Commenter ‘fifthmonarchyman’ from UD has joined us and has some thoughts on active information.
I’ll cross-post his comment from the Ewert thread below.
148 thoughts on “Active information thread for fifthmonarchyman and others”
Leave a Reply
You must be logged in to post a comment.
This is as good a place as any for an experimental toe in the water. Let see how it goes.
Tom English asks
Your “search” process decides when to stop and produce an outcome in the “search space.” A model may do this, but biological evolution does not.
I say,
I’m not Dr Ewert but I’d like to give it a go.
In my opinion your observation is a key to understanding the incite of the paper.
We act as if evolution has stopped any time we appeal to evolution as an explanation for a particular specific outcome. I don’t think we can be justified in doing so precisely because evolution is an ongoing process and not an search for anything in particular.
I feel what is at issue is not the validity of evolution per se but it’s explanatory utility.
let me elaborate with an illustration
Suppose we observed an iceberg off the coast of Florida and I offered “ocean currents” as the explanation. Would you find my account to be satisfying?
While it might be true that “ocean currents” are the mechanism by which the iceberg got to where it is this fact has limited explanatory value.
In order to satisfy us an explanation would need to detail what it was that caused “ocean currents” to place this particular object at this particular place and time. IOW what made the process “stop” in such an unusual place.
Those are the sorts of things that we really want to know.
If I understand correctly It is those details that constitute active information.
peace
let the test begin
Cross-posting from the Ewert thread:
fifthmonarchyman on May 12, 2015 at 10:08 pm said:
This is as good a place as any for an experimental toe in the water. Let see how it goes.
Tom English asks
Your “search” process decides when to stop and produce an outcome in the “search space.” A model may do this, but biological evolution does not.
I say,
I’m not Dr Ewert but I’d like to give it a go.
In my opinion your observation is a key to understanding the incite of the paper.
We act as if evolution has stopped any time we appeal to evolution as an explanation for a particular specific outcome. I don’t think we can be justified in doing so precisely because evolution is an ongoing process and not an search for anything in particular.
I feel what is at issue is not the validity of evolution per se but it’s explanatory utility.
let me elaborate with an illustration
Suppose we observed an iceberg off the coast of Florida and I offered “ocean currents” as the explanation. Would you find my account to be satisfying?
While it might be true that “ocean currents” are the mechanism by which the iceberg got to where it is this fact has limited explanatory value.
In order to satisfy us an explanation would need to detail what it was that caused “ocean currents” to place this particular object at this particular place and time. IOW what made the process “stop” in such an unusual place.
Those are the sorts of things that we really want to know.
If I understand correctly It is those details that constitute active information.
peace
let the test begin
Hi fifthmonarchyman,
Welcome to TSZ!
In order not to disturb Tom’s thread, I’ve created a new thread for you and others to discuss active information.
keiths,
Excellent response. Thanks.
Fifthmonarchyman, I truly did not want Winston to be faced with a quagmire of comments on arrival. It’s not a matter of saving my thread, but instead his.
Hi fifth,
Again, welcome to TSZ.
Regarding your iceberg example, here’s how I would put it, speaking informally:
Suppose we see a large iceberg calving off a glacier in Greenland. Several weeks later we set out to find the iceberg again.
A truly blind search would look for the iceberg in random places throughout the universe. The likelihood of finding it this way is obviously extremely low.
If we take advantage of our knowledge of physics, we can limit our search to the oceans of earth. But we can do even better than that. Exploiting our knowledge of ocean currents, including real-time measurements, we can further limit our search to a small fraction of the ocean’s surface.
At each step we have improved the probability of success for the search. ‘Active information’ is Dembski and Marks’s attempt to quantify this.
Hey Keiths,
In your example the active information is our background knowledge including the knowledge of the original position of the iceberg. Do I understand you correctly?
If so I would have no problem with that.
Now suppose I asked you how you found the iceberg.
Do you think “I found it by a simple process of elimination” would be a satisfying response?
Next suppose I wanted to know how you found things in general would “I just look for them” capture the gist of your approach?
peace
fifthmonarchyman:
Not quite. Active information is a measure of how much better (or worse, in the case of negative active information) our search is when compared to the default blind search. Our background knowledge improves the search and thus contributes to its active information, but it is not identical with the active information.
Have you read the Dembski/Ewert/Marks paper?
That depends on your reason for asking the question. But if you are trying to equate evolution with a blind search for an iceberg, that won’t float (so to speak). Do you understand why?
No, but it also fails to capture the gist of how evolution works.
Steps out for popcorn…
I have advocated the ID movement abandon or at least de-emphasize information theory arguments in favor of expectation analysis.
Here was my cordial exchange with my colleague Winston in December on that topic:
http://creationevolutionuniversity.com/forum/viewtopic.php?f=3&t=105
Hey Sal,
I’m actually thinking more about the strengths and limitations of algorithmic processes lately with an emphasis on non-computability as the hallmark/definition of design. I’d like to kick around some stuff in that regard but I’m not sure if this will turn out to be a good place to do so.
I don’t begrudge any approach however and I’ll check out your link when I get a minute.
peace
I moved FMMs original comment here.
Welcome FMM, BTW!
That would be interesting. How would you distinguish that from irreducible complexity, or from Behe’s Edge?
petrushka says,
How would you distinguish that from irreducible complexity, or from Behe’s Edge?
I say,
I’m in the “found a hammer so everything is a nail” phase of discovery right now. 😉
I would say that the notion of non-computability is a unifying concept for most everything that’s come down the ID pike. That includes CSI and active information
The light came on for me when I read this paper concerning Integrated Information Theory.
http://arxiv.org/abs/1405.0126
It hit me that Integrated Information the phenomena that Tononi and others associate with consciousness is synonymous with irreducible complexity.
The kicker is that paper demonstrates that algorithmic processes like evolution are mathematically incapable of producing Integrated Information. Thus proving Behe right and giving a mathematical basis for his edge.
That realization rocked my world.
It’s been quite a ride from there. I would love to talk about it among critics without getting into the tired old culture war rut.
peace
Not really my cup of tea, but with a conclusion like that, I couldn’t possibly turn you down. Hopefully Keith will join in also. He’s a very bright guy, if you haven’t picked up on that already.
Let me finish what I’ve got on my plate. In the meantime, you might work on firming up your arguments. Some formalism would be nice. I’ll mention that the “search” process of Dembski, Ewert, and Marks is not necessarily algorithmic. And dynamical systems like recurrent neural networks supposedly can realize super-Turing computations. The strong Church-Thesis is only a thesis, you know. Arguments about what natural processes can and can’t compute are difficult to make. I’m not throwing that at you as an ID proponent. It’s just the way things are.
fifth,
By all means, write up your argument and post it here.
We’re tough but fair.
Okay… so this “integrated information” dealie is just another term for “irreducible complexity”, such that one term can be substituted for the other without any loss/distortion of meaning.
Substituting terms: “[T]hat paper demonstrates that algorithmic processes like evolution are mathematically incapable of producing
Integrated Informationirreducible complexity.” Alas for Behe (and ID), evolution can produce irreducible complexity. In layman’s terms, evolution produces irreducible complexity by means of what has been called the ‘Mullerian two-step’; step 1 is to introduce a new part to an existing system, and step 2 is to modify one of the other parts (i.e., the parts that existed before the new part got introduced in step 1) in such a way that the system can no longer function in the absence of that new part. See Genetic Variablity, Twin Hybrids and Constant Hybrids, in a Case of Balanced Lethal Factors, by Hermann J Muller, in Genetics, Vol 3, No 5, Sept 1918, pp 422-499 for further details.Apart from that, the abstract of the paper you cited says:
Note: “if unitary consciousness exists”. Does it? Also, “This result implies that… unitary consciousness… cannot be modelled computationally”. Is that implication true?
There is so much going on here It’s hard to know where to start. I guess I’ll just jump right in,
cubest says
Alas for Behe (and ID), evolution can produce irreducible complexity.
I say,
Perhaps it would be helpful to be sure we are talking about the same thing.
Here is the rough and ready definition of Integrated Information
quote:
In a system composed of connected “mechanisms” (nodes containing information and causally influencing other nodes), the information among them is said to be integrated if and to the extent that there is a greater amount of information in the repertoire of a whole system regarding its previous state than there is in the sum of the all the mechanisms considered individually.
end quote:
That is the definition that I’m calling synonymous with IC. There are lots of other definitions floating around but that is the one I’m using. just so there is no confusion
Next it’s important to know that the integration in the paper is a specific kind of “nonlossy” integration
quote:
An integrating function’s output is such that the information of its two (or more) inputs is completely integrated.
i.e, the knowledge of m (z) does not help to describe m(z′),when z and z′ are close.
end quote:
This kind of integration is indeed non-computable it is indeed irreducibly complex. There is no debate here it’s a matter of simple mathematics.
This I believe is what Behe is talking about and if such a phenomena actually exists evolution is incapable of producing it by definition.
let me know if you guys understand this and if so we can move on
peace
Welcome to TSZ, fifthmonarchyman!
Welcome!
I’m not sure I do understand it.
Could you give a specific example perhaps? And how is the “amount of information” calculated?
cubist says
“if unitary consciousness exists”. Does it?
I say,
I believe this is also beyond debate. Unitary consciousness implies that there is one whole unified person known to me as “cubist”. To deny it would I believe render our communication impossible. I don’t want to go there.
you say,
Also, “This result implies that… unitary consciousness… cannot be modelled computationally”. Is that implication true?
I say,
You can read the paper and come to your own conclusion. I know I’m convinced
peace
OMagain asks,
Could you give a specific example perhaps?
I say,
Lets use the worn out example of a mouse trap. The integrated system is indeed more that the sum of it’s parts. integrated it’s a working mouse trap dissembled it’s a pile of parts.
It’s important to know that the integration takes place in the observers head. I’m the one who decides if a particular assembly of parts is a mouse trap and not something else
you say,
And how is the “amount of information” is calculated?
I say,
One way would be to calculate the information in the best approximating model and subtract that value from the information in the system itself .
The difference would be the key number.
peace
[shrug] I take “irreducible complexity” to be what Behe says it is, a quality in which a system cannot function if even one of its parts is broken/absent. “Integrated Information” I’m not so sure about. But if this “Integrated Information” thingie is indeed synonymous with “irreducible complexity”, then any alleged proof that evolution can’t produce “Integrated Information” is flawed, because yes, Virginia, evolution can produce “irreducible complexity”. So the question is, is “Integrated Information” truly synonymous with “irreducible complexity”? If so, any evolution-can’t-produce-Integrated-Information proof must necessarily have some sort of flaw that renders it inapplicable to evolution, if not just plain invalid.
Cool. Now all you need to do is (a) identify some System X which genuinely is irreducibly complex; (b) specifically identify every last one of the “mechanisms” of which System X is composed; (c) measure the information content of every last one of the “mechanisms” in System X; and (d) demonstrate that the “amount of information in the repertoire of [System X] ” actually is greater than “the information among [the ‘mechanisms’]”.
Okay, fine. Does any existing physical system exhibit that particular type of “‘nonlossy’ integration”?
Your conclusion doesn’t follow unless Muller’s 1918 paper is wrong. Which it could be, of course, but “could be” is most assuredly not synonymous with “is” or “has been proven”. So your next step is to demonstrate that the so-called ‘Mullerian two-step’ cannot produce irreducible complexity.
Fine, but the paper argues that “it is more natural to frame consciousness as a lossless integrative process”. If consciousness is indeed a “lossless” process of whatever sort, how come it’s possible to forget anything?
Groovy.
Exactly how does one go about “calculat[ing] the information in the best approximating model”?
And since you need to know “the information in the system itself” couldja tell us how to measure the “information” in a system? Since Omagain asked about mousetraps, perhaps you could show us all how this works by determining the “information in the best approximating model”, and “information in the system itself”, for a mousetrap.
I’m not sure what that means, if it even means anything at all. What precisely is the meaning of “information among them” and how is it measured (so that one can talk of “greater amount”). And what does “previous state” mean here?
What does “nonlossy” mean here? I know what “non-lossy” compression means. It means that one is dealing with what is explicitly Shannon information, and the compression is such that the exact original sequence of symbols can be recovered. But now you are talking of “integration” and using “information” in a was that is far from clear. Can you flesh out the details?
Again, it is unclear what this is saying.
If you are talking of Shannon information, then it seems to be talking of a finite problem. And finite problems are computable.
If you are not talking of Shannon information, then it isn’t clear that computation is actually involved. And that would suggest that “non-computable” is just a bullshit term thrown in to obfuscate.
Could you do that then, as a starting point?
If not, perhaps we should take a step back and re-assess?
Not to derail the thread but has anyone heard a peep from Winston Ewert? Since his debacle over at UD he seems to have vanished. He’s certainly shown no interest in the questions he was asked at Google Moderator.
Perhaps not entirely on topic, but over at UD, there’s a thread on junk DNA. I’ll post here because the concept of junk is related tto the concept of biological information.
Mung asks:
Interesting question. I wonder why biologists never thought of that?
fifthmonarchyman,
Keiths invited you to make a post. That’s what I had in mind. I’m not looking for an essay describing math you haven’t actually done, but feel is easy to do. (Take Dembski’s No Free Lunch as an object lesson.) I’m calling on you to make it clear how you arrive at your conclusions. You can’t do this without showing us the math. Put {latexpage} (replace curly braces with square braces) at the beginning of your post. I think you’ll need to edit in the raw text mode to get full access to QuickLaTeX. Please don’t post now. I want to interact with you, but can’t squeeze any more formal stuff into my brain until I’m through addressing Winston.
Show me. (Not in this thread.)
You should state your assumptions explicitly. There was a reason that I said above that the strong Church-Turing thesis is just a thesis. There was a reason that I brought up super-Turing computation in recurrent neural nets. I had a pretty good idea of how you had to proceed with your argument.
If you genuinely want a discussion, then you should not address us as though you’re our teacher.
Given that this is the “holding pattern” thread, I’ll say that Winston posted last Saturday, and that today is Thursday. The “Ask Dr Ewert” questions that he’s received through Google Moderator are challenging. It’s unreasonable to expect him to respond to any of them before the weekend.
Note that DiEb posed three of the four questions, and has since been banned from UD. I’m not saying that correlation implies causation. But the fact of the matter is that DiEb cannot comment on answers that Winston posts at UD.
cubist asks,
Exactly how does one go about “calculat[ing] the information in the best approximating model”?
I say,
Lets take the Pi for example.
If we were to represent Pi numerically we would have a irrational constant. In other words Pi is non-computable in the sense that we are using the term here
This is intuitively obvious given that the information in the concept PI is fully integrated. So the K complexity necessary to reproduce PI entirely would be effectively infinite, Again that is just another way of saying that PI is non-computable.
Now there are many algorithms that can approximate PI to any given level we choose I happen to like this one
M = 4-4/3+4/5+4/7-4/9…..
now the K complexity of M will vary depending on the programing language we use but it will be finite and the resulting equation would look like this.
K complexity (Pi) – K complexity (M) = K complexity (Pi – M)
Infinity – K complexity (M) = K complexity (Pi-M)
Infinity = K complexity (Pi – M)
in other words no matter how we slice it the extra information in the Target/Specification that is PI is infinitely greater than that the information in any model that approximates it.
Which is just another way of saying that the Integrated information is Pi is not computable.
peace
cubist asks,
If consciousness is indeed a “lossless” process of whatever sort, how come it’s possible to forget anything?
I say,
If you forget something it is not a part of your unified consciousness by definition.
I have forget many things that I used to know but none of those things are a part of my unified conscious self.
When we are talking about your unified consciousness are talking about what it is about you that makes you you. Stuff you might have knew at one time but now don’t obviously is not part of who you are. It it were you would not be the same person you were back then.
peace
fifth,
How about writing up your entire argument and posting it?
Tom English says,
Keiths invited you to make a post. That’s what I had in mind.
I say,
I appreciate the offer and might take you up on it eventually but if this thread so far is any indication such an enterprise would quickly get bogged down in the preliminary definition phase.
perhaps once we can agree on some terminology I’ll work on a summery of the argument
We have already seen some difficulty coming to an agreement on something as simple as the equivalence of Integrated information and Irreducible complexity.
My fear was that folks have too much invested in not giving the interloping fundi even an inch of ground and so far it’s turning out that way
Peace
fifth,
Why not simply define your terms and present your argument?
Tom English says.
There was a reason that I brought up super-Turing computation in recurrent neural nets.
I say,
My argument is not about materialism verses dualism so I have no problem with Hypercomputation in neural nets. Why should I?
It’s possible that information integrating is an entirely phyiscal process but it’s not an algorithmic process.
peace
FMM, for future reference…
I got the aligned equations right in my preferred TeX system. I put {latexpage} at the beginning of this comment, and then copy-and-pasted the LaTeX code between {latex} and {/latex} (replace curly braces with square braces). It seems there’s no fix if you don’t get this right in your initial comment. I can’t get editing to work.
Keiths says,
Why not simply define your terms and present your argument?
I say,
lets see if we can come to a consensus on this one simple definition first.
It is highly possible this is just not the place to discuss this stuff.
peace
I see an “edit” near the top right of your message. If you can see that, give it a try.
There’s another alternative, but again I’m not sure if you can use it except on your own thread. If you click the little box near the top of the screen (looks like the speech bubble in comics), that should get you to the comments screen. If you can find your comment there, see if there is a “Quick Edit” near the bottom. That give plain text editing.
The third option is to ask an administrator for help.
fifthmonarchyman,
I won’t be commenting further on your thread-within-a-thread. Write a detailed OP, connecting all of the dots. I will give it close attention if you wait a while to post.
+1.
I see that only in comments on my own thread. It does in fact allow editing of LaTeX.
All I see here is “0 comments awaiting moderation.”
Well… Years back, we had the option to work with HTML in comments. Was there a problem with that?
keiths:
fifth:
Just state your definitions within your OP. We’re capable of keeping your definitions in mind as we evaluate your argument.
I doubt that there’s a better place on the Internet to discuss this. There are people here who understand ID, computation theory, Kolmogorov complexity, and various theories of consciousness including Tononi’s.
You’re unlikely to find that anywhere else, and certainly not among the regulars at UD.
Richardthughes says,
How about writing up your entire argument and posting it?
I say,
1) Why is having the whole argument laid out all at once so important? Is it possibly because instead of an open dialogue you are hoping to find an easy technical flaw that will allow you to dismiss the whole idea with out taking the time to understand the general drift.
2) I’ve not taken the time to put together a systematic treatment as of yet. When I do I’d like to be sure we are talking about the forest argument and not loosing ourselves in trees of definitional minutia.
3) I’m not even sure what part of the argument I want to emphasize.
Do I need to demonstrate that evolution is an algorithmic process?
Do I need to explain how specifications and targets and integrated information are all the same thing?
Do I need to show how Active Information is simply another way of thinking about CSI?
Do I need to establish that Informational integration/Irreducible complexity is observer dependent or is this obvious?
etc etc etc
you get the idea before long a simple discussion becomes a paper.
I’m not sure I’m ready or capable of such a step
peace
I think it might be good to know why you speak of specifications and targets, since no one in biology thinks evolution has specifications and targets.
Just do your best, fifth. We’ll ask for clarification if we need it.
You say that non-computability is “the hallmark/definition of design.” Write up an OP explaining clearly why you believe this.
Because it makes it easier for us to understand what you are trying to argue. If one statement is confusing, we can read ahead and see if the rest of the argument helps to make your position clearer.
petrushka asks,
I think it might be good to know why you speak of specifications and targets, since no one in biology thinks evolution has specifications and targets.
I say,
This is a great question.
I think a big part of the communication gap is that folks on my side tend to think in terms of specifications and targets. Things like birds and species and bacterial flagellum. Specific particular things existing in space and time that we intuitively think about when we think about life.
We want to know if “evolution” explains these things. It’s what interests us.
Poplar science writing seems to imply that evolution can explain stuff like that and EDM are trying to show that it can’t. I happen to agree
It has been convincingly argued that evolution is not even interested in things like these. If evolution is not looking for particular things I don’t understand how it can be appealed to explain them.
It’s like explaining the iceberg on the Florida cost in July by appealing to “currents”.
regardless
In order to even have a discussion we need to agree just what kind of things birds and bacterial flagellum are.
Folks on my side understand them to be specifications/targets if they are not what are they?
peace
Is there an ID explanation for an iceberg in Florida, assuming the explanation isn’t human? Or an ID explanation for the number of hurricanes in a season?
You seem to be saying that retrospective astonishment is evidence for ID.
Keiths says
You say that non-computability is “the hallmark/definition of design.” Write up an OP explaining clearly why you believe this.
I say,
Ok against my better judgement I’ll put something short together. Don’t let it be said that I did not bend over backwards to appease you guys
😉
Do me a favor and give me a step by step on how to post.
thanks in advance
peace