has no known correlations with the players, or the rules of the game.
Lastly, proving a particular sequence of bits/digits totally random is impossible because the search necessary to prove no correlation existed would be impossible.
IP: Logged
|
|
sanc olde
Member
Member # 1939
|
posted 20. April 2006 22:26
quote:
by GSchultz
A definition of randomness I have found useful is, "information that is completely uncorrelated--both within itself, and with anything outside of it."
The mechanistic worldview before the 20th century was the universe was winded like a clock at the beginning and that everything can be accounted for x seconds after that. This was shattered after quantumn mechanics came along. It made the game more interesting by introducing uncertainty. But with a catch, certainty about the magnitude of the uncertainty.
I think randomness rides on this uncertainty (which has affinity with ignorance). If this uncertainty is certain (it's computed all the time in QM), can randomness be also formulated so it becomes measurable?
IP: Logged
|
|
Bruce Fast
Member
Member # 924
|
posted 20. April 2006 23:40
sanc olde, your paralell between QM and genetic randomness is intriguing, and seems valid to me. I think, however, that there is a standard measure of the "amount" of the randomness. The molecular clock hypothesis provides a measure of the number of point mutations which occur in junk DNA. The proposed number seems to be 1% of nucleotides mutated per million years in mammals. Is that what you are looking for?
IP: Logged
|
|
IIIKrazyKiDDIII
Member
Member # 2151
|
posted 19. August 2006 16:51
Im not in all of the details but it feels like all of you are trying to calculate infinity. Vainly calcualting steps 1 and 2 and getting closer but will never reach the asymtote.
Isnt that what true randomness is?
IP: Logged
|
|
sanc olde
Member
Member # 1939
|
posted 08. December 2006 06:13
quote:
by Bruce Fast
sanc olde, your paralell between QM and genetic randomness is intriguing, and seems valid to me. I think, however, that there is a standard measure of the "amount" of the randomness. The molecular clock hypothesis provides a measure of the number of point mutations which occur in junk DNA. The proposed number seems to be 1% of nucleotides mutated per million years in mammals. Is that what you are looking for?
Am uncomfortable with analyzing how Shakespeare honed his writing skills by comparing his early works with the works he did when he was at his peak by counting how many times he used a particular word or little of it. Same thing with DNA. Until we know how to interprete the data in the DNA, I'll be apprenssive in analyzing it by the occurence or absence of some letters in it.
Given two chopsticks we can have random configuration of them. A cross, L, T, and others. The outcome will be consistent with the nature of our two sticks. For example, we don't expect to encounter a letter S by ramdomly arranging the two sticks even if we do it in a billion years. I am of the opinion that this is the same with DNA. It has wide leeway for mutations, but all those mutations would be consistent with the permissible configurations the DNA allows. It's not everything is possible, but just many things are possible.
IP: Logged
|
|
|
Printer-friendly view of this topic