"It from bits"--the Universe is computable

[Anselm33]

Well, I’ve not read Shadows of the Mind, but I have read parts of Penrose’s The Emperor’s New Mind, and lots of related discussion online, over the years. I think this bit from Penrose on the subject here is relevant:

This I think clearly locates Penrose’s objection in his understanding of Gödel Incompleteness. He talks later in the article about various criticisms of his take on this, and allows that Daniel Dennett’s “bottom up” argument is one that bears more analysis (which I think an understatement, but a fair one), but the problem is more fundamental, I say: Gödel’s theorem needn’t apply for strong AI, or “computational consciousness”. All computational systems we have now that are formal such that Gödel’s theorem applies still compute, still resolve. Gödel’s insight doesn’t invalidate or change that. Rather, it just notes that the system has transcendental limitations. It’s self-consistency necessarily means it cannot possibly be complete in its expositions, it’s proofs.

That in no way denies strong AI in principle, and if you’ve read Penrose on this, you know that he doesn’t claim is does. He just thinks there’s something “simply non-computable” in consciousness. You don’t need to rely on Gödel for that, and Gödel can’t even help you get there, or show that to be true.

Penrose says this later in the article:

There’s nothing to “get around” to establish the viability of strong AI, per Gödel, that I can see. Any “computable consciousness” would have true propositions that were non-computable, but so does any formal system, and we have plenty of examples where that is no crisis of computation. I’ve read a lot of Penrose, but not all, but from what I gather, he never shows us where the limitation actually obtains. That makes sense, and is fairly obvious, because we don’t have a working, robust model of consciousness yet as a reference point. We don’t even know what we are applying Gödel to, if it is even applicable.

Have I missed the main substantive point from Penrose here?

-TS

“Shadows of the Mind” is more expanded than “The Emperor’s New Mind” (ENM) and it answers critiques raised against the arguments in ENM. In Shadows Penrose has an extended discussion of why mathematical intelligence and insight is a non-computable process, a much more extended discussion than in ENM. The Halting Theorem business is in fact, not his only argument against computable awareness.
Now I won’t deny that digital computing can emulate many aspects of human intelligence, whether from top-down or bottom-up type programming (e.g. for the former, chess-playing and for the latter medical diagnoses of images via neural network training). Nevertheless, if there is only one mathematical procedure that can’t be done computationally IN PRINCIPLE (not because of present day limitations on computing power), then that indicates to me that there are things–concepts, thought processes–that humans can do that computers will not be able to do and that therefore the universe (which includes humans?) can not be computable.
And I’m not necessarily using the argument about formal systems, which Penrose uses to say that humans can do math that a computer will never be able to do.
Here is the specific URL for the polyominoes (taken from Shadows of the Mind):
books.google.com/books?id=gDbOAK89tmcC&pg=PA30&lpg=PA30&dq=non-computable+polyominoes&source=bl&ots=8RIsObpC2O&sig=bMTYGofqh2IukA89kVELecPjavs&hl=en&ei=8BbBTJjHIIX6lwfHmNDYCQ&sa=X&oi=book_result&ct=result&resnum=2&ved=0CBUQ6AEwAQ#v=onepage&q=non-computable%20polyominoes&f=false
Here is the link to non-computable functions:
arxiv.org/abs/math/0406416
and
www-formal.stanford.edu/jmc/basis1/node6.html
and links contained therein.

I should add that I don’t have a dog in this fight with respect to religious belief. I can imagine “that great programmer in the sky” as easily as any Creator God, so my problem with “It from bit” is not a theological one, but the following. I’m trying to see if mentality–mind can be understood on a physical basis. At the moment I’m not sure it can.

 

[Touchstone]

“Shadows of the Mind” is more expanded than “The Emperor’s New Mind” (ENM) and it answers critiques raised against the arguments in ENM. In Shadows Penrose has an extended discussion of why mathematical intelligence and insight is a non-computable process, a much more extended discussion than in ENM. The Halting Theorem business is in fact, not his only argument against computable awareness.
Now I won’t deny that digital computing can emulate many aspects of human intelligence, whether from top-down or bottom-up type programming (e.g. for the former, chess-playing and for the latter medical diagnoses of images via neural network training). Nevertheless, if there is only one mathematical procedure that can’t be done computationally IN PRINCIPLE (not because of present day limitations on computing power), then that indicates to me that there are things–concepts, thought processes–that humans can do that computers will not be able to do and that therefore the universe (which includes humans?) can not be computable.
And I’m not necessarily using the argument about formal systems, which Penrose uses to say that humans can do math that a computer will never be able to do.
Here is the specific URL for the polyominoes (taken from Shadows of the Mind):
books.google.com/books?id=gDbOAK89tmcC&pg=PA30&lpg=PA30&dq=non-computable+polyominoes&source=bl&ots=8RIsObpC2O&sig=bMTYGofqh2IukA89kVELecPjavs&hl=en&ei=8BbBTJjHIIX6lwfHmNDYCQ&sa=X&oi=book_result&ct=result&resnum=2&ved=0CBUQ6AEwAQ#v=onepage&q=non-computable%20polyominoes&f=false
Here is the link to non-computable functions:
arxiv.org/abs/math/0406416
and
www-formal.stanford.edu/jmc/basis1/node6.html
and links contained therein.

I should add that I don’t have a dog in this fight with respect to religious belief. I can imagine “that great programmer in the sky” as easily as any Creator God, so my problem with “It from bit” is not a theological one, but the following. I’m trying to see if mentality–mind can be understood on a physical basis. At the moment I’m not sure it can.

Thanks for the links – I think our posts crossed each other in transit, I found the book link, too. For my part, I don’t know if mind can be virtualized or made physically computable, either. But my reservations may be somewhat different; I don’t have a working definition of “mind” that is discrete and robust enough to put to the test.

Or, how would I know if mind could be virtualized in computing machinery? How would I know if I succeeded? The various flavors of the Turing Test, as impressive as they have been in terms of framing the problem and isolating basic issues, are likely not adequate for a great many people.

If not that, though, what would you (or Penrose) accept as a dispositive test?

I can’t say at the moment, and don’t know that any one can. Which is why claims of “can’t be done”, especially when invoking Gödel (??), seem unreasonably premature, reliance on an invincible intuition, if not outright dogmatic.

Do you have a test or criterion in mind that would be compelling for you on this question? What would cause you to agree, or not, with the idea of a ‘computable consciousness’?

I think ‘a working example’ would be my high-level answer, but as I said, I couldn’t give a detailed breakdown of what specifically would satisfy that request right now. We’re a lot closer than we were just 10 years ago, but the problem of “what is cognition, specifically” is still an obstacle that prevents us from getting to your question. That will have to wait, I think, until we make much more progress on what we mean by “consciousness” in a precise, technical, non-fluffy way.

-TS

 

[Anselm33]

Thanks for the links – I think our posts crossed each other in transit, I found the book link, too. For my part, I don’t know if mind can be virtualized or made physically computable, either. But my reservations may be somewhat different; I don’t have a working definition of “mind” that is discrete and robust enough to put to the test.

Or, how would I know if mind could be virtualized in computing machinery? How would I know if I succeeded? The various flavors of the Turing Test, as impressive as they have been in terms of framing the problem and isolating basic issues, are likely not adequate for a great many people.

If not that, though, what would you (or Penrose) accept as a dispositive test?

I can’t say at the moment, and don’t know that any one can. Which is why claims of “can’t be done”, especially when invoking Gödel (??), seem unreasonably premature, reliance on an invincible intuition, if not outright dogmatic.

**Do you have a test or criterion in mind that would be compelling for you on this question? What would cause you to agree, or not, with the idea of a ‘computable consciousness’?

I think ‘a working example’ would be my high-level answer, but as I said, I couldn’t give a detailed breakdown of what specifically would satisfy that request right now. We’re a lot closer than we were just 10 years ago, but the problem of “what is cognition, specifically” is still an obstacle that prevents us from getting to your question. That will have to wait, I think, until we make much more progress on what we mean by “consciousness” in a precise, technical, non-fluffy way.**
-TS

( emphasis added).

I think the primary capability for mentality (not necessarily intelligence) is self-awareness. How do we determine that? I’m not sure. There are other aspects that would also speak to mentality (again, not necessarily intelligence): awareness of values, goodness, beauty, etc., without having specific case-by-case instructions programmed in (i.e. what we Catholics call Natural Law, presumably, except for psychopaths and sociopaths, built into the human mind, just like the gene for language–see Francis Collins).

It has always seemed to me, as a father of five and grandfather of 11, that watching a baby develop intelligence, mind, etc. should give some clue as to what intelligence and mind are all about, both for normal and for pathological studies. The only such studies I"m aware of are those by Piaget, and they deal, as I recall, mainly with how various categories of intelligence develop.

By the way, there seem to be sexual differences in types of intelligence, despite what the PC crowd say: I recall an article from long ago in which a picture of a tilted glass of water was shown with the air-water interface at the top parallel to the bottom of the glass; some large percentage of female viewers did not see anything wrong with the picture, whereas almost all the male viewers did (they were college freshmen…strike that…freshpersons).
I don’t think the Turing test would work…it’s too easy to program. (I’ll give that anecdote later).

 

[Anselm33]

I don’t think the Turing test would work…it’s too easy to program. (I’ll give that anecdote later).

Here’s software that emulates a psychotherapist; sample dialog is given below:
Computer program in UPPER CASE; patient in lower case.

HELLO. I AM THE DOCTOR. PLEASE TELL ME ABOUT YOUR PROBLEM. TYPE IN COMPLETE SENTENCES AND ALWAYS END YOUR SENTENCES WITH A PERIOD.

My mother makes me so mad sometimes.

CAN YOU GIVE ME AN EXAMPLE?

She phoned me and expected me to drop everything.

TELL ME ABOUT YOUR FAMILY.

I hate my parents.

WHY DO YOU HATE YOUR PARENTS?

They won’t let me go.

IS IT IMPORTANT THAT THEY WON’T LET YOU GO?

I’m suffocating.

(and so forth).

This computer program was either at Stanford or MIT (don’t remember which) and the story goes that a visiting Russian software expert was introduced to the program and started telling it (the program) in great detail about his marital difficulties, how his wife was unfaithful, how he couldn’t satisfy her, etc…to the great embarrassment of his hosts… So there is your Turning Test, passed successfully, but I don’t think that program/computer had self-awareness. (Ot maybe the Russian was having it on–private joke–his hosts? hadn’t thought of that.)

 

[Touchstone]

( emphasis added).
I think the primary capability for mentality (not necessarily intelligence) is self-awareness. How do we determine that? I’m not sure. There are other aspects that would also speak to mentality (again, not necessarily intelligence): awareness of values, goodness, beauty, etc., without having specific case-by-case instructions programmed in (i.e. what we Catholics call Natural Law, presumably, except for psychopaths and sociopaths, built into the human mind, just like the gene for language–see Francis Collins).

As you’ve likely considered in this, those “other aspects” implicate a wider scope than just brain/mind. One of the more recently appreciated features of cognition is that reasoning and computation (biological version) are not nearly exhaustive or sufficient for any working model of mind. There’s a good body of evidence out there now and some fairly shrewd reasoning on it that says that for a moral judgment, for example, we “pre-decide” in a sub-conscious, sub-rational way based on emotions, based on emotional and subliminal responses, responses that are tied up in physiological reactions, and the “rational” part of the judgment is “rationalization” in the cynical sense.

That is, we “decide”, subliminally what our answer or response is going to be, and the “reasoning” part is just so much working backwards to justify after the fact what was a given before it hit the frontal lobes, so to speak.

Without committing on way or another on that idea, what’s important about that hypothesis is the radical ramifications it has for any computational model of mind. On that model, roughly, you need a softtware “thinker/reasoner” that just cleans up and does the filling in for a “feeler”, non-rational/sub-rational responsive system that depends on emotional (name removed by moderator)uts and outputs and various sensory stimuli. In this model, the computation becomes, literally, a matter of afterthought, and the real challenge, in terms of virtual machinery, is some kind of facsimile of humans in the non-rational/sub-rational processing, because on that view, the the “thinker” is the tail, and the “feeler” is the dog wagging that tail.

All of which to say that it’s increasingly seen as naive this notion of “brain-distinct-from-body”, of mind apart from natural person. If you want to build a virtual human brain, you won’t get anywhere satisfying until you build a whole personal body around it, with all the fantastically complex loops and interactions that has. Or, to put it another way, if you took a human mind out of context and analyzed it as 'just a mind", it would be as pathetic and crude as our existing software-based neural networks.

It has always seemed to me, as a father of five and grandfather of 11, that watching a baby develop intelligence, mind, etc. should give some clue as to what intelligence and mind are all about, both for normal and for pathological studies. The only such studies I"m aware of are those by Piaget, and they deal, as I recall, mainly with how various categories of intelligence develop.

In Marc Hauser’s Moral Minds, Hauser recounts and reviews a lot of the kind of applied research I think you are interested in here. This isn’t Hauser’s particular area, so he’s “curating” the research of others, but it’s a pretty good scan. Hauser’s earlier Evolution of Communication is also good on this subject, but as someone with a good friend who works in child neurology, that’s a big, complicated field of science and medicine. Definitely an area where it’s difficult to make smart judgments from the outside.

By the way, there seem to be sexual differences in types of intelligence, despite what the PC crowd say: I recall an article from long ago in which a picture of a tilted glass of water was shown with the air-water interface at the top parallel to the bottom of the glass; some large percentage of female viewers did not see anything wrong with the picture, whereas almost all the male viewers did (they were college freshmen…strike that…freshpersons).

Yeah, those differences aren’t scientifically controversial – it is what it is. It’s fodder for all sorts of politics, of course. But that’s a good to raise, not just because of the challenge of now emulating “male mind” or “female mind”, but the idea that just within humans, different “formats” of mind and cognition exist makes the computing model all the more difficult as emulation, but at the same time suggests that “mind” is plastic in some sense, and can be rendered in different ways. It’s not a “cosmic substance” of some particular constitution, that suggests to me.

I don’t think the Turing test would work…it’s too easy to program. (I’ll give that anecdote later).

OK, thanks.

-TS

 

[Anselm33]

As you’ve likely considered in this, those “other aspects” implicate a wider scope than just brain/mind. One of the more recently appreciated features of cognition is that reasoning and computation (biological version) are not nearly exhaustive or sufficient for any working model of mind. There’s a good body of evidence out there now and some fairly shrewd reasoning on it that says that for a moral judgment, for example, we “pre-decide” in a sub-conscious, sub-rational way based on emotions, based on emotional and subliminal responses, responses that are tied up in physiological reactions, and the “rational” part of the judgment is “rationalization” in the cynical sense.

That is, we “decide”, subliminally what our answer or response is going to be, and the “reasoning” part is just so much working backwards to justify after the fact what was a given before it hit the frontal lobes, so to speak.

Without committing on way or another on that idea, what’s important about that hypothesis is the radical ramifications it has for any computational model of mind. On that model, roughly, you need a softtware “thinker/reasoner” that just cleans up and does the filling in for a “feeler”, non-rational/sub-rational responsive system that depends on emotional (name removed by moderator)uts and outputs and various sensory stimuli. In this model, the computation becomes, literally, a matter of afterthought, and the real challenge, in terms of virtual machinery, is some kind of facsimile of humans in the non-rational/sub-rational processing, because on that view, the the “thinker” is the tail, and the “feeler” is the dog wagging that tail.

All of which to say that it’s increasingly seen as naive this notion of “brain-distinct-from-body”, of mind apart from natural person. If you want to build a virtual human brain, you won’t get anywhere satisfying until you build a whole personal body around it, with all the fantastically complex loops and interactions that has. Or, to put it another way, if you took a human mind out of context and analyzed it as 'just a mind", it would be as pathetic and crude as our existing software-based neural networks.

In Marc Hauser’s Moral Minds, Hauser recounts and reviews a lot of the kind of applied research I think you are interested in here. This isn’t Hauser’s particular area, so he’s “curating” the research of others, but it’s a pretty good scan. Hauser’s earlier Evolution of Communication is also good on this subject, but as someone with a good friend who works in child neurology, that’s a big, complicated field of science and medicine. Definitely an area where it’s difficult to make smart judgments from the outside.

Yeah, those differences aren’t scientifically controversial – it is what it is. It’s fodder for all sorts of politics, of course. But that’s a good to raise, not just because of the challenge of now emulating “male mind” or “female mind”, but the idea that just within humans, different “formats” of mind and cognition exist makes the computing model all the more difficult as emulation, but at the same time suggests that “mind” is plastic in some sense, and can be rendered in different ways. It’s not a “cosmic substance” of some particular constitution, that suggests to me.

OK, thanks.

-TS

I think we’re agreed that it’s a mystery (at this time). Bucknell University Library has “Moral Minds” so I’ll check it out… get to keep it for an academic year. Thanks for the reference.

 

[Touchstone]

Here’s software that emulates a psychotherapist; sample dialog is given below:
Computer program in UPPER CASE; patient in lower case.

HELLO. I AM THE DOCTOR. PLEASE TELL ME ABOUT YOUR PROBLEM. TYPE IN COMPLETE SENTENCES AND ALWAYS END YOUR SENTENCES WITH A PERIOD.

My mother makes me so mad sometimes.

CAN YOU GIVE ME AN EXAMPLE?

She phoned me and expected me to drop everything.

TELL ME ABOUT YOUR FAMILY.

I hate my parents.

WHY DO YOU HATE YOUR PARENTS?

They won’t let me go.

IS IT IMPORTANT THAT THEY WON’T LET YOU GO?

I’m suffocating.

(and so forth).

This computer program was either at Stanford or MIT (don’t remember which) and the story goes that a visiting Russian software expert was introduced to the program and started telling it (the program) in great detail about his marital difficulties, how his wife was unfaithful, how he couldn’t satisfy her, etc…to the great embarrassment of his hosts… So there is your Turning Test, passed successfully, but I don’t think that program/computer had self-awareness. (Ot maybe the Russian was having it on–private joke–his hosts? hadn’t thought of that.)

OK, but this is a kind of tinker-toy version of the Turing Test, and on more sophisticated implementations, there are no programs available right now that can reliably beat a human investigator intent on discerning the machines talking back and the humans.

But even so, the Turing Test is an important evolutionary step in thinking about AI, but it’s hopelessly steeped in naive behaviorism, and artifact of the reigning psychology of his day, and this severely limits if not nullifies the Turing Test as the benchmark for strong AI, or weak AI, for that matter. It’s a very good stone to throw in the pond to start the conversation, though. Would you accept the passing of a strong Turing Test (not something simplistic like the above)? OK, well, what then?

It’s hard to see how to progress until there’s clarity on the benchmark the “measure of mind”.

-TS

 

[Anselm33]

Would you accept the passing of a strong Turing Test (not something simplistic like the above)? OK, well, what then?

It’s hard to see how to progress until there’s clarity on the benchmark the “measure of mind”.

-TS

Well, all the Asimov robot stories essentially give a humanoid robot (who does pass sophisticated type Turing tests, many times).

The “measure of mind” has various constructions by philosophers, AI experts, etc. I like Penrose’s discussion with respect to mathematical intelligence–he claims that mathematicians “discover” mathematical truths (located in a Platonic world of ideal truths.) Maybe the Mysterians, Chalmer et al, are right. At this stage I myself am still on a very steep learning curve, so outside of vague statements, like those in posts above about value judgments–good, beautiful, funny–I couldn’t presume to make a definition.

But to the original point, if it is conceded that aspects of mentality are non-computable, then it would follow that the Universe (upper case) is not computable… Would that be a correct inference?
and again thanks for all your comments.

 

[sinnerdexter]

OK, so I ask, if these functions are non-computable in principle, how can the Universe be computable?

I await enlightenment.

anselm

(logic of dominoes)
Everything have an general algorithm in our universe.Everything is connected (no free willing).If somthng have no connections than its not existed.So one part of universe can be calculated by other part but the whole universe can not be calculated by itself.

If you want to calculate our universe with a bigger one than they have become a universe because they interacted.

Its like you have a x86 p4 3ghz processor that simulates a pic.But you can not ask it to simulate a x86 p4 3ghz processor.

Hevvy stuff

 

[Anselm33]

(logic of dominoes)
Everything have an general algorithm in our universe.Everything is connected (no free willing).If somthng have no connections than its not existed.So one part of universe can be calculated by other part but the whole universe can not be calculated by itself.

If you want to calculate our universe with a bigger one than they have become a universe because they interacted.

Its like you have a x86 p4 3ghz processor that simulates a pic.But you can not ask it to simulate a x86 p4 3ghz processor.

Hevvy stuff

Interesting and reasonable notion. However, if one concedes (which I don’t necessarily) that the Universe (big U!) is computable, then that which is outside the Universe (i.e. God) is that which is the programmer.

 

[Touchstone]

Well, all the Asimov robot stories essentially give a humanoid robot (who does pass sophisticated type Turing tests, many times).

The “measure of mind” has various constructions by philosophers, AI experts, etc. I like Penrose’s discussion with respect to mathematical intelligence–he claims that mathematicians “discover” mathematical truths (located in a Platonic world of ideal truths.) Maybe the Mysterians, Chalmer et al, are right. At this stage I myself am still on a very steep learning curve, so outside of vague statements, like those in posts above about value judgments–good, beautiful, funny–I couldn’t presume to make a definition.

But to the original point, if it is conceded that aspects of mentality are non-computable, then it would follow that the Universe (upper case) is not computable… Would that be a correct inference?
and again thanks for all your comments.

I think that would be the unavoidable conclusion. If we understand minds to be a) non-computable and b)part of the universe then the universe would be necessarily non-computable, for there are non-computable minds in it.

-TS

 

[In_Spiration]

J.R. Lucas’s “Minds, Machines and Godel” may interest you. Here.