Register | Sign In


Understanding through Discussion


EvC Forum active members: 65 (9162 total)
4 online now:
Newest Member: popoi
Post Volume: Total: 915,817 Year: 3,074/9,624 Month: 919/1,588 Week: 102/223 Day: 0/13 Hour: 0/0


Thread  Details

Email This Thread
Newer Topic | Older Topic
  
Author Topic:   The Minkowski's challenge
Admin
Director
Posts: 12998
From: EvC Forum
Joined: 06-14-2002
Member Rating: 2.3


Message 61 of 120 (355289)
10-08-2006 8:20 PM
Reply to: Message 60 by Albert Godwin
10-08-2006 3:03 PM


Albert Godwin writes:
I see Admin's last post unexplainably offensive and personal. He is becoming irrationally offensive although I NEVER had the intention of getting myself into battles here.
1-day suspension.

--Percy
EvC Forum Director

This message is a reply to:
 Message 60 by Albert Godwin, posted 10-08-2006 3:03 PM Albert Godwin has not replied

  
Barbarian
Inactive Member


Message 62 of 120 (355321)
10-09-2006 2:54 AM
Reply to: Message 55 by Albert Godwin
10-08-2006 8:26 AM


Hi, Albert -
clarification:
quote:
> There is a question about whether the decryptor code must be correctly executable in all generations. I would, just to keep it realistic, propose that it does not need to be.
IF it is not correctly executable then the file will not work. Sorry but all the offspring from the first file to the goal must be executable.
The 'files' would be 'executable' all right, even more, they would be required to produce an exact copy of themselves as a result of being executed (they will also be allowed to produce multiple copies, totally unexecutable code etc. - as long as they also write at least one exact copy of themselves). What I was referring to here was the requirement that the decryption logic has to work from its inception in all intermediate generations, not just in the final one(s). My reading would be that the intermediate generations don't matter, and only the final result should have the mutually dependent and hence irreducibly complex structure of decryption + encrypted code.

This message is a reply to:
 Message 55 by Albert Godwin, posted 10-08-2006 8:26 AM Albert Godwin has not replied

  
Barbarian
Inactive Member


Message 63 of 120 (355322)
10-09-2006 3:50 AM
Reply to: Message 56 by Albert Godwin
10-08-2006 8:36 AM


quote:
> We are approaching this stuff from very different positions. The challenge is not yet entirely clear, e.g. see the discussion about encryption above. Depending on the clarifications, it could turn out to ask for the impossible or for the impractical.
So far, do you accept the challenge or think that it is fair enough?
See my previous post for a clarification request; apart from that, pretty much the only thing I still don't have a handle on is the required volume of the simulation. I have a few possible selector sequences which would lead to the irreducible complex setup we talk about here, but the system is complex enough to surprise me even before it is ready to run. Apart from the uncertainty regarding the necessary time, another nice problem is that in such a reverse-engineered situation, the winner of every 'geological period' must be the one I intended it to be, but it can always happen that the artificial world discovers a better solution to the constraints of the selector, and this better adapted program completely displaces the one I tried to evolve to serve as an intermediate step.
quote:
But to me, it is an excellent method to practically show the programmers how evolution is very unlikely to happen.
Many other posters have already pointed out to you that this challenge has little to do with biological evolution, in particular the feasibility of the two processes are all but unrelated. I get it that you think irreducibly complex structures cannot evolve at all, but this particular sort of structure might be unevolvable for entirely different reasons than unfeasibility of biological evolution. It could happen that the only selector for which encryption + decryption is the best solution is one similar to the WEASEL selector (measuring the distance of certain dead code from a prepared decryption + encrypted code structure - this would be cheating and of course I won't use it), and all other selectors understandable by humans admit better solutions than the encryption+decryption one. It could happen that the scaffolding needed to evolve this structure is so huge that the program won't finish evolving and then dismantling it in our lifetime. Etc. etc.
This is why I try to handle this as a programming challenge. The most severe constraint I see right now is that it has to happen within reasonable time. I do not intend to, nor can easily enlist anything beyond a machine with a 2400 MHz Intel processor, and the ratio between what can be done on such a machine and what biological evolution had at hand to work with simply defies imagination.

This message is a reply to:
 Message 56 by Albert Godwin, posted 10-08-2006 8:36 AM Albert Godwin has not replied

  
Admin
Director
Posts: 12998
From: EvC Forum
Joined: 06-14-2002
Member Rating: 2.3


Message 64 of 120 (355379)
10-09-2006 12:27 PM
Reply to: Message 60 by Albert Godwin
10-08-2006 3:03 PM


Hi Albert,
You seem terminally clueless, but I'll take one shot at explaining why you find yourself in trouble here.
Imagine that you've somehow got it into your head that automobiles are impossible. You walk into a bar and announce that automobiles are impossible, and challenge the patrons to prove otherwise. Some explain that there are cars out in the parking lot and that if you'd just go take a look you'd see you're wrong, but you express no interest in doing this. Finally, someone agrees to build an automobile just to show you you're wrong, but you specify some requirements: the automobile must have 1000 horsepower and weigh less than 500 pounds.
It is explained to you that if it turns out that an automobile meeting those requirements isn't really possible that it doesn't mean that automobiles are impossible. It's actually more a design challenge than a "proof of automobile" challenge. You apparently understand none of this and again declare that automobiles are impossible, and you taunt the patrons about how only one person dared accept your challenge. Just to save your hide, the bartender tosses you out.
You're doing the analogous thing here with evolution. The evolution of encryption, whether or not it's possible, is irrelevant to the question of whether evolution is possible, just as whether a 1000 horsepower car weighing less than 500 pounds is possible is irrelevant to whether automobiles are possible. You're refusal to learn about artificial life and genetic algorithms is analogous to refusing to go out to the parking lot to see all the automobiles. You're making silly assertions about something you know very little about, and you seem determined to maintain your state of ignorance.
You can continue the dialogue with Barbarian about the encryption evolution program, but further nonsense will earn further suspensions. You might want to give the Forum Guidelines a careful look, particularly rule 4 and rule 10. My advice to you remains to learn something about the topic. When you begin becoming embarrassed by some of the naive things you've said in this thread then you'll know you're beginning to learn something.

--Percy
EvC Forum Director

This message is a reply to:
 Message 60 by Albert Godwin, posted 10-08-2006 3:03 PM Albert Godwin has not replied

  
Albert Godwin
Inactive Member


Message 65 of 120 (355527)
10-10-2006 4:21 AM


Dear Barbarian,
>My reading would be that the intermediate generations don't matter, and only the final result should have the mutually dependent and hence irreducibly complex structure of decryption + encrypted code.
Yes, sure. I am not asking you to evolve encryption from the first generation. As long as all intermediates are executable, it's all right with me. Sorry for misunderstanding that.
>See my previous post for a clarification request; apart from that, pretty much the only thing I still don't have a handle on is the required volume of the simulation. I have a few possible selector sequences which would lead to the irreducible complex setup we talk about here, but the system is complex enough to surprise me even before it is ready to run. Apart from the uncertainty regarding the necessary time, another nice problem is that in such a reverse-engineered situation, the winner of every 'geological period' must be the one I intended it to be, but it can always happen that the artificial world discovers a better solution to the constraints of the selector, and this better adapted program completely displaces the one I tried to evolve to serve as an intermediate step.
well, if there is any thing that you must add to solve that, just tell me and I will approve it.
>It could happen that the only selector for which encryption + decryption is the best solution is one similar to the WEASEL selector (measuring the distance of certain dead code from a prepared decryption + encrypted code structure - this would be cheating and of course I won't use it),
Yes, I object to using it, not because it is cheating, but because many intermediates will not be executable in this way.
> This is why I try to handle this as a programming challenge. The most severe constraint I see right now is that it has to happen within reasonable time. I do not intend to, nor can easily enlist anything beyond a machine with a 2400 MHz Intel processor
Well, definately you know better than me, but I don't know. I think that's too much to evolve encryption... It doesn't seem that complex after all...
Dear Admin,
I will fully respond to you after my discussion with Barbarian is over.
Cya

Replies to this message:
 Message 66 by Barbarian, posted 10-10-2006 8:09 AM Albert Godwin has replied
 Message 67 by Admin, posted 10-10-2006 9:09 AM Albert Godwin has not replied

  
Barbarian
Inactive Member


Message 66 of 120 (355540)
10-10-2006 8:09 AM
Reply to: Message 65 by Albert Godwin
10-10-2006 4:21 AM


Hi, Albert -
quote:
>It could happen that the only selector for which encryption + decryption is the best solution is one similar to the WEASEL selector (measuring the distance of certain dead code from a prepared decryption + encrypted code structure - this would be cheating and of course I won't use it),
Yes, I object to using it, not because it is cheating, but because many intermediates will not be executable in this way.
Since I do not intend to use such a trick anyway, it might be superfluous to point out that you object to the use of a WEASEL selector for entirely the wrong reason. Executable programs may contain garbage code if said code is never executed or is executed only after an exact copy of the program was written out and declared final (i.e. the 'file' was 'closed'). Such garbage code could be evolved to whatever I want it to be using a WEASEL selector. The reason it should not be used, besides smelling like cheap cheating, is simply that results obtained using such a selector usually fail to impress young-Earth creationists, who can claim that the end result was too directly specified by this method. The original WEASEL program came under exactly this kind of attack.
Perhaps this is the moment when I have to bring up the fact that selectors are situated on a continuum, with WEASEL at one extreme and no selector whatsoever at the other one. Given this fact, it is easy to foresee the opportunity for bitter dispute about a successful selector being "too specific". So far none of the selectors making up the sequences I considered were even close to being too specific, but I still wonder how would you draw the line between WEASEL-like and acceptably general selectors.
All else considered, I think we have touched on all the necessary basics here. I will summarize my understanding of the discussed details probably before Sunday this week and try to force it into the resource envelope of a usual hobby project. It could take a long time even if it proves to be doable on small machines, potentially lasting for months, since I can only work on it on weekends, although I expect the whole environment to be ready to run by the end of this week.
Edited by Barbarian, : English grammar ...

This message is a reply to:
 Message 65 by Albert Godwin, posted 10-10-2006 4:21 AM Albert Godwin has replied

Replies to this message:
 Message 68 by Albert Godwin, posted 10-11-2006 1:50 PM Barbarian has replied

  
Admin
Director
Posts: 12998
From: EvC Forum
Joined: 06-14-2002
Member Rating: 2.3


Message 67 of 120 (355554)
10-10-2006 9:09 AM
Reply to: Message 65 by Albert Godwin
10-10-2006 4:21 AM


Albert Godwin writes:
Dear Admin,
I will fully respond to you after my discussion with Barbarian is over.
No response is desired or necessary, though you can certainly respond in the General discussion of moderation procedures - Part thread if you feel the need. You've replied helpfully and constructively to Barbarian, and as long as your contributions continue in the same vein I see no need for further interaction.
Edited by Admin, : Grammar.

--Percy
EvC Forum Director

This message is a reply to:
 Message 65 by Albert Godwin, posted 10-10-2006 4:21 AM Albert Godwin has not replied

  
Albert Godwin
Inactive Member


Message 68 of 120 (355911)
10-11-2006 1:50 PM
Reply to: Message 66 by Barbarian
10-10-2006 8:09 AM


>All else considered, I think we have touched on all the necessary basics here. I will summarize my understanding of the discussed details probably before Sunday this week and try to force it into the resource envelope of a usual hobby project. It could take a long time even if it proves to be doable on small machines, potentially lasting for months, since I can only work on it on weekends, although I expect the whole environment to be ready to run by the end of this week.
Dear Barbarian,
Take your time, I don't want to feel that I am distracting you from your work. Post the summary whenever you are free. Even if you can't make it, at least post how much you have done, this will help anybody later who would like to take the challenge to know more about its details.
Perhaps you should consider publishing it with your real name as well... its legal, why not?
One more time take your time and good luck (How can i put a flower in this forum ?)
Dear Admin,
I am afraid that you misunderstood my last post. I just wanted to tell you ( all the people here) about myself, how i came to know this thing and why I am that excited about it. Anyway, if that is not allowed here, it's ok.
Cya

This message is a reply to:
 Message 66 by Barbarian, posted 10-10-2006 8:09 AM Barbarian has replied

Replies to this message:
 Message 69 by Barbarian, posted 10-15-2006 4:03 PM Albert Godwin has not replied

  
Barbarian
Inactive Member


Message 69 of 120 (356718)
10-15-2006 4:03 PM
Reply to: Message 68 by Albert Godwin
10-11-2006 1:50 PM


As promised earlier, I have summarized the rules I try to work within. I am doing this for my own amusement - I am done with the tedious work of testing the assembler, the virtual machine, the debugger and the mutation mechanism - and have no idea when will I finish. At the current rate, sometime around the end of the year ...
Markings such as "(-> number)" mean that the notion being used is elaborated under the paragraph marked "(number)". This was needed because I could not be bothered to give the stuff a pedagogical structure, so references run both forward and backward.
So, the rules:
(1) the task is to evolve (-> 2), with appropriate choice of selectors (-> 3), a program (-> 4) featuring viral behavior (-> 11) and encryption (-> 5), starting from one which shows viral behavior only (-> 6).
(2) evolution proceeds over a finite population of possibly different programs (-> 4), by means of the selector (-> 3) granting them a chance to procreate (-> 7) proportional to their fitness value (-> 10). The task is not simply to evolve such a specimen once but to provide a way to repeat the experiment as many times as one wants.
(3) the selector functionality is based upon a fitness function (-> 10), with the fitness values directly influencing the chances of leaving offspring. Multiple selectors can be used either sequentially - i.e. evolving a trait, then switching to the next selector - or in parallel, simulating co-existing biological niches. Most likely only the sequential variant will be used.
(4) programs are codes in a low level language defined by me and runnable in a virtual machine. A program is a list of segments, while a segment is a list of instructions. The language will be publicly specified only together with the solution, because it might still change in order to make evolution fast enough for my ancient PC. (It is a stack-based language much like Forth, but without Forth's structured approach - apart from not having registers, it is as non-stuctured as any machine code can be.)
(5) presence of encryption is defined as the observed execution of all special instructions during program runs, together with their absence from the program code.
(6) the lack of encryption in the start-up code is established by simply inspecting its assembly source.
(7) procreation proceeds under the supervision of the framework, which selects a member of the population with a probability linked to its fitness value (-> 10). The selected individual is copied verbatim and then, depending on a random choice, zero or more mutations (-> 9) are applied to it (most likely this will be simplified to zero or one). Finally, the new individual is rated for fitness and admitted into the population, unless it has a fitness rating of zero (stillborn).
(8) special instructions are instructions which from the POV of the virtual machine do nothing. As such, they act as NOPs and cannot alter the state of the machine in any significant way (except the standard update of the current instruction pointer). They do, however, interact with the environment, and for the sake of this particular simulation I plan to use one of them for writing out a program segment (-> 4), another one for declaring the written-out list of segments closed (closing the file), and some number of them to do nothing except register the fact of their execution with the environment. I do not know beforehand how many of them will be useable within a reasonable time frame - originally I proposed a total of 12 such instructions, but it could be less if I want to ever finish.
(9) mutations are applied by a mutator mechanism, which has some understanding of the program structure (-> 4). This means that in addition to individual instructions being altered, deleted or inserted, whole segments can be fused, split, duplicated or dropped. The probabilities of these mutation events can be finely adjusted, but they stay unchanged during the whole experiment.
(10) a fitness function attaches some numerical value to a particular program. The value shall be a non-negative integer, with a finite upper bound. In order to exclude WEASEL-like selectors, the fitness function is limited to seeing only certain aspects of the program, such as the number and individual length of program segments and knowing about the presence of special instructions (-> 8) in the code. In addition, the fitness function will be able to inspect the result of effects of special instructions (-> 8) upon the environment.
(11) viral behavior is defined as the capability of the program to write out an exact copy of itself. The program has a way to delimit units (close files) it writes out. It does not matter if units before or after the exact copy are additional exact copies or perhaps just junk. In addition, if the copy is written out, it does not matter how the program terminates - by error, by exceeding the time limit or by normal termination.

This message is a reply to:
 Message 68 by Albert Godwin, posted 10-11-2006 1:50 PM Albert Godwin has not replied

Replies to this message:
 Message 70 by Percy, posted 10-16-2006 9:43 AM Barbarian has replied

  
Percy
Member
Posts: 22392
From: New Hampshire
Joined: 12-23-2000
Member Rating: 5.3


Message 70 of 120 (356835)
10-16-2006 9:43 AM
Reply to: Message 69 by Barbarian
10-15-2006 4:03 PM


Barbarian writes:
As promised earlier, I have summarized the rules I try to work within. I am doing this for my own amusement...
That's good, because...
... - I am done with the tedious work of testing the assembler, the virtual machine, the debugger and the mutation mechanism - and have no idea when will I finish. At the current rate, sometime around the end of the year ...
That's a lot of effort. Albert has already demonstrated a lack of interest in discussion or in investigating the relevant fields. His sole source of information is a fictional novel where the accompanying software example is naive in the extreme (manual selection, for one very significant example). I assume that Albert's attention span on this topic will be similarly shallow, and I doubt that he will be anywhere to be found by year's end.
Even if Albert sticks around, he's incapable of understanding the results anyway, so even if you successfully evolve encryption he will cast unsubstantiated and nonsensical criticisms that's he's unable to defend, accompanied by an inability to comprehend that he lacks the necessary background and knowledge to assess or discuss the results, but he'll keep posting anyway.
Normally I'm not so personal and critical of another member, but I have limited myself to those qualities that Albert has already amply demonstrated and that are substantively affecting this thread, for example, his reluctance to engage in discussion or to inform himself of the current state of the art. You're making a substantial investment of your personal time, and so I think you should be very certain that you are doing this for it's own rewards, because proving to Albert that he's wrong won't be one of them.
--Percy

This message is a reply to:
 Message 69 by Barbarian, posted 10-15-2006 4:03 PM Barbarian has replied

Replies to this message:
 Message 71 by Barbarian, posted 10-17-2006 4:54 AM Percy has not replied

  
Barbarian
Inactive Member


Message 71 of 120 (356992)
10-17-2006 4:54 AM
Reply to: Message 70 by Percy
10-16-2006 9:43 AM


I hear you, Percy. I started this exchange with the hope of making Albert re-think his position while we argue the rules, then set the somewhat lower aim - still not entirely dropped - that if this challenge ever makes it to the Index (of creationist claims), it would be nice to be able to post the rebuttal at the same time, and finally set it as a programming challenge. So no, my primary goal is not to convince Albert, but I do not discard the possibility of such an outcome either (mostly because the possibility is inconsequential to me at this point).
I do expect some arguments about the ways my proposed virtual machine programs differ from usual machine code. I included some behaviors usually not found in real-world machine code in order to make the necessary evolutionary timeframe shorter, but I always tried to do it in a way mimicking some part of biological evolution. E.g. insertions/deletions of instructions usually do not dislocate jump targets, as jumps refer to labels attached to individual instructions, not to whatever instruction is located at a certain address (in order to localize the effect of mutations, as they are localized - in a different way - in the genome, but not localized in the usual machine code). Even so, the analogy between my toy world and biology is, for lack of a better world, fractured. E.g. instructions could be claimed to correspond to codons (I even have a STOP and a START instruction, suspending between them the execution - but not the scanning - of instructions), and perhaps segments can be thought as corresponding to chromosomes, but there is no direct correspondence to genes (perhaps the execution history between two writings of segments could have a claim to this), and even less correspondence with expression of genes (I use a mix of "structure of program", "presence of special instructions" and "effect of executing the special instructions" to correspond to the real-world test of the result of gene expressions). These programs are more like ribozymes catalyzing their own synthesis, but then why do I include a mechanism like the chromosome? So, there is ample opportunity for arguing away their relevance. This whole experiment is very far removed from biology.

This message is a reply to:
 Message 70 by Percy, posted 10-16-2006 9:43 AM Percy has not replied

Replies to this message:
 Message 72 by Albert Godwin, posted 10-17-2006 5:16 AM Barbarian has replied

  
Albert Godwin
Inactive Member


Message 72 of 120 (356993)
10-17-2006 5:16 AM
Reply to: Message 71 by Barbarian
10-17-2006 4:54 AM


That's enough
Dear Barbarian,
I have been trying to hold my nerves as much as I could with Mr. catlover to keep this discussion going on with you. But I am really getting sick of Percy (Admin)'s continuous personal attacks on me.
He has banned me for writing much less than what he wrote in his last post, and definitely I shouldn't expect justice to be done for the very reason that he seems to own this place.
I am willing to continue my discussion with you here:
http://www.youdebate.com/cgi-bin/scarecrow/topic.cgi?foru...
But I am not reading or writing a single post here any more.
Edited by Albert Godwin, : none

This message is a reply to:
 Message 71 by Barbarian, posted 10-17-2006 4:54 AM Barbarian has replied

Replies to this message:
 Message 73 by Barbarian, posted 10-17-2006 6:03 AM Albert Godwin has not replied
 Message 74 by Admin, posted 10-17-2006 10:10 AM Albert Godwin has not replied

  
Barbarian
Inactive Member


Message 73 of 120 (356994)
10-17-2006 6:03 AM
Reply to: Message 72 by Albert Godwin
10-17-2006 5:16 AM


Re: That's enough
Hi, Albert -
thanks for the invitation to go to the other forum but I have to decline it for now. I have a certain aversion to signing up to new boards without a decent amount of previous lurking, and besides, there is little else to discuss about this particular experiment. This whole challenge has now become just a programming pet project for me, which I will try to complete under the terms we discussed but otherwise I would rather have it completed first and justify my additional design decisions later.

This message is a reply to:
 Message 72 by Albert Godwin, posted 10-17-2006 5:16 AM Albert Godwin has not replied

  
Admin
Director
Posts: 12998
From: EvC Forum
Joined: 06-14-2002
Member Rating: 2.3


Message 74 of 120 (357027)
10-17-2006 10:10 AM
Reply to: Message 72 by Albert Godwin
10-17-2006 5:16 AM


Re: That's enough
Albert Godwin writes:
But I am really getting sick of Percy (Admin)'s continuous personal attacks on me.
It is fully within your power to bring the feedback you so dislike to an end. All you have to do is:
  • Follow the Forum Guidelines.
  • Respond to replies and rebuttals.
  • Discuss the topic. This is a discussion board, not an "issue challenges" board. Issuing a challenge is fine, but I'm very unhappy that there is no accompanying discussion of the topics of artificial life and genetic algorithms.
  • Gain some familiarity with the topic.
  • Refrain from issuing taunts.
  • Do not make allegations that you're unwilling to discuss and defend.
These are same things I've been telling you, and that you've been ignoring, all along. EvC Forum's goal is informed and constructive discussion, and unless and until you share this goal you will continue to have difficulty here. Ignoring moderators is another sure way to get into trouble. You're doing a good job of covering the gamut of ways to bring difficulties upon yourself here.

--Percy
EvC Forum Director

This message is a reply to:
 Message 72 by Albert Godwin, posted 10-17-2006 5:16 AM Albert Godwin has not replied

  
Jazzns
Member (Idle past 3912 days)
Posts: 2657
From: A Better America
Joined: 07-23-2004


Message 75 of 120 (357035)
10-17-2006 11:39 AM
Reply to: Message 31 by Dr Jack
09-29-2006 5:45 AM


Do you have any idea how radically irresponsible that would be?
Computer programs that self-replicate already exist, going under the name Computer Viruses - writing them is illegal.
Just to try to keep the quality of the opposition up, that is a HIGHLY ignorant statement Mr Jack.
First of all, it is most certainly not impossible. Like Percy said, it might be very difficult in that identifying the selector or selectors to get your desired result is a matter of some discovery and fine tuning. Also, evolving RSA encryption for example would be VASTLY more difficult than evolving ROT13.
Second, there are examples of programs that self-replicate for fun and profit that are NOT viruses or worms. There are even programs that evolve that are an interesting area of research in computer science.
IT seems as though this response from you was very knee jerk as well as being uninformed.

Of course, biblical creationists are committed to belief in God's written Word, the Bible, which forbids bearing false witness; --AIG (lest they forget)

This message is a reply to:
 Message 31 by Dr Jack, posted 09-29-2006 5:45 AM Dr Jack has replied

Replies to this message:
 Message 76 by Dr Jack, posted 10-17-2006 12:23 PM Jazzns has replied

  
Newer Topic | Older Topic
Jump to:


Copyright 2001-2023 by EvC Forum, All Rights Reserved

™ Version 4.2
Innovative software from Qwixotic © 2024