I want to thank OMagain in advance for doing the heavy lifting required to make my little tool/game sharable. His efforts will not only speed the process up immeasurably they will lend some much needed bipartisanship to this endeavor as we move forward. When he is done I believe we can begin to attempt to use the game/tool to do some real testable science in the area of ID . I’m sure all will agree this will be quite an accomplishment.
Moving forward I would ask that in these discussions we take things slowly doing our best to leave out the usual culture warfare template and try to focus on what is actually being said rather than the motives and implications we think we see behind the words.
I believe now would be a good time for us to do some preliminary definitional housework. That way when OMagain finishes his work on the gizmo I can lay out some proposed Hypotheses and the real fun can hopefully start immediately.
It is always desirable to begin with good operational definitions that are agreeable to everyone and as precise as possible. With that in mind I would like to suggest the following short operational definitions for some terms that will invariably come up in the discussions that follow.
1. Random– exhibiting no discernible pattern , alternatively a numeric string corresponding to the decimal expansion of an irrational number that is unknown to the observer who is evaluating it
2. Computable function– a function with a finite procedure (an algorithm) telling how to compute the function.
3. Artifact– a nonrandom object that is described by a representative string that can’t be explained by a computable function that does not reference the representative string
4. Explanation –a model produced by a alternative method that an observer can’t distinguish from the string being evaluated
5. Designer– a being capable of producing artifacts
6. Observer– a being that with feedback can generally and reliably distinguish between artifacts and models that approximate them
Please take some time to review and let me know if these working definitions are acceptable and clear enough for you all. These are works in progress and I fully expect them to change as you give feedback.
Any suggestions for improvement will be welcomed and as always please forgive the spelling and grammar mistakes.
peace
Checkmate
Are we talking about recognising patterns or recognising meaning? These are completely different things.
fG
recognizing patterns.
I would not say that are completely different but there are differences none the less. You are right to point that out.
Observers might pick up on different patterns in the same string and they might pick up different meanings from the very same pattern.
The bottom line is that we are looking to see if an observer can distinguish between strings by looking at the overall patterns and not just individual digits
peace
What do you think spellcheckers do?
fG
Pattern and meaning are like syntax and semantics. Patterns are properties of the object itself, are observer-independent and can be quantified. Meaning relies on extraneous conventions between the originator and the observer (the ‘users’), can’t really be quantified, and is pretty much independent from the object because any object can be assigned meaning, if the users so decide.
You have to be very clear and precise on what you want your game to detect, and on what the implications are if it succeeds.
fG
Good question.
I’m no expert but I think they have a database of words and look to see if an unknown word is close to a word in the database and offer suggestions.
If they were good at looking at the over all patterns in a text I would say we would not get the silly suggestions we sometimes get.
As has been pointed out humans can often get the gist of a long text even when every word is misspelled. I don’t think computers are good at that sort of thing
I could be completely wrong about this though
I think you are correct that spell checkers have lists of acceptable words. Some appear to have rules for possessive and plural endings. Some have lists of common phrases, and some have lists of common phrases that are frowned upon.
Computers can be very good at general pattern recognition. Voice recognition has gotten really good in recent years. My $70 phone can hear better than I can and transcribe spoken english better than I can.
Getting better at tasks like translation is just a matter to time, now, and not a problem in principle. The Google translator operates on probabilities rather than word lookup. It is just a matter of time before translators have evolved intrinsic understanding rather than lookup tables. Like the tone discriminator.
I think this has to do with the Y axis that I talk about.
take the string 123456789
What is the pattern? Neil is right to point out that the only objective thing we can say is that we have 9 characters with no repeats.
However subjectively I know that we have the first 9 digits of the Arabic numerical system in ascending order.
Moving even higher on the axis I can say this particular string is the string that we have been using in this particular thread to illustrate the concept of integrated information.
The height on the Y axis is a function of the shared contextual framework between the “observer” and the “designer”.
It seems that the more context that is shared the less information is necessary to make a design inference.
I know that 3.14 is PI even though those digits are not all that unlikely to arise at random.
Peace
PS thank you again for the interaction. Your comments are thought provoking. That is what I wanted from this thread
peace
Understanding.
What is your endeavor relevant to?
That is correct: they look at the overall input string and compare it with already known strings. They don’t look at each letter in the text by itself (in your words: ‘overall strings not individual digits’). So what you are after is being done already.
Here you are already confusing pattern and meaning. What you should have said is: “If they were good at looking at the over all meaning in a text I would say we would not get the silly suggestions we sometimes get”.
Although the meaning of a sentence can differ depending on the pattern of the words it is formed from, this is not necessarily the case. In English, yes, re-arranging the words in a sentence can radically alter the meaning. In Latin, for instance, this is not the case – you can pretty much put the words of a sentence in any order and the meaning would stay the same. A context-sensitive English spell checker would look at the order of the words (the pattern), whereas a context-sensitive Latin spell checker would ignore the word order and concentrate on other aspects of the grammar (inflections for instance).
Pattern is not meaning; meaning is not pattern.
I don’t know the current status of this field but I suspect that it is possible to write expert systems that are robust, to some degree, to spelling mistakes in the user input. Probably not yet as robust as actual humans, so it may require a bit more interaction to iron out misunderstandings, but I’m sure it can be done to a high degree of effectiveness. The lexicon available to the computer would be the context in which it would interpret the user’s input.
It is a problem in principle if we are dealing with nonlossy integrated information. and I think we are.
Time will tell
peace
You have not convinced anyone that non lossy integrated information exists. Brains are certainly not non lossy. You last attempt to defend this was a disaster.
But just a flesh wound.
Ok I am beginning to understand where you are going with this. The problem you will face is that what you call a ‘design inference’ is not actually that. It is merely you applying knowledge that was transferred to you in your past. Someone taught you the meaning of the symbols 1, 2…9, and someone taught you that a certain order of these digits represents an ascending rank. If you hadn’t been taught these conventions you would never notice anything special about 123456789 except perhaps that it looks a bit like script (assuming someone taught you to read and write).
Understanding the meaning of unknown (dead) languages from inscriptions is notoriously hard. Look up Rosetta Stone if you want a good example. Egyptian hieroglyphs display many known patterns, yet it required a one-to-one comparison with a known language to be able to assign meaning to those patterns. People suspected that the hieroglyphic text had meaning, of course, from the overall context in which they were found, but if all we had was just a few isolated examples of hieroglyphs they might have been nothing more than cute little drawings, as far as we could tell.
Same with Pi. You have been taught what Pi is therefore you recognise it when you see it. There is nothing special about the string 3.14 seen in isolation from this knowledge.
Your design inference is really nothing more than your internal lookup table in context.
fG
I think pattern and meaning are closely related but not synonymous.
FYI
I have run numerical representations of texts in languages that I don’t know through the game and right away I pick up on patterns like no repeated words.
The better the fake string I use the more subtle the “pattern” must be for me to distinguish between the two strings.
So far no fake has been good enough to fool me but some have gotten close.
My hypothesis is that I will always be able to distinguish between the real and the fake with feedback. We shall see
I find it all to be gobs of fun. I can’t wait to share it
peace
Perhaps you could think of that that way, If you did you would need to specify that there is at some level a universal entry that codes to “design” in the table.
something like====== thinks like me
peace
I intend to submit excerpts from the Voynich manuscript. Your discriminator should be able to crack this puzzle.
What exactly do you mean by real and fake?
have you fead anything in this thread? 😉
FG suggested we substitute “fake” for “explanation” in my list of definitions
“real” is simply the original string up for evaluation
peace
fifthmonarchyman,
No, I’m talking about individual memories and groups of related memories, just as the paper does.
The paper says nothing like that. If it did, it would be wrong. Memories do change when they are remembered.
Here’s the argument from the paper (anyone uninterested in the details can skip down to “tl;dr” below):
From the Introduction:
Note the claim that each observation is integrated with previous memories.
From Requirement 2: Generating Integrated Information comes the discussion of the difference between human memory and the authors’ mental model of a smell detector:
As I’ve previously explained, it’s possible to train a neural network to categorize smells without using the database approach suggested in the paper. Such a network will have a similar response to changes as a brain, with connections and their weights involved in categorizing multiple smells. By the criteria expressed in the paper, such a neural network would contain integrated information.
In the Quantifying Integrated Information section is the demonstrably incorrect assertion about human memories:
However undesirable the authors find it, that is how human memories work.
The following paragraph summarizes the intent of the rest of the paper:
Remember, they aren’t talking just about cognition, they are talking about “the requirement for a conscious observation to be integrated with previous memories, hence generating integrated information.”
The authors then discuss Kolmogorov complexity of binary strings (maybe they’d like a copy of a sonnet in binary). They attempt to conflate data compression with integrated information. Their argument includes more than a bit of hand waving, but I’ll leave that aside for now since there are larger problems with the paper.
With integrated information magically linked to data compression, they then talk about Quantifying Integration Using Edit Distance. There’s more hand waving here attempting to tie data compression to the way human brains behave. This reification of a bad metaphor is unconvincing.
In the section On the Computability of Integration the authors claim to demonstrate “that lossless information integration cannot be achieved by a computable process.” Leaving aside, again, whether or not their math means what they say it means, their conclusion is:
tl;dr
In summary, the paper’s argument is this:
1) Conscious observation must be integrated with previous memories, which generates integrated information.
2) Memory functions must be vastly non-lossy, otherwise retrieving them repeatedly would cause them to gradually decay.
3) Lossless information integration cannot be achieved by a computable process.
4) Therefore we have to abandon either the idea that people enjoy genuinely unitary consciousness or that brain processes can be modelled computationally.
Since (2) is demonstrably false, the conclusion in (4) is not supported.
The paper is thus refuted without even going into the questionable mathematical arguments or criticisms of Tononi’s theory.
Actually something like that has already been done but from a different angle
check it out
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0067310
peace
petrushka,
Or perhaps a sonnet?
agreed, your subjective idea of smell of chocolate is a function of among other things all the times you have smelled chocolate
Correct, the more I think about my subjective idea of the smell of chocolate the better that idea is ingrained in my mind, My subjective idea does not degrade with recall
mathematically proven
It necessarily follows of course.
Notice it does not say “perfect unitary memory” but “unitary consciousness” there is a world of difference between the two concepts.
Apparently you still don’t get the point of the paper.
Try again and this time don’t assume the authors are stupid.
peace
If you code a sonnet in base 10 instead of binary we can do this together and start as soon as OMagain finishes. Or If you want I can do it now.
If you insist on binary it will take some modifications and time
peace
How do you construct the fake?
You should, because I am finding it hard to discuss this with you when I don’t know the details of your game.
fifthmonarchyman,
Yes, it does.
That claim in the paper is solidly refuted. It can no longer be used to support any argument.
fifthmonarchyman,
You can take any of the remaining binary strings I provided and easily convert them to base 10 if that’s what your program requires.
Right now I take the original string and randomize it as to order.
Then I run it through a crude EA in which I create several copies that vary slightly compare them with the real string and keep the one with the highest r squared value. I repeat this process until the r squared reaches 80 or so
I open to any algorithmic process that does not specifically target the exact digits in the original string
peace
My game is just a re purposing of the one found here.
http://arxiv.org/pdf/1002.4592.pdf
You might give it a try to understand what I’m talking about
I play it in a excel spreadsheet.
I’ve also coded it (again crudely) in the processing language
https://processing.org/
I would be happy to share this with you if you would let me know how to contact you or we can just wait on OMagain
peace
fifthmonarchyman,
Just curious. How does that differentiate from one numerical string or another? Unless you have some other input about how the string was formed, I don’t see any way, by just looking at a sequence of numbers, you can say anything about how it was generated. For example, take a roulette wheel. Rather than equally-sized slots, let’s make them adjustable. Make 36 slots, but make them adjustable, from 0° to 360°, so we could have one slot of 360° and 35 of 0°, or any other combination. How biased from 10° would the slots need to be for a string generated by this roulette wheel for this to show up as biased?
ETA degarbalization
Since you apparently are for some reason incapable of understanding what the paper is saying let me make this easy for you,
when you are successful in your little hack the papers findings will be refuted. Why don’t you get after it ?
peace
fifthmonarchyman,
My detailed analysis and summary of the paper that I posted this morning disproves this assertion. The paper’s conclusion depends on a claim demonstrated to be false. Therefore the paper’s conclusion is not supported.
You’ve mentioned that you want to talk about science. Here’s your chance. The scientific evidence shows that human memory functions are not lossless. That evidence rebuts the paper you like so much. If you’re really interested in science, this should please you. You now have knowledge you did not have before and you have an opportunity to correct your understanding of the world.
Or, you could just cling to a paper that you think supports your beliefs despite it being disproven.
faded_Glory,
For some of the history behind the trainwreck, see this thread:
Design as the Inverse of Cognition
It’s a comedy of (fifth’s) errors.
What if the original pattern is 1111111? The “fake” one will also be 1111111.
What if one picks 11115555? The “fake” one will be 55551111.
Shuffling a pattern is not guaranteed to produce the appearance of randomness. In fact, your choice of picking the highest square r will tend to produce non randomness
I saw that paper earlier. I don’t see what this has to do with ID. After all, financial market returns are not designed. I don’t know what exactly causes people to spot a fake return from a real one, but one being designed and the other one not cannot be the answer.
Are you now proposing a tool (or game) that shows a ‘designed’ (by you) string next to a perturbation, and then let people choose which one is the original and which one is the perturbation?
fG
Patrick it does no such thing it shows that human memory is fallible to that I say. duh. Everyone who ever lost their keys or tried to cram for a test knew that already.
For some reason you are equating lossless with infallible which is just plain silly.
When we say a memory function is lossless all we mean is that you can’t change any part of a memory with out altering it. It has nothing to do with how accurate our memories are accurate or whether we tend to forget stuff over time.
This has been repeatedly pointed out to you yet you seem to keep going back to the same dry well
I find the fact that you are not getting this to be odd. If I did not know better I might think you are being deliberately obtuse
peace
If the pattern is 11111111 we will be unable to distinguish it from a randomized string. and therefore we will not infer design
When we actually have a shareable game I will explain the way we decide if a string is complex enough to evaluate. It’s really cool and simple.
Keep in mind that the strings run in a loop
basically what we do is is split the real sting at a random place so that the beginning shifts places then we see if we can distinguish the new string from the real one if we can then we don’t have enough complexity to evaluate the string.
does that make sense?
randomness is only the first part of the game. in order to infer design a string needs to me nonrandom and noncomputable.
That means we need distinguish the real string from both a randomized string and algorithmicly produced model that is close to it.
Again this is all pretty self exclamatory when you see the actual game
peace
fifthmonarchyman,
“You can’t change any part of a memory with out altering it.” Of course not, that’s what change means.
The fact remains that memory function is not lossless as asserted in the paper. Recalling memories does change them.
Since the conclusions of the paper depend on that disproven claim, they are unsupported. Science. Supposedly what you want to discuss.
You can’t say any thing about how it was generated. All you can say is that is was not produced by a random process and is not computable.
What you infer from that is another matter entirely.
I think we can demonstrate that humans tend to infer design when we encounter an object that is not random or computable.
That does not mean that such an inference is necessarily accurate true or that we mathematically compelled to make it. It only means that is what we do.
Like I have repeatedly said This will boil down to the problem of other minds
The critic can relax they will always be an escape hatch.
I can’t prove that a designer exists anymore than I can prove that I am not the only mind in the universe
peace
Exactly now you perhaps are finally getting it.
I apologize in advance for even doing this because I know you are more familiar with this stuff that I will ever be.
But perhaps you are having some sort of mental block
Maybe jogging your memory will help you understand what we are discussing,
Think of a jpeg or MP3 file verses a phone number.
a jpeg file can loose many bits of info with out any change to the image. The phone number on the other hand can’t loose any information at all or it is rendered worthless
get it?
peace
fifthmonarchyman,
No. It has nothing to do with the fact that I’ve provided a reference to evidence that contradicts one of the main claims of the paper you’ve been pushing and hence eliminates any support for the conclusion.
If you disagree, prove it with reference to the paper and the evidence provided.
What sort of poof do you need?
The paper stands on it’s own it’s not difficult
You are completely misunderstanding it by bring up stuff about fallible memories.
Think again about the smell of chocolate.
The smell of chocolate is a lossless memory. If you change one part of the memory it is no longer the same memory.
This is so basic that is seems condescending to point it out to you. I apologize but I just don’t know how to make it any simpler
peace
fifthmonarchyman,
That’s not how memories work, as the references I provided illustrate. Read them.
Recalling memories changes them. That’s a fact. The paper you’re pushing relies on the opposite being true. That’s a fact. The conclusions of the paper are therefore unsupported.
I would strongly disagree, Market returns are the artifact of millions of discrete decisions by conscious agents. I can’t see how you could possibly think of them as anything but designed
It doesn’t have to be designed by me.
The game is simply a nifty means for us to see if a particular real string can be distinguished from fake (randomized or algorithmic) ones.
peace
You just don’t get it. I have read them they are irrelevant to the point of the paper.
let me try again
Take the memory of your phone number if I substitute one digit does that change the memory? If it does then we can say that your phone number is a lossless memory
Do you understand that much?
peace
Or look at it like this suppose I secretly created a fake Patrick that was exactly like you in every way except he was a Amish farmer.
Would my fake be you?
Come on man use your head
peace
fifthmonarchyman,
I pointed out the exact references in the paper that make them relevant.
If you disagree, show how my references don’t do so. Direct argument, not bad analogies.
no you pointed out references in the paper that apparently you completely misread.
When you read “memory” you are thinking of squishy reminiscence when you should be thinking “data compression”.
Why you are thinking reminiscence when the name of the paper is “Is Consciousness Computable” and it has sections entitled “Consciousness as Integrated Information” “Data Compression as Integration” is beyond me.
from the paper
quote:
In the following section we formally prove that, given the
Partial Information Decomposition (Williams & Beer, 2010)
formulation of synergy, the amount of integrated informa-
tion an information-lossless process produces on statistically
independent inputs is equivalent to the data compression it acheives
end quote:
There is nothing about telephone games or forgetting where your keys are.
Why can’t you see that?
peace
Patrick,
In an effort to give you every benefit of the doubt I have reread your “response” to the paper
quick question
In your opinion what is “just” cognition? How can cognition occur with out integrating an experience with previous memories?
peace
This is unwarranted. Plenty natural phenomenons produce patterns, and we design random generators, so we design things that produce randomness. I’m afraid this is going nowhere if you intend to somehow build a design detector.
Would you say that “natural phenomenons” can produce patterns that are nonrandom and noncomputable?
can you provide some examples?
peace