• TheMadFool
    2.4k
    I would really like to hear more on what you mean by numbers being passive and active. I'm currently struggling to get my thoughts together on this, but I think we might have very similar views.Jerry

    Active numbers have causal import. Numbers can cause things to happen. Passive numbers are, very roughly speaking, the effects of active numbers and can be read like a book. My idea of numbers is very rudimentary.
  • TheMadFool
    2.4k
    I couldn't answer the OP's question directly because it depends on the meaning of "exist". Too much speculation - healthy, yes, but not verifiable in a practical sense.
  • fishfry
    469
    Maths is unreasonably effective. It’s abstractions are more than mere intellectual accidents. There must be a reason for their Platonic seeming necessity. So therefore that is why the nature of mathematical truth remains so central to physicalist inquiry.apokrisis

    Agreed.

    Neuroscience believes thoughts to be informational processes, not biochemical ones.apokrisis

    Of course nothing of the sort is true, but I see where you're going in the next couple of paragraphs so this is not a crucial point.

    What's true is this. Computationalism s the claim that the mind (or the universe, in a more grandiose version) is a computation. Now those neuroscientists who are computationalists believe that thoughts are informational processes; and those who aren't, don't.

    I hope you will agree with me that this is a true statement about the states of belief of neuroscientists, and that this is NOT a settled issue by any means. If nothing else, if mind is a computation, what's the algorithm? When you bring me some computer code and say, "Here, this is how you implement an mind. It's 875,356 of C++. Some grad student figured it out," then maybe I'll believe you. Till then, the burden of proof is on you.

    It's fine if you want to argue from a computationalist point of view. Just so you don't claim that it's the only point of view and that the neuroscientists have all agreed on it. That last part isn't true.

    To use the easily abused computational analogy, the "material physics" explains nothing. You could implement the logic of a Turing machine in some system of tin cans and bits of twine.apokrisis

    Yes I understand substrate independence. Any Turing machine computes the same thing whether you implement it with pencil and paper or on a supercomputer or in the wetware of the brain. So if you believe that mind is a computation, then mind is a TM executed in the hardware of the brain. However if you DON'T believe that mind is a computation, you no longer necessarily have substrate independence. I hope you would grant me this. Searle makes the same point. He's said that he thinks mind is a physical but not computational aspect of the brain. I agree with him about that.

    So a science of the mind definitely does need a dualist physicalism of some kind. There has to be some ontic difference between information and entropy, even if they also arise in some common (mutual) fashion.apokrisis

    You are describing the distinction in computer science between a program and a computation. Take the Euclidean algorithm to find the greatest common divisor of two integers. By itself, it does nothing. It csn not find the GCD of two integers. The only way to do that is to execute the algorithm on physical hardware. That is a physical process involving an input of energy and an output of heat. Something a physicist could observe and quantify.

    So yes we always have that dualism. Where does the algorithm itself live? Well it lived first in Euclid's brain. But isn't Euclid's mind a physical process? His abstract thoughts are physical processes, and his thoughts can be implemented as physical processes. But I don't see why we need dualism.


    But putting that aside, the issue here is the epistemic one of a distinction between observers and observables. Classical physics just presumes that observers are free agents, able to make measurements of reality without disturbing that reality. And this supports the idea that thoughts and rocks are unproblematically separate. Not only are our conceptions of reality a free invention of the human mind, but so do our perceptions of reality enjoy a matching freedom from our ability to invent.apokrisis

    I would say that I think this is a little off-target in the sense that it's one complication too many. If we try to weave quantum theory into this we may lose our way. It's hard enough to try to pin down the difference between the abstract and the physical. So I'm not going to try to think about this. You have to start somewhere, and perhaps we could agree that for purposes of this conversation, there is the number pi and there is a rock, and that we don't have to consider their quantum relationship to each other, if any.


    This epistemic shock doesn't seem to have registered with the mathematical community as far as I can see.apokrisis

    I don't see why it should. When Wiles proved FLT, he didn't say to himself, "Well, quantum theory says that the integers are just like rocks." Why would this come up? Mathematicians do math. Some mathematicians sometimes do philosophy. But when mathematicians do philosophy they're acting as philosophers, not as mathematicians. Wiles doesn't sit around thinking about the nature of the integers. To a number theories, integers are as real as rocks. I doubt Wiles would agree that he's written a work of fiction. Or even give the matter any thought at all.

    The ontological options are still either that maths is a free invention or a perception of Platonic reality.apokrisis

    Right. Philosophers are rightly concerned with this question. But mathematicians aren't. There is no epistemic shock or even any thoughts at all of philosophy. Some mathematicans care about these things but it's not a requirement of the job.

    It has to conform to the rules of an informational process - the syntax that is grounded in set theory, or category theory, or whatever other fundamental notion of a closed syntactical system happens to be in vogue at the time.apokrisis

    Ooh you are on shaky ground here! Gödel told us that math is NOT an informational process! No algorithm can determine the truth of mathematical statements. First you say mind is a computation and then you say that math is. Well we know for a fact that math is NOT a computational process. Perhaps mind isn't either! You know Penrose has made that argument, that incompleteness shows that mind is not a computation. Nobody takes Penrose's argument seriously, but after all he is Sir Roger and the rest of us aren't.

    I hope you can see that your computationalist bias may be leading you astray. There are important things in the world that are not computations. Like mathematical truth.


    Sanity is not having to think, it appears.apokrisis

    Well you can drive yourself nuts thinking about this too hard. If every single thing in the world needed to make perfectly logical sense, we couldn't get out of bed in the morning. Life does not make sense! Of course a computationalist like yourself would not understand that. You think we're all just a computer program. That's a silly idea. There I said it. I really disagree with computationalism.

    Yes, it is out there as a ratio capturing a primal relation of a physical world with some kind of limit-state perfect symmetry. Let that world be not perfectly flat, let it be non-Euclidean, and the value of pi starts to wander accordingly.apokrisis

    Jeez that sounds a little mystical. You're saying that Euclidean geometry is the midpoint between elliptic and hyperbolic geometry. Yes this is a true mathematical fact, but it is not mystical. It's just pi. There are plenty of other real numbers out there too. Pi's not that big a deal really. I only used it as a familiar example of a number whose physical existence can be argued against. I didn't mean it in a mystical sense.

    Between the hyperbolic and the hyperspheric, there is only one geometry that is absolutely balanced enough that the value of pi is as stable as far as the eye can see. Whether your circles are big or small, now pi remains always the same.apokrisis

    Yes but you're going all woo-woo about a trivial mathematical fact. Well not trivial, non-Euclidean geometry was a big deal when it was discovered. And as Kant noted, we do seem to have an intuition of Euclidean geometry. I'll grant you that. But I think you're making too much of this.

    Also I have a picky little pedantic point. The ratio of a circle's circumference to its diameter is different depending on the geometry. But the number pi is always 3.14159... Pi is a particular real number. It's not defined geometrically these days. If we called the circle ratio the foozle, then you can say that in Euclidean geometry the foozle is pi and in non-Euclidean geometry it's not. But pi is always pi. It's a number. It's like asking what's the value of 3 in hyperbolic space. It's 3.


    So pi pops out of reality, out of nature, not by accident but because the very possibility of a "physical relation" has some emergent invariant limit. It arises out of the broken symmetry that is a perfect orthogonality.apokrisis

    You and Kant. He was wrong. You're wrong. Euclidean geometry's not special. It's just something we seem to have an intuition of. I'll grant you the psychological and philosophical interest of that fact. But not the mathematical importance.

    In short, you are a Euclidean-chauvinist!


    Thus on the one hand, pi - as a position on the number line - looks the purest accident. Why should it have that exact value?apokrisis

    I'll answer that question, if you'll first tell me why 3, a point on the number line by the purist accident, has that exact value. It's a mystery!! Why does 3 have the exact value of 3? It's because it's the number 3. And pi is the number pi. It's just a real number.

    It's true that it's the ratio of a circle's circumference to its diameter is pi, but if it were 3 or 47 or 18, you'd be asking why it's that? It's just what it is. The only really interesting thing is that the ratio is always the same no matter what size the circle is! That's the real breakthrough here, that was a great discovery once. [Edit - You made the point that this is only true in Euclidean geometry. Point taken].

    On the other, pi is the identity relation when it comes to a limit notion of orthogonal dimensionality.apokrisis

    You are really into pi mysticism. What I mean is, what you wrote here is pretty word salad-y. I have to repeat, I only picked pi because it's a good candidate to make the point that numbers are abstract and not physical. I could have made the exact same point with 3, but people have a harder time understanding that 3 isn't any more physical than pi.

    I did NOT intend to inspire any pi mysticism. There is nothing special about pi. It's just a real number. There are plenty of real numbers, many of them interesting for various reasons. It's a matter of historical contingency which ones got discovered first.

    We might as well just give its value as 1. Everything else that is less perfectly broken can be measured as some difference to that.apokrisis

    Yes, linear scaling factors don't matter. If you think of the number line, it doesn't matter what names we give anything. If we called 3 by the name 47, everything else would work out the same. Math wouldn't change.

    * So to sum up:

    - You are arguing from a computationalist point of view, but I'm not sure what point you are trying to make. Looking back I see that now. Even if I agree with you that mind is computation, there are still numbers and rocks. I possibly did not follow your argument.

    - You are wrong that math is a computation. And like many computationlists, you underestimate or ignore the importance of non-computable phenomena in the world. Remember even Tegmark distinguishes between the mathematical universe hypothesis and the computable universe hypothesis. Computationalism is a very strong assumption.

    * Mathematicians do math, not philosophy. My sense is that the vast majority of working mathematicians never give any thought to philosophy. When an engineer is building a bridge, do you want him spending his time contemplating the fact that there is no difference between him and the bridge? Or do you want him calculating the load factors according to state of the art engineering principles?

    Well I hope some of this was on target.
  • apokrisis
    4.5k
    Thanks for the lengthy reply.

    What's true is this. Computationalism s the claim that the mind (or the universe, in a more grandiose version) is a computation. Now those neuroscientists who are computationalists believe that thoughts are informational processes; and those who aren't, don't.

    I hope you will agree with me that this is a true statement about the states of belief of neuroscientists, and that this is NOT a settled issue by any means. If nothing else, if mind is a computation, what's the algorithm? When you bring me some computer code and say, "Here, this is how you implement an mind. It's 875,356 of C++. Some grad student figured it out," then maybe I'll believe you. Till then, the burden of proof is on you.
    fishfry

    I'm definitely not claiming computationalism - or at least not Turing machine computation as you seem to suggest. The mainstream neuroscience view - since Sherrington's "enchanted loom" or Hebbs's learning networks - is some kind of neural net form of "computation".

    And more to the point, it is mainstream to emphasise that the brain is involved in informational activity, not merely biochemical activity. Otherwise why is neuroscience interested in discovering the secrets of the neural code, or brain's processing architecture? It knows the biophysics of what makes a neuron fire. But how that firing then represents or symbolises something with felt meaning is the big question. And that can only be approached in terms of something other than a biochemical materialism. It demands a semiotic or information theoretic framework. Which in turn has already considered Turing computation and found it not the answer.

    So broadly speaking, neuroscientists think thoughts are informational processes and not biochemical events. At the same time, they don't think the brain is literally a Turing machine or programmable computer. That might be a helpful analogy, like calling the eye a camera. But just as quickly, the caveats would begin.

    There are important things in the world that are not computations. Like mathematical truth.fishfry

    Computers are machines. They are devices that construct patterns. So yes, of course, human minds seem to operate in a fundamentally different fashion. We can grasp the whole of some pattern. We can understand it "organically" as a system of constraints, rather than as an atomistic construction.

    Our abductive or intuitive approach to reasoning begins with this ability to see the whole that "stands behind" the part. We can make inferences to the best explanation. And then, having framed an axiom or hypothesis, we are also quite good at deducing consequences and confirming by observation.

    So when it comes to mathematical truth, that is what we think we are doing. We notice something about the world. We then leap towards some rational principle that could "stand behind" this something as its more general constraint.

    Turing machines are really bad at making such a holistic generalisation. Neural network computers are our attempt to build machines that are good at implementing this precise inferential leap.

    However if you DON'T believe that mind is a computation, you no longer necessarily have substrate independence. I hope you would grant me this.fishfry

    Yeah. I don't claim complete substrate independence. But then my "computationalism" is a semiotic or embodied one. The whole point is that it hinges on a separation which then allows an interaction.

    A Turing machine does not self-replicate. A Turing machine does not have to manage its material flows or compete with other TMs. But a living thing is all about regulating its physics with information. So an independence from physical substrate (an epistemic cut) is required by life and mind. But only so as to be able to regulate that physics - bend it in the direction which is making the autopoietic wholeness that is "an organism".

    The only way to do that is to execute the algorithm on physical hardware. That is a physical process involving an input of energy and an output of heat. Something a physicist could observe and quantify.fishfry

    Yes, you can measure one side of the computational story in terms of entropy production. But how do you measure the other side of the story in terms of "negentropy" production? The fact that your computer runs either hotter or colder doesn't say much about whether its eventual output is righter or wronger.

    Where does the algorithm itself live? Well it lived first in Euclid's brain. But isn't Euclid's mind a physical process? His abstract thoughts are physical processes, and his thoughts can be implemented as physical processes. But I don't see why we need dualism.fishfry

    We are labouring the point. If you really can't see the difference between syntax and semantics by now, things are likely hopeless.

    You keep talking about the physical events as if they are the informational processes. Of course a neuron or a transistor or a membrane receptor or a speedometer can be described in terms of their "physics". But it is hardly the level of description that explains "the process" which we are interested in.

    To reduce functional or informational processes to atomistic material events becomes a nonsense. Especially for true computationalism. The only time we are interested in the physics of a logic gate is when it doesn't behave like a logic gate - that is when it has some uncontrolled physical process going on.

    So algorithms are extreme mechanistic dualism in fact. You don't even have to run a programme for it to "have a result". The result could only be different if the physics of the real world somehow intruded, And then we would say the computer had a bug. It over-heated or something.

    And maths is kind of like that. We imagine it as transcendent and eternal truths - things that would be true without ever needing the reality of physical instantiation. Pure information. It is crazy to talk of Euclidean maths as existing in some geezer's long dead brain.

    Jeez that sounds a little mystical. You're saying that Euclidean geometry is the midpoint between elliptic and hyperbolic geometry. Yes this is a true mathematical fact, but it is not mystical.fishfry

    Why do you interpret that as a mystical statement? My point was that it is not a mystery because it is what you would expect from principles of physicalist symmetry. If every kind of difference gets cancelled (as the negatives erase the positives) then what you are left with is the mid-point balance. It would be natural to expect "flatness" as the emergent limit state.

    So I'm not going to try to think about this. You have to start somewhere, and perhaps we could agree that for purposes of this conversation, there is the number pi and there is a rock, and that we don't have to consider their quantum relationship to each other, if any.fishfry

    Well it is your choice to ignore what we know to be fundamental in preference for what we know to be emergent.

    I can't agree that it makes for good metaphysics. And I think you just want to avoid having to make a better argument.

    To a number theories, integers are as real as rocks. I doubt Wiles would agree that he's written a work of fiction. Or even give the matter any thought at all.fishfry

    Fine. The philosophical issue here is not the pragmatics of mathematical research. And I even agree that mathematical research - in being an informational theoretic exercise - would deliberately insulate itself from such fundamental metaphysical issues. Maths doesn't really want to even concern itself with geometry - the physical constraints of space - let alone with actual materiality, or the constraints of energy, the possibilities of change. So - as institutional habit - integers are as real as rocks.

    Except they are then ... ideas? Constructs? Thoughts in the head?

    You seem to want it both ways. And that winds up in Platonism.

    That is why my own position is the semiotic one where the integers are the ideal limits on materiality. That is a formula of words that both accepts a strong difference and a strong connection between the two sides of the semiotic equation. Information is real if it is causal. And being an actual limit on material freedom is pretty clearly causal.

    Ooh you are on shaky ground here! Gödel told us that math is NOT an informational process! No algorithm can determine the truth of mathematical statements.fishfry

    See earlier where I spoke about abductive reasoning and our ability to make inferential leaps. Gödel validates my approach here. The failure of logical atomism is the solid ground for the holist. It is why a semiotic approach to reality is justified.

    Yes but you're going all woo-woo about a trivial mathematical fact. Well not trivial, non-Euclidean geometry was a big deal when it was discovered.fishfry

    You mentioned pi. I am just highlighting how the usual woo-woo aspect - the fact that there is just this "one number" picked at random out of all the numbers on the number-line - masks a bigger story. The woo-woo evaporates when you see there is a "material" process that picks out a value for "being flat". Two kinds of possible curvature had a mid-point balance. Pi is a number that emerges due to something more holistic going on. The fact that it emerges "right there" on the number-line is not some kind of weird magic.

    It is even easier to see with other constant like e that are directly derived from growth processes. There the contrasting actions that produce the emergent ratio are in plain sight. It is funny that e should be 2.71828. But then that becomes obvious when it is realised that growth always has to start from some thing that is just itself 1. There is no reason to think of e as anything but natural after that.

    You and Kant. He was wrong. You're wrong. Euclidean geometry's not special. It's just something we seem to have an intuition of.fishfry

    But I am not Kantian, except in a loose sense. I'm Peircean in the way Peirce fixed Kant.

    And I'm arguing flatness is special as the mid-point of opposing extremes of curvature. It has physically important properties too. Only flat geometries preserve invariance under transformations of scale. That is a really important emergent property when it comes to things like Universes.

    It's true that it's the ratio of a circle's circumference to its diameter is pi, but if it were 3 or 47 or 18, you'd be asking why it's that? It's just what it is. The only really interesting thing is that the ratio is always the same no matter what size the circle is! That's the real breakthrough here, that was a great discovery once. [Edit - You made the point that this is only true in Euclidean geometry. Point taken].fishfry

    And as I repeat, it is very important metaphysically that absolute scale invariance only appears at a particular numeric value of pi. That is how a Universe is even possible.

    So you are focused on the triviality of pi being given some particular position on the number line - look guys, its 3.141592653589793238462643383279502884197169399375105820974944592307816406286 208998628034825342117067982148086513282306647093844609550582231725359408 ...

    And that is what makes folk go woo. It seems both weirdly specific and weirdly random. There seems no natural reason for the value.

    But it's a ratio derived from the radius being granted as the natural unit. Let's call the radius 1. Let's get a grip on this weird thing called curvature by starting with the "most natural part of the story" - a line segment. That gets to be "1" on the number-line.

    Well, as I say, once mathematicians woke up to the fact that flatness was a rather special case of curvature, and once physicists in turn woke up to the fact that scale invariance was essential to any kind of workable Universe (its called rather grandly the cosmological principle), well, maybe it is the ratio that should be called "1". A straight line segment is only a natural unit in the context of an already flat space which supports unlimited scale transformations. It depends on the emergent fact of parallel lines or infinite rays being an actual possibility.

    You are really into pi mysticism. What I mean is, what you wrote here is pretty word salad-y. I have to repeat, I only picked pi because it's a good candidate to make the point that numbers are abstract and not physical. I could have made the exact same point with 3, but people have a harder time understanding that 3 isn't any more physical than pi.fishfry

    I am being anti-mystical in pointing out the very physical basis of pi as a number. It is a ratio that picks out a critical geometric balance.

    The number 3 is trivial by comparison. Well there are physical arguments for why the geometry of universes are optimal if they have just three orthogonal spatial directions. But 3 as a member of the integers has no numeric specialness by design. The special or natural numbers are 1 and 0. We see this in the symmetries captured by identity operations. There is something basic or universal when we hit the bedrock that is a symmetry or invariance.

    You would call it a mystical fact perhaps. I see it as quite reasonable and self-explanatory.

    * So to sum up:

    - You are arguing from a computationalist point of view, but I'm not sure what point you are trying to make. Looking back I see that now. Even if I agree with you that mind is computation, there are still numbers and rocks. I possibly did not follow your argument.
    fishfry

    Nope. At least not your notion of computation as Turing machine/programmable computation.

    I take an information theoretic perspective. And more specifically, a semiotic one. In technology terms, neural networks come the closest to implementing that notion of computation.

    And numbers vs rocks is a distinction that relies on a classical metaphysics - one in which the divide between observers and observables does not present an epistemic difficulty. The epistemic cut - the necessary separation of the information from the physics - can be treated as an ontological fact.

    So my positions on both "mind is a computation" and "reality is classical" are the same. Semiotics starts from the view that there is no fundamental ontic division of observers and observables. But that is also the division which must emerge via some epistemic cut. It is the basis of intelligibility. And even the Universe can only exist to the degree it hangs together in intelligible fashion.

    Hence why maths tends to be unreasonably effective at describing the Universe. Or being in general.

    - You are wrong that math is a computation. And like many computationlists, you underestimate or ignore the importance of non-computable phenomena in the world. Remember even Tegmark distinguishes between the mathematical universe hypothesis and the computable universe hypothesis. Computationalism is a very strong assumption.fishfry

    Labouring the point still, but I'm sorry. I'm not a computationalist in the sense you are hoping for. Indeed, that was what I was accusing you of. You seem to believe reality is a machine. An account of physical events is sufficient.

    But yes, you also seem to say the opposite. This is a symptom that your metaphysics is "commonsensical" and not well thought out.

    * Mathematicians do math, not philosophy. My sense is that the vast majority of working mathematicians never give any thought to philosophy. When an engineer is building a bridge, do you want him spending his time contemplating the fact that there is no difference between him and the bridge? Or do you want him calculating the load factors according to state of the art engineering principles?fishfry

    Again, bully for mathematicians. Bully for engineers. Bully even for most physicists (as very few are employed in frontier theory construction).

    But it is curious to be complaining about metaphysics where metaphysics is appropriate.

    And so far you haven't put forward any clear exposition of your own epistemic position, let alone given a clear justification for it. You just hoped to be able to label me with some obviously weak ontology that I spend most of my time arguing against.
  • fishfry
    469
    Much of your argument centers around your belief that neural networks are a different mode of computation than Turing machines. I do not believe you are correct but there's a fair amount of confusion of this point online. Before replying to your specific points I'll lay out my understanding, and if you or anyone else can clarify or amend my thoughts, please do.

    * First, there are Turing machines (TMs). The Church-Turing thesis says that anything we can compute, can be computed by a TM. This is an eighty year old core idea in computer science that has never been refuted. If tomorrow morning professor so-and-so in Helsinki publishes a paper called, "A mode of computation that's not a TM," it would rock the computer science world and it would make the popular media. "80 year old computer theory debunked," etc.

    This hasn't happened. As far as anyone in the world knows, everything that we would call a computation can be implemented as a TM.

    * Real-world neural nets are TMs. This must be true; if not, Church-Turing would be broken and we'd all have heard about it.

    What is a neural net (NN)? It consists of a set of nodes, each assigned a numeric weight. Then we apply some logic: If the weight of this node is such and so and its immediate neighbors are such and so and their neighbors are such and so, then do something. This is perfectly conventional programming. The greatest weak AI in the world, such as AlphaGo Zero, is a conventional computer program implemented on conventional hardware. A real-world implementation of a TM.

    To be sure, neural nets are very clever ways go organize a conventional computation. But they are conventional computations nonetheless.

    * Theorestical neural nets. In the abstract model, the numeric weights of the nodes can be real numbers. Since in general it takes an infinite amount of information to specify a real number, there are no real-world implementations of theoretical NNs. I'm not aware of the theory of computation behind NN and how they relate to Church-Turing.

    * Wetware NNs such as the brain. Since brains are physical, even if they are NN's, they are TMs. Moreover, the idea that the mind implemented by the brain is a NN is a speculative idea. Nobody has proof. There's no reason to believe a mind/brain is a TM and plenty of reasons to doubt it.

    I'm definitely not claiming computationalism - or at least not Turing machine computation as you seem to suggest. The mainstream neuroscience view - since Sherrington's "enchanted loom" or Hebbs's learning networks - is some kind of neural net form of "computation".apokrisis

    Any physical implmentation of a NN is a TM. If you disagree then you either believe the Church-Turing thesis has been falsified (which it hasn't) or that the brain contains nodes that can represent arbitrary real numbers (absurd) or you have some other justification for your claim. Please provide such justification.

    As far as what constitutes mainstream neuroscience, I'm not qualified to judge. I hope you would agree that the opinions of neuroscientists donot constitute a refutation of Church-Turing, merely ignorance of it.


    And more to the point, it is mainstream to emphasise that the brain is involved in informational activity, not merely biochemical activity.apokrisis

    It may be mainstream speculation, but it is certainly not mainstream established fact. But again, why are you so hung up on the opinions of neuroscientists? Since any real-world NN must be implemented as a TM, the burden is on you to explain yourself.


    Otherwise why is neuroscience interested in discovering the secrets of the neural code, or brain's processing architecture?apokrisis

    How does that prove a real-world NN isn't a TM? You are committed to your argumentum ad populum but you don't seem to be able to reason on your own.

    It knows the biophysics of what makes a neuron fire. But how that firing then represents or symbolises something with felt meaning is the big question.apokrisis

    Oh, I thought it was a computer program as you keep claiming. Or a NN, which is just a particular kind of computer program. Now you admit that we DON'T understand how firing neurons give rise to mind. Well then at last we agree.


    And that can only be approached in terms of something other than a biochemical materialism.apokrisis

    Why?

    It demands a semiotic or information theoretic framework.apokrisis

    Why?

    Which in turn has already considered Turing computation and found it not the answer.apokrisis

    Funny you should say that, since many people (wrongly) believe the mind is literally a TM computation. But if the mind is ANY kind of computation, you have to explain how it could be a computation yet not be a TM. This point does not seem to be appreciated in the literature


    So broadly speaking, neuroscientists think thoughts are informational processes and not biochemical events.apokrisis

    Some do, some don't. Some scientists used to think heat was caused by phlogiston. What of it? You are continually trying to substitute claims about the opinions of some neuroscientists for thinking things through on your own.


    At the same time, they don't think the brain is literally a Turing machine or programmable computer. That might be a helpful analogy, like calling the eye a camera. But just as quickly, the caveats would begin.apokrisis

    I'm glad that you agree with me on at least this point. The problem is that there is no other mode of computation as far as we know.


    Computers are machines. They are devices that construct patterns. So yes, of course, human minds seem to operate in a fundamentally different fashion. We can grasp the whole of some pattern. We can understand it "organically" as a system of constraints, rather than as an atomistic construction.apokrisis

    If you agree with me why do you keep trying to disagree? I have no idea what your point is. You go back and forth on your own opinion.

    Our abductive or intuitive approach to reasoning begins with this ability to see the whole that "stands behind" the part. We can make inferences to the best explanation. And then, having framed an axiom or hypothesis, we are also quite good at deducing consequences and confirming by observation.apokrisis

    Yes. Which neither confirms nor denies that mind is a computation, since even the weak AI's are quite impressive these days in seeing the whole, as in facial recognition.

    So when it comes to mathematical truth, that is what we think we are doing. We notice something about the world. We then leap towards some rational principle that could "stand behind" this something as its more general constraint.apokrisis

    You are eloquently agreeing with my point.

    Turing machines are really bad at making such a holistic generalisation.apokrisis

    These days, strangely and counterintuitively, TMs are incredibly good at generalization and "gestalt," at least in constrained domains. AlphaGo Zero is mind-blowing in its philosophical implications and AlphaGo Zero is a TM.

    Neural network computers are our attempt to build machines that are good at implementing this precise inferential leap.apokrisis

    Agreed. NN's are a clever way of organizing a conventional TM. But every NN is implemented as a TM. They're computer programs implemented on conventional hardware. Please tell me you understand this point. There are no magic NN computers. They're NN algorithms implemented on TMs.

    Yeah. I don't claim complete substrate independence. But then my "computationalism" is a semiotic or embodied one. The whole point is that it hinges on a separation which then allows an interaction.apokrisis

    If it's a computation then it's a TM. You need to deal with this point.

    A Turing machine does not self-replicate.apokrisis

    Neither does a person without children. What does that have to do with the subject at hand? Red herring.


    A Turing machine does not have to manage its material flows or compete with other TMs.apokrisis

    Ever hear of core wars? Oldtime hackers used to write programs that would compete with each other for machine resources. Of course TMs can be programmed to compete with other TMs.


    But a living thing is all about regulating its physics with information.apokrisis

    You are confusing the issue by bringing up living things. Nothing to do with the subject at hand.


    So an independence from physical substrate (an epistemic cut) is required by life and mind.apokrisis

    I don't see why. Searle believes mind is a function of the physical brain, just not a computational one. You don't need mysticism or duality.

    But only so as to be able to regulate that physics - bend it in the direction which is making the autopoietic wholeness that is "an organism".apokrisis

    Autopoeietic. Whatever. What's that mean? I could look it up but I'd like you to explain this in your own words what your point is. NNs are TMs and if you think the mind is a computation then you think the mind is a TM. You have to deal with that by denying it (with evidence) or accepting it.



    Yes, you can measure one side of the computational story in terms of entropy production. But how do you measure the other side of the story in terms of "negentropy" production? The fact that your computer runs either hotter or colder doesn't say much about whether its eventual output is righter or wronger.apokrisis

    I don't follow the relevance of that para.



    We are labouring the point. If you really can't see the difference between syntax and semantics by now, things are likely hopeless.apokrisis

    If you don't see the difference between fish and bicycles, things are likely hopeless. WTF? You think I don't know the difference between syntax and semantics? You're flailing.


    You keep talking about the physical events as if they are the informational processes.apokrisis

    No no. But informational processes ARE physical events. Running Euclid's algorithm in a supercomputer or with pencil and paper are physical processes [not events]. They require energy and output heat. The description of the algorithm, the program, does not compute anything.


    Of course a neuron or a transistor or a membrane receptor or a speedometer can be described in terms of their "physics". But it is hardly the level of description that explains "the process" which we are interested in.apokrisis

    It doesn't explain mind. It only points out that you don't need duality to explain computation. Computation is a physical process. [And not the converse as you tried to claim I said earlier].

    To reduce functional or informational processes to atomistic material events becomes a nonsense.apokrisis

    Why? Who's the dualist now? What kind of mystical process are you believing in? If my mind is not a function of my physical brain, what do you think it is, exactly? Are you a dualist or not?

    Especially for true computationalism. The only time we are interested in the physics of a logic gate is when it doesn't behave like a logic gate - that is when it has some uncontrolled physical process going on.apokrisis

    Uncontrolled physical process? You know you are not speaking coherently these past few paragraphs. You're flailing randomly.

    So algorithms are extreme mechanistic dualism in fact.apokrisis

    Dualism, why? I program a computer to add 2 + 2, it outputs 4. Where is the dualism? The computer inputs electricity, outputs heat, and performs a computation. I really don't understand your mysticism around this very commonplace and well-understood phenomenon of computation.


    You don't even have to run a programme for it to "have a result".apokrisis

    That's just wrong. If I write down the Euclidean algorithm, it has no result. Only when I implement the program on a physical substrate and execute the algorithm does it produce a result. If you don't understand this there really is nothing to talk about.

    The result could only be different if the physics of the real world somehow intruded, And then we would say the computer had a bug. It over-heated or something.apokrisis

    What of it? You just claimed a program need not be executed to produce a result. That's "not even wrong." Its a profound misunderstanding of computation.


    And maths is kind of like that. We imagine it as transcendent and eternal truths - things that would be true without ever needing the reality of physical instantiation. Pure information.apokrisis

    Ok we're Platonists today. Fair enough. But where do these truths live? You are quite the mystic.

    It is crazy to talk of Euclidean maths as existing in some geezer's long dead brain.apokrisis

    Really? Crazy? That the best you can do in lieu of an actual argument? You haven't made a single rational argument in this entire post. I don't think you have one.


    Why do you interpret that as a mystical statement? My point was that it is not a mystery because it is what you would expect from principles of physicalist symmetry. If every kind of difference gets cancelled (as the negatives erase the positives) then what you are left with is the mid-point balance. It would be natural to expect "flatness" as the emergent limit state.apokrisis

    Nonsense. Mathematical nonsense and physical nonsense. You must have missed the Einstenian revolution. It's not 1900 anymore.



    Well it is your choice to ignore what we know to be fundamental in preference for what we know to be emergent.
    apokrisis

    Ah, emergence. Another murky concept. Hydrogen's not wet and oxygen's not wet but water is wet. Zowie, cosmic.

    I honestly have no idea what you are going on about. I really don't think you are making any sense at all.

    Fine. The philosophical issue here is not the pragmatics of mathematical research. And I even agree that mathematical research - in being an informational theoretic exercise ...apokrisis

    Didn't I already remind you earlier that Gödel disproved that math is an information-theoretic exercise? Why are you doubling down on a claim I've already falsified?


    Maths doesn't really want to even concern itself with geometry - the physical constraints of space - let alone with actual materiality, or the constraints of energy, the possibilities of change. So - as institutional habit - integers are as real as rocks.apokrisis

    You seem to be back in 1840, railing against the great discovery of non-Euclidean geometry. Is that your complaint? That math isn't physics? I'm sure the physicists agree with you.

    Except they are then ... ideas? Constructs? Thoughts in the head?apokrisis

    The nature of mathematical truth is indeed an open question.

    You seem to want it both ways. And that winds up in Platonism.apokrisis

    I just want you to say something that's reasonably on topic and that makes some sort of sense. I have no idea what you're going on about here.


    That is why my own position is the semiotic one where the integers are the ideal limits on materiality.apokrisis

    "the integers are the ideal limits on materiality" -- This is supposed to make logical sense to me? Is this some sort of postmodern theory? I confess I don't take postmodern mathematical musings very seriously. Perhaps you do. If you would take the time to explain what you mean by "the integers are the ideal limits on materiality" then perhaps I'd learn something.


    That is a formula of words that both accepts a strong difference and a strong connection between the two sides of the semiotic equation. Information is real if it is causal. And being an actual limit on material freedom is pretty clearly causal.apokrisis

    You sound like a raving postmodernist. Perhaps you are a postmodernist and I'm insuffiently appreciative of that point of view. That may well be the case.


    See earlier where I spoke about abductive reasoning and our ability to make inferential leaps. Gödel validates my approach here. The failure of logical atomism is the solid ground for the holist. It is why a semiotic approach to reality is justified.apokrisis

    Semiotic. Whatever. Explain yourself clearly if you can. Can you?

    You mentioned pi. I am just highlighting how the usual woo-woo aspect - the fact that there is just this "one number" picked at random out of all the numbers on the number-line - masks a bigger story. The woo-woo evaporates when you see there is a "material" process that picks out a value for "being flat". Two kinds of possible curvature had a mid-point balance. Pi is a number that emerges due to something more holistic going on. The fact that it emerges "right there" on the number-line is not some kind of weird magic.apokrisis

    Tell me something. What is the true value of 3? And why does it emerge "right there" on the number line, halfway between 2 and 4?


    It is even easier to see with other constant like e that are directly derived from growth processes. There the contrasting actions that produce the emergent ratio are in plain sight. It is funny that e should be 2.71828.apokrisis

    It's hilarious.

    But then that becomes obvious when it is realised that growth always has to start from some thing that is just itself 1. There is no reason to think of e as anything but natural after that.apokrisis

    Mystical word salad.


    But I am not Kantian, except in a loose sense. I'm Peircean in the way Peirce fixed Kant.apokrisis

    Maybe we better not go there. You remember how that went last time. Although I suspect this is the problem. You are arguing from a very particular point of view. But you are not willing or able to explain yourself. So everything that anyone says is wrong from your point of view, and you're right, and you're condescending, but you can't explain your ideas in everyday English. And in my experience, you don't really understand a lot of the ideas whose terminology you carelessly sling around.

    And I'm arguing flatness is special as the mid-point of opposing extremes of curvature. It has physically important properties too. Only flat geometries preserve invariance under transformations of scale. That is a really important emergent property when it comes to things like Universes.apokrisis

    Oh to be back in 1840 when Euclidian geometry was given to us by God.


    And as I repeat, it is very important metaphysically that absolute scale invariance only appears at a particular numeric value of pi. That is how a Universe is even possible.apokrisis

    What is the specific numeric value of 3? I already explained to you that pi is a particular real number whose value does not change and has nothing to do with geometry.


    So you are focused on the triviality of pi being given some particular position on the number line - look guys, its 3.141592653589793238462643383279502884197169399375105820974944592307816406286 208998628034825342117067982148086513282306647093844609550582231725359408 ...

    And that is what makes folk go woo. It seems both weirdly specific and weirdly random. There seems no natural reason for the value.
    apokrisis

    What is the natural reason for the value of 3?

    But it's a ratio derived from the radius being granted as the natural unit. Let's call the radius 1. Let's get a grip on this weird thing called curvature by starting with the "most natural part of the story" - a line segment. That gets to be "1" on the number-line.apokrisis

    Bearing in mind that the unit distance is arbitrary. A linear scaling factor would make no difference.

    Well, as I say, once mathematicians woke up to the fact that flatness was a rather special case of curvature, and once physicists in turn woke up to the fact that scale invariance was essential to any kind of workable Universe (its called rather grandly the cosmological principle), well, maybe it is the ratio that should be called "1". A straight line segment is only a natural unit in the context of an already flat space which supports unlimited scale transformations. It depends on the emergent fact of parallel lines or infinite rays being an actual possibility.apokrisis

    I have no doubt that what you're saying makes perfect sense to you. It makes no sense to me. I don't get where you're coming from.

    I am being anti-mystical in pointing out the very physical basis of pi as a number. It is a ratio that picks out a critical geometric balance.apokrisis

    Even if I grant your point, what of it? Why are you going on about pi?

    The number 3 is trivial by comparison.apokrisis

    Then explain to me what mystical geometric balance causes 3 to have the exact value that is has.


    Well there are physical arguments for why the geometry of universes are optimal if they have just three orthogonal spatial directions. But 3 as a member of the integers has no numeric specialness by design. The special or natural numbers are 1 and 0. We see this in the symmetries captured by identity operations. There is something basic or universal when we hit the bedrock that is a symmetry or invariance.apokrisis

    Look, maybe Pearce said all this. I haven't the knowledge to comment. You seem to know a lot about this, or at least you know the buzzwords. It's pointless to argue with you about it.

    You would call it a mystical fact perhaps. I see it as quite reasonable and self-explanatory.apokrisis

    I can no longer converse with you on this point.

    Nope. At least not your notion of computation as Turing machine/programmable computation.apokrisis

    What is your model of computation? And how do you square it with the Church-Turing thesis?

    I take an information theoretic perspective. And more specifically, a semiotic one. In technology terms, neural networks come the closest to implementing that notion of computation.apokrisis

    Yes and any NN that can exist in the physical world is a TM. You have to refute this or accept it. I'm perfectly willing to be shown wrong, since I'd learn something. Do it.

    And numbers vs rocks is a distinction that relies on a classical metaphysics - one in which the divide between observers and observables does not present an epistemic difficulty. The epistemic cut - the necessary separation of the information from the physics - can be treated as an ontological fact.apokrisis

    You are the buzzword king. You do have a hard time translating your buzzwords into ideas that you can explain to people.

    So my positions on both "mind is a computation" and "reality is classical" are the same.apokrisis

    If mind is a computation and if it's implemented on a physical brain, then it's a TM.


    Semiotics starts from the view that there is no fundamental ontic division of observers and observables. But that is also the division which must emerge via some epistemic cut. It is the basis of intelligibility. And even the Universe can only exist to the degree it hangs together in intelligible fashion.apokrisis

    I yield to your facility with buzzwords.

    Hence why maths tends to be unreasonably effective at describing the Universe. Or being in general.apokrisis

    It's a puzzler alright. Or perhaps math is only telling us something about our own minds, and not the universe at large. That's a possibility too.


    Labouring the point still, but I'm sorry. I'm not a computationalist in the sense you are hoping for.apokrisis

    Well we're back to the NN = TM issue again. That's the core issue here. You think there's a mode of computation that is not a TM and that can be physically implemented in the brain. This I deny. Computer science is on my side I believe. I could be wrong. I await clarity.

    Indeed, that was what I was accusing you of. You seem to believe reality is a machine. An account of physical events is sufficient.apokrisis

    I have never believed that and I strenuously oppose it. But it's normal for you to think I've said the opposite of what I think I said. We may just have to live with that.

    But yes, you also seem to say the opposite. This is a symptom that your metaphysics is "commonsensical" and not well thought out.apokrisis

    I admit I'm not much of a philosopher. If you ever took the trouble to explain your points of view to me, I'd learn something. But that never seems to happen.


    Again, bully for mathematicians. Bully for engineers. Bully even for most physicists (as very few are employed in frontier theory construction).

    But it is curious to be complaining about metaphysics where metaphysics is appropriate.
    apokrisis

    You were complaining that mathematicians aren't in a state of "epistemic shock." I pointed out that when mathematicians are doing math, they're not doing philosophy. You want people to do their jobs, not get lost in the wonder of it all.


    And so far you haven't put forward any clear exposition of your own epistemic position, let alone given a clear justification for it.apokrisis

    I haven't got much of an epistemic position today. I vaguely recall originally making some minor point which I no longer remember. I don't believe the mind is a TM and I don't believe real-world NN's are anything other than TMs. That's plenty of position. By the way don't you mean ontic position? What is, rather than how we know what is? I'm even confused about your nonstandard use of "epistemic." I never understand anything you say.

    You just hoped to be able to label me with some obviously weak ontology that I spend most of my time arguing against.apokrisis

    LOL. And you are doing the same to me.

    Bottom line, why don't you just explain to me why you think a real-world NN is anything other than a TM. That's at least one subject that might be of interest, and it's one where we're reasonably on the same page even though we have different opinions.
  • apokrisis
    4.5k
    My understanding is that we can accommodate abstract mental constructs quite easily within physicalism. Abstractions are thoughts, biochemical processes in my brain.fishfry

    I don't believe the mind is a TM and I don't believe real-world NN's are anything other than TMs....

    Bottom line, why don't you just explain to me why you think a real-world NN is anything other than a TM.
    fishfry

    Hmm. So what I have got from this exchange is that you struggle to keep track of your own arguments because you don't actually have a well constructed metaphysical position. And when you encounter someone who does, you bluster and ad hom. Nice.

    And so here now you have diverted the discussion to something that you hope might be safe ground.

    I said that mainstream neuroscience would reject the reductive materialist notion that abstract thoughts are just biochemical processes in the brain. In some fashion - still not fully understood of course - they would be considered informational and semiotic processes.

    You then leapt to the idea that this meant the activities of the brain are computational processes - Turing machine computational.

    I replied no, a TM is a dualistic device. The software is absolutely divorced from the world which gives it rule-bound play any material meaning. It is presumed that the hardware supporting the action has no entropic cost. It is presumed that the input and the outputs of this finite state machine are meaningful to some further intelligence outside it. So a TM is just a syntactic device. It can blindly follow rules. But at no point in its mathematical-strength definition is there any semantics included.

    And then, so far as neuroscientists would consider the brain some kind of computer, it would be like a neural network. Which is different from neuroscientists thinking the brain IS a neural network. Rather, it is neural networks which are like a semiotic relation.

    Neural networks are meant to learn from the world by experience. They don't have a programming language and so they don't have a set of syntactic tokens to shuffle about according to some set of computational grammar. And while they can of course emulate a Turing Machine - just like we can emulate a TM too - that doesn't mean they are TMs. It just means they can follow rules that shuffle symbols without needing to understand anything about what they are doing. Semantics is optional to blind programmatic rule following.

    So you made a wild claim - thoughts are nothing more than biochemistry. Now you want to defend the opposite thesis - thoughts are nothing more than Turing computation. Or no, you realise that is ridiculous. So you want to pretend that is my position instead.

    The circle of mathematics is an ideal circle, a pure mental abstraction.fishfry

    Then you don't seem to be interested in metaphysics even as it touches on the reality of numbers. It appears largely that you reject what physicalism might have to say about "reality" just because looking up "buzzwords" is such a tiresome chore ... when you already have all the answers.

    I was hoping that focusing on the reality of mathematical constants might have got us somewhere. Yet it appears you haven't even really thought about the reason constants emerge as limits on material action in physical systems. So that was a waste of time too.

    Oh well. I was expecting too much, obviously.
  • fishfry
    469
    Hmm. So what I have got from this exchange is that you struggle to keep track of your own arguments because you don't actually have a well constructed metaphysical position. And when you encounter someone who does, you bluster and ad hom. Nice.apokrisis

    I'd love to chat with you without the snark. This isn't fun. I'm going to withdraw. Suggest you investigate the relationship between neural nets and Turing machines. I did a little research since my last post and apparently neural nets are computationally weaker than TMs. Some neural nets are Turing complete but some aren't. And of course none have computational abilities beyond the TM, since we know of no such thing in the physical world.

    All the best.
  • apokrisis
    4.5k
    Maybe you don’t realise that snark is pretty routine on your part.

    And I have investigated neural nets. Your posts on the issue reveal you haven’t really.
  • fishfry
    469
    And I have investigated neural nets.apokrisis

    Then you must have a substantive response to my point that if mind is an informational function of a neural net executing in the brain, then it can be implemented as a Turing machine. So it's not logically consistent to claim that mind is a neural net without admitting that your position requires you to also agree that mind must be a TM.

    That is:

    * If you claim that mind is a neural net; then you must also agree that mind is a TM.

    I could well be wrong. If you have a substantive response to my point then perhaps I'll learn something. If you don't, you don't.
  • apokrisis
    4.5k
    That is:

    * If you claim that mind is a neural net; then you must also agree that mind is a TM.
    fishfry

    Either you understand the difference between emulating a TM and being a TM, or you don't. Either you understand the difference between analog computers and digital computers, or you don't. Either you understand the difference between semantics and syntax, or you don't. Etc, etc.

    Maybe you could start with this famous philosophy of mind argument - https://en.wikipedia.org/wiki/Chinese_room#Chinese_room_and_Turing_completeness

    To understand my biosemiotic take on the issue, this is a nice foundational paper - http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.18.1316&rep=rep1&type=pdf
  • fishfry
    469
    So you are making the point that a NN somehow implements semantics? That's wrong. We'll have to agree to disagree here. And once again you have an insult but not a substantive argument. Everyone here can see that. And since I've referenced Searle twice you'd have to know that I'm thoroughly familiar with the CRA and the surrounding arguments.

    Are you actually incapable of making a substantive argument? "You are wrong because Point 1, Point 2, and Point 3?" You simply can't do it?

    Make a substantive argument or show everyone on this site that you're not capable of doing so.

    Are you actually making the claim that even though a NN can be emulated by a TM, the NN somehow implements semantics? That's so wrong ... well it's wrong. Leave it at that. A thing that can be implemented or emulated as an algorithm is an algorithm. But if that's your argument, so be it. But you haven't provided an argument. You do not possess the ability to outline a rational argument. You make a claim but you provide no substantive argument.

    And for what it's worth, you completely miss the point of the CRA. If you think an NN that can be emulated by a TM can nevertheless embody semantics, you simply have no idea what the CRA says.
  • apokrisis
    4.5k
    Are you actually making the claim that even though a NN can be emulated by a TM, the NN somehow implements semantics?fishfry

    Nope. I made the point that humans and NNs can emulate TMs. (You did claim to be familiar with the CRA?) However that doesn't make either of them TMs.

    I also said neuroscientists find NNs to be biologically realistic models of neural processes. There is no reason to think brains are finite state automata. There is no reason to think they are programmable computers (von Neumann machines). There is no reason to think they are Turing complete. But - given that NNs are inspired by the biology - it is not much of a surprise that NNs implemented even as logic devices show some of the important functionality we associate with nervous systems.

    So NNs are goodmodels. TMs, by contrast, are woeful models of brain function.

    Can an NN have semantics or is it also just a syntactic device? Well, it all rather depends now on how you define semantics. And that is what biosemiotics concerns itself with. One would need a general physicalist theory of semantics to answer the question in some quantitative fashion.

    I would say the NNs built to date aren't really semantic. They are just pattern matching systems. And they require supervised learning, so the semantics are clearly "in the mind" of their human trainers. But arguably they are getting near the abilities of an ant or cockroach.

    I would say that is still only in terms of pattern matching ability. An embodied view of cognition would say that a hell of a lot is still missing in terms of an actual ability to "makes sense of the world" even at that level. NN designers haven't even got their heads around the kind of functionality they need to start implementing as the "learning algorithms" on that score.

    I could say a lot more about the semantic issue, but it's way off topic for this thread.

    The issue was whether maths is Platonically real or a free creation of the human mind. I argued for a third position - one which says the maths that is "unreasonably effective" when it comes to physicalist theories, is so because it describes real physicalist limits on reality.

    So enough of the sideshow. You only turn anything I say back to front anyway.
  • Harry Hindu
    1.5k
    I voted other.

    Numbers are real, and establish causal relationships, and exist in space and time. Numbers exist as little scribbles or sounds that refer to a quantity. Numbers exist in space and time as little scribbles and sounds, and also the quantity they refer to exist in space and time. Some number has a different causal effect than some other number. If you need 100 points to win a game, you keep playing until you get 100 points, at which point you stop playing and start a new game.
  • fishfry
    469
    Nope. I made the point that humans and NNs can emulate TMs.apokrisis

    Oh I quite agree with that. I can take a pencil and paper and step through a program. It's a common debugging technique. And you're right, that doesn't make me a TM. But I can do many things that CAN'T be emulated by a TM. Like understand Chinese. If I could understand Chinese. That's one of Searle's points that he's made over the years. When he speaks English he's doing something very different in character from what he's doing when he speaks Chinese using the symbolic rules.

    You only turn anything I say back to front anyway.apokrisis

    Peace brother. I'm out.
  • apokrisis
    4.5k
    But I can do many things that CAN'T be emulated by a TM. Like understand Chinese.fishfry

    Err, yeah. As I was saying.

    But the problem was you began this by claiming biochemistry is capable of things like understanding Chinese. :)
  • fishfry
    469
    But the problem was you began this by claiming biochemistry is capable of things like understanding Chinese.apokrisis

    That's my understanding of Searle's position as well. The mechanism remains to be discovered.
  • apokrisis
    4.5k
    Do you really want to argue that Searle thinks "biochemical processes" are a necessary and sufficient condition of conscious thought?

    It is well know that Searle fluffs around the issue because he has some broke-arse property dualism in mind.

    But he usually talks about neural processes and brain structures as the likely level where first person experience might "pop out" into existence as an emergent property of a third person material world.

    I've not seen him make a positive assertion that consciousness would be emergent just from "biochemistry", sans all that rather suggestive neural circuitry. So you might want to check your understanding.

    Peace out, as they used to say.
  • fishfry
    469
    Do you really want to argue that Searle thinks "biochemical processes" are a necessary and sufficient condition of conscious thought?apokrisis

    Now that you mention it, I can narrow down my claim. There are really two issues here: What I think about mind; and what I think Searle thinks about mind.

    My belief in the correctness of what I think is in no way affected by the state of my knowledge about what Searle thinks.

    My thesis here is that mind arises from a physical process in the brain; but that it is not a computational process in any way that we currently understand computation. It's not a TM or an NN or a cellular automata or anything else along those lines.

    I don't know what the actual mechanism might be. I'm not sure even what it means to say that there is a process that follows physical law but that is not a computation. I think these are matters for future geniuses to work out. I think this will take another revolution in physics.

    Now as it happens, the way I got my opinion is that I read something Searle wrote, or perhaps said in a video. I may well have misunderstood what he said; but in any event it sparked this thesis in my mind, or gave clarity to some vaguer notions I'd been having.

    So Searle is the inspiration for my opinion but may not himself actually share my opinion. But really -- isn't that classic Searle? For decades people have been misunderstanding the CRA yet have been intellectually inspired by their own misunderstandings.

    ps -- Oh wait you said necessary and sufficient. No I don't think biochemistry is necessary. Or sufficient. It just "happens to be the case" in this instance. It's possible that machinery might become conscious, so biochemistry's not necessary. And there's plenty of biochemical matter walking around that's not particularly conscious, so biochemistry is not sufficient.
  • apokrisis
    4.5k
    My thesis here is that mind arises from a physical process in the brain; but that it is not a computational process in any way that we currently understand computation. It's not a TM or an NN or a cellular automata or anything else along those lines.fishfry

    It seems curious that it was only just a few posts back that you were trumpeting the mind-like abilities of NNs. So if they were inspired by the "computational" structure of the brain, it is surprising they should indeed be so effective at machine learning, and yet the brain itself would not function along these lines.

    I don't know what the actual mechanism might be ... I think this will take another revolution in physics.fishfry

    Sounds legit.

    No I don't think biochemistry is necessary. Or sufficient. It just "happens to be the case" in this instance. It's possible that machinery might become conscious, so biochemistry's not necessary. And there's plenty of biochemical matter walking around that's not particularly conscious, so biochemistry is not sufficient.fishfry

    So that is a retraction of your original statement coupled to a backtrack on the retraction?

    It is the structure of the matter that matters and not the particular matter. But you don't want to say the structure implements any kind of informational process?
  • fishfry
    469
    It seems curious that it was only just a few posts back that you were trumpeting the mind-like abilities of NNs.apokrisis

    I oppose that notion every chance I get. It's not possible that I expressed such an opinion. Two possibilities:

    * You misunderstood something I wrote; or

    * I expressed myself so badly that I communicated the opposite of what I intended.

    One's just as likely as the other.

    But no, NN's are not "mind-like." It's starting to become my mission in life to explain to people why NN's are *NOT* "mind-like." To the extent that I fail to make a convincing argument, I need to work on my argument. To the extent that I'm giving you the opposite impression of what I'm actually trying to say, I have to work harder not to do that.

    So if they were inspired by the "computational" structure of the brain, it is surprising they should indeed be so effective at machine learning, and yet the brain itself would not function along these lines.apokrisis

    Airplanes are stunningly effective at flying, yet birds don't work that way.

    Sounds legit.apokrisis

    You agree with me that perhaps the explanation of mind must await the next revolution (or two) in physics? If you agreed with that point you'd be halfway to agreeing with the rest of my thesis. There's a lot we don't know.


    So that is a retraction of your original statement coupled to a backtrack on the retraction?apokrisis

    You asked me if I thought biochemistry was both necessary and sufficient for mind. I said it's neither necessary nor sufficient. You are seeing this as a retraction? How so?

    It is the structure of the matter that matters and not the particular matter.apokrisis

    I don't know. Perhaps it has to be biological. Perhaps not. I don't think it's relevant to my argument. Whatever mind is, it's not a computation. But I take no position on whether it has to be biological or not. Is that more clear? I don't know what you think I'm retracting.

    But you don't want to say the structure implements any kind of informational process?apokrisis

    Hmmm ... that's kind of an interesting technical question. So there's the neural wetware of the brain, and you are asking me if it is possible that SOME informational process is implemented.

    Um ... well ... sure. Why not. If I blink my eyes at you in morse code I'm digitizing my thoughts. For that matter, I can execute the Euclidean algorithm with pencil and paper. So yes, wetware can certainly implement computational processes. But not everything wetware does can be explained by a computation.
  • Janus
    6.1k
    And the best way out of that bind is to look to causality and treat that as the best definition of "physical reality".apokrisis

    You agree with Schopenhauer:

    Thus also, whoever has recognised the law of causation, the
    aspect of the principle of sufficient reason which appears in what
    fills these forms (space and time) as objects of perception, that
    is to say matter, has completely mastered the nature of matter as
    such, for matter is nothing more than causation, as any one will
    see at once if he reflects. Its true being is its action, nor can we
    possibly conceive it as having any other meaning. Only as active
    does it fill space and time; its action upon the immediate object
    (which is itself matter) determines that perception in which alone
    it exists. The consequence of the action of any material object
    upon any other, is known only in so far as the latter acts upon the
    immediate object in a different way from that in which it acted
    before; it consists only of this. Cause and effect thus constitute
    the whole nature of matter; its true being is its action.
  • apokrisis
    4.5k
    But no, NN's are not "mind-like." It's starting to become my mission in life to explain to people why NN's are *NOT* "mind-like."fishfry

    Fine. I would agree that NNs are not biologically realistic in some fundamental ways. But also, NNs are an attempt to be more biologically realistic in some important structural or information-processing fashion.

    So this could easily be an argument over whether the glass is half full or half empty. That is why the epistemology of NNs demands especial care in a Philosophy of Mind discussion.

    Airplanes are stunningly effective at flying, yet birds don't work that way.fishfry

    But what is the "unreasonably effective" feature they share? Is it an aerofoil wing that creates lift?

    I agree that human machines are just basically different from biological organisms. However again, you need some actual general metaphysical argument to spell out the precise nature of that difference. And that is what I'm talking about with biosemiosis, autopoiesis and other "buzzwords".

    You need a theory of the distinction if you want to say anything definite on the matter. And you seem quite dismissive of the literature here.

    You agree with me that perhaps the explanation of mind must await the next revolution (or two) in physics?fishfry

    No. I was being sarcastic.

    Physics is already undergoing the right kinds of revolution anyway. Thermodynamics is becoming foundational. Physics is becoming information theoretic. Holism and emergence can now be modelled in a variety of ways.

    So Newtonian materialism is out-dated. Existence can be understood as a dissipative process. And that is a framework which biology and neurology slot straight into.

    I don't know. Perhaps it has to be biological. Perhaps not. I don't think it's relevant to my argument.fishfry

    Well I would say this shows you don't have an appropriate general metaphysical framework. It has to be a central issue if you are arguing either for or against artificial life and mind.

    That is why I urged you to read that Pattee paper.

    Whatever mind is, it's not a computation.fishfry

    That's a hand-waving statement, so not much use in a serious debate here.

    At the moment I have no clue what you even mean by "mind". I get the impression it is probably the standard dualistic substance ontology - a sensing stuff, a bunch of "feels".

    So we wouldn't even be on the same page for a serious discussion in terms of a comparison of neurological processes and computational mechanisms. You are likely already convinced that there is no physicalist understanding of what brains do.

    Hmmm ... that's kind of an interesting technical question. So there's the neural wetware of the brain, and you are asking me if it is possible that SOME informational process is implemented.

    Um ... well ... sure. Why not. If I blink my eyes at you in morse code I'm digitizing my thoughts. For that matter, I can execute the Euclidean algorithm with pencil and paper. So yes, wetware can certainly implement computational processes. But not everything wetware does can be explained by a computation.
    fishfry

    You seem to entirely miss the point.

    You appear to believe that TMs completely define all possible notions of computation, information and semiosis. And so any question about "information processes" or "processing architecture" gets immediately translated into a TM view.

    But just maybe TMs are a very tiny fragment of a much larger landscape.

    Of course, there is something immensely powerful about TMs in being (almost) pure syntax/no semantics. In short, they are (near) perfect machines. They represent a completely constrained and rule-bound universe. And so they leave out all the "messiness" of the physical and biological world. They leave out, in fact, information as traditionally understood - ie: information as meaning.

    It is like the syntax of Boolean logic. To reconnect to the OP, there is something "unreasonably effective" about reaching the limits on a de-semanticised view of reality - one where we just model reality in terms of its simplest syntactical rules.

    So TMs and Boolean logic idealise reality. They abstract away the materiality or particularity of physicalist semantics to arrive at the simplest, sparest, syntactical forms.

    Great. Defining the ultimate limits of reality is what it is all about. But maybe there is such a thing as over-simplification.

    Machines are rule-bound artificial systems. And so they can't construct themselves. They can't give themselves purposes , they don't have autonomy. Machines are useful to us humans as it is we who get to design the machines, build them to serve some purpose.

    However organisms are systems with evolved designs and purposes. They have an irreducible causal complexity. And that is their "secret". There is always semantics - or semiosis - involved.

    So the whole mechanical paradigm of nature is flawed at root if it excludes the basic causal complexity of real living and minding creatures.

    We can see that TMs and Boolean logic leave out formal and final cause. Well they leave out material cause as well. All they are is pure syntax. They can be used - by an organism with a purpose and a design - to represent a formal system of entailment. They can capture the description of a syntactic structure. But being such a rarified representation of reality, the computational patterns that result have an extreme real-world brittleness.

    In practice, any computer program or computer circuit is incredibly prone to bugs. Just one broken link and the whole finite state automata grinds to a halt.

    Organisms by contrasts not only thrive on physical instability, their very existence depends on it. Life and mind arise on the "edge of chaos" as where things are perched on the verge of falling apart, that is where the slightest extra informational nudge can push them instead into falling together.

    So life and mind thrive on material dynamism. TMs and other machines only flourish where all the uncertainties of the real world have been managed out of existence by their human designers. Mindless routine following becomes possible where minds have made that a safe thing to do.

    Anyway, my point is that any biologist or neurologist would understand that computers and organisms are different in this fundamental way. There is a reason why TMs are both such "universal" machines, and also the most biologically helpless of physical structures.

    There is a general metaphysical paradigm that accounts for why brains aren't computers, and yet also, we could build computers that start to have some of that biological realism designed into them.

    A "true" NN has to learn for itself. That's both its advantage and disadvantage. It is essentially a black box to its human owner.

    I know a "mad genius" who has developed one of the currently most advanced neural network computers in the world. It runs his company for him. But he has no clue how it works inside. It grew its own "programme". And if it failed, he couldn't transfer its software to another hardware rack. He can't even do a memory back-up as such.

    But because the memory doesn't work like a traditional TM device, and instead is more like a brain, that is not such a problem as it has natural fault tolerance. The failure of individual links can't corrupt the whole system.

    So yep, the whole NN issue isn't clear-cut. But the field has a history now. Computer science has been exploring the degree to which neurologically realistic architectures can lead to a more organismic notion of a machine.

    We already have a mathematical definition of the most non-organismic one - a TM/Boolean one - as the theoretical limit of a machine that is all syntax, no semantics. So the next question for the engineers is how to start building back in some useful biological realism. And that in turn demands a general metaphysical theory about how to define "semantic processing", or semiosis.
  • apokrisis
    4.5k
    Not sure he said it best. But yep. Materiality is located action - action with a direction.

    Then the other half of the causal story is the global form which constrains actions to locations and directions.

    To reconnect to the OP, that is why I would be a realist about global constraints as well as local degrees of freedom. The two together make for a reality that has an observable structure and regularity.
  • Janus
    6.1k


    Yes, Schopenhuaer's prose is clear enough, but just a little dated and pompous in tone. ;)

    For Schopenhauer, the other half of the story is Will, understood as the primal urge (he didn't want to reduce it to physical forces for they are causal conditions of the manifestation itself) that we see manifested as gravity, mechanical action, chemical reaction, organic growth, hunger, sexual desire, instinct and so on. For Schopenhauer this primal urge or striving is completely devoid of any characteristics that could be understood as being possessed by anything in the world as idea.

    So, no idea of time, space, causality, differentiation and so on can be coherently applied to Will. The emergence of the world as idea would be, in your terms. the symmetry breaking. The world as will is a virtual world; the thing in itself (note, it cannot be things in themselves for Schopenhuaer and their can be no causal relation between the thing in itself and the world as idea; the latter just is the former as manifest). According to Schopenhauer we can 'feel' the world as will in our experience of our own bodies. I think Peirce has a similar notion of the experience of "firtsness", but maybe I have misunderstood.
  • apokrisis
    4.5k
    So, no idea of time, space, causality, differentiation and so on can be coherently applied to Will.Janus

    I think Peirce has a similar notion of the experience of "firtsness", but maybe I have misunderstood.Janus

    This is the problem for me. Peirce makes sense of causality as the development of reasonable habits. I can follow that as an intelligible metaphysics.

    But not Schopenhauer. In the end, I can't piece together a logical description of a coming into being as a concrete self-organising process. The bits don't fit together.

    Peirce liked Schelling. I can see why.

    Peirce at first disliked Hegel but then came to appreciate him. Again, I see why.

    Peirce seems to have been silent on Schopenhauer. Perhaps Schop just wasn't systematic enough for there to be a real metaphysical thesis to critique?
  • Janus
    6.1k
    Peirce makes sense of causality as the development of reasonable habitsapokrisis

    I wonder, though, whether Peirce can make sense of the development of reasonable habits in terms of something more fundamental?

    Schop could say that Will comes to manifest in ever more habitual ways, which become the more reasonable as the world as idea unfolds; he could say that Will gains its increase by establishing habitual manifestations.
  • apokrisis
    4.5k
    I wonder, though, whether Peirce can make sense of the development of reasonable habits in terms of something more fundamental?Janus

    But isn't that the problem? The way you phrase it suggests that you have certain beliefs about the nature of fundamentality.

    The semiotic view is that tychism - chance or spontaneity - is the most fundamental starting point because it has the least regularity or stability. It is the least concrete possible state. So it is not a "something" - some more basic level of substance. It is a state of unfettered anythingness. It is pure instability without habit or regulation.

    This then means reality arises by a restriction on a fundamental anythingness. So Peirce has a metaphysics we can recognise from Anaximander. And one that also is now straight out of modern quantum physics and thermodynamics.

    So any metaphysics that tries to get something from nothing does have a problem. But reducing everything to just something is easy, by contrast.

    And because we know something does indeed exist - us and our cosmos - we already know that nothingness couldn't have been the case. So whatever our metaphysical reasoning leads us to as the "primal condition" has to be the best answer we are going to get. Which is why we would believe in Firstness, vagueness, the Apeiron, a quantum foam, or whatever best represents a condition of chaotic symmetry, a realm of utterly unstable fluctuation.

    This foment may indeed sound a little like a primal raging will to exist. But does any connection to Schop go deeper than that?

    Schop could say that Will comes to manifest in ever more habitual ways, which become the more reasonable as the world as idea unfolds; he could say that Will gains its increase by establishing habitual manifestations.Janus

    Yes. Maybe Schop could be mapped to this kind of "anythingness" based metaphysics. I mean it is the general alternative option that runs through all creation stories.

    Either there was nothing, and existence got created. Or else existence is a result of a disorganised everythingness that got regulated.

    I just think that Peirce developed the best account of the mechanism for a self-organising everythingness. He realised it had to be a triadic tale, not merely a dualistic one.

    For me, when you describe Schop's Will, it seems to be trying to stand for two things at once - both the material spontaneity and the formal constraints. It is the primal source of the energy and also the end towards which that energy is directed.

    Again, a triadic view allows everything to arise emergently. It stands against the usual view where something can only come from something - the view that presumes substance to be a conserved quantity in the process of creation. Peirce's metaphysics is an open systems view - one which starts in the unlimited and develops its concreteness through self-bounding or self-closure.

    So the question about what is "fundamental" is flipped. Any beginning - and any ending - have to be the least concrete kinds of causes imaginable. As the concrete is what arises in the middle between them.

    That means the beginning is a Firstness or vagueness. Just pure fluctuation. And the ending is Thirdness or generality. Just the fixity of "a habit". A state where all differences are assimilated to a common idea.

    So, as I say, the beginning and the end are real (ie: not nominalist). But they are logically opposed (one being vague, the other general) and both are arrived at as being as "insubstantial" as can be imagined.

    The Will, by contrast, seems to exist as an efficient cause that drives the action. It is definite at the start, and gets to where it alway intended by the end.

    It just lacks the formal dichotomous division of the vague and the general (as that to which the principle of non-contradiction and the law of the excluded middle respectively fail to apply). And it lacks the insubstantiality that can then stand as the contrast to the actuality, the particularity of secondness, that arises emergently "in the middle" - in good old hylomorphic fashion.

    So Peirce has deep roots that are sunk right into logic itself - the laws of thought. It connects in direct fashion to Aristotelian metaphysics as well.

    I don't get any of this kind of rigour from Schop. But then I've never looked into him that deeply.
  • fishfry
    469
    Fine. I would agree that NNs are not biologically realistic in some fundamental ways.apokrisis

    Then we are fundamentally in agreement.

    But also, NNs are an attempt to be more biologically realistic in some important structural or information-processing fashion.[apokrisis

    Of course. I hope you don't think I'm dismissive of the amazing achievements of weak AI these days. From chess to Go to driving cars to facial recognition.

    But isn't it true that just because we can cleverly simulate an approximation of certain aspects of the human mind, that this does not necessarily mean that this is literally how the human mind works?

    In other words we've invented flying machines; but that doesn't mean we've discovered the mechanism by which birds fly. On the contrary, we've shown that the way humans make flying machines is different than the way God (or nature) does. And by analogy, perhaps the way humans make thinking machines is different than the way nature does. In which case, NN's can be as clever as you like, but minds are still not NNs. Just like birds aren't jet engines with fixed wings and small bags of salted peanuts.

    So this could easily be an argument over whether the glass is half full or half empty.apokrisis

    Humans can build machines that emulate what nature does, but we do it differently than nature does.

    That is why the epistemology of NNs demands especial care in a Philosophy of Mind discussion.apokrisis

    Can you explain what the epistemology of NNs means? That went over my head.

    Airplanes are stunningly effective at flying, yet birds don't work that way.
    — fishfry

    But what is the "unreasonably effective" feature they share? Is it an aerofoil wing that creates lift?
    apokrisis

    Well that's the point. Just because we can make flying machines doesn't mean that nature's flying machines work the same way. And just because we can make thinking machines doesn't mean nature's thinking machines work that way.

    I agree that human machines are just basically different from biological organisms.apokrisis

    Well we have to be careful here. If I'm making a physicalist (but not computationalist) argument, then I must admit that we are machines. But we are not computations. I'm making that distinction. We are machines that think, but we are not computations.

    However again, you need some actual general metaphysical argument to spell out the precise nature of that difference.p/quote]

    Sadly I have none. I am a philosophical ignoramus. My shame knows no bounds. Yet must I have a general metaphysical argument before I can have an opinion about whether minds are algorithms?
    apokrisis
    And that is what I'm talking about with biosemiosis, autopoiesis and other "buzzwords".apokrisis

    My complaint about the buzzwords is that you often fail to explain your ideas to me in ways I can understand. And even if I went back to school and got a Ph.D. in philosophy, I might still miss out on Peirce. So if you care to explain your viewpoints I'd learn something.

    You need a theory of the distinction if you want to say anything definite on the matter. And you seem quite dismissive of the literature here.apokrisis

    No, only dismissive of your unwillingness to explain things that you know and I don't.

    You agree with me that perhaps the explanation of mind must await the next revolution (or two) in physics?
    — fishfry

    No. I was being sarcastic.
    apokrisis

    Ah, a crushing disappointment then.

    Physics is already undergoing the right kinds of revolution anyway. Thermodynamics is becoming foundational. Physics is becoming information theoretic. Holism and emergence can now be modelled in a variety of ways.apokrisis

    So physics is already pretty much complete, and all we need to do is fill in the details? That's exactly what they thought in 1900, right before the quantum revolution. Surely history teaches us that there are radical paradigm shifts in physics every couple of centuries. The next one might not happen in our lifetimes but it's certain that it will happen sooner or later.

    So Newtonian materialism is out-dated.apokrisis

    Just as our current ideas may soon be equally as out-dated.

    Existence can be understood as a dissipative process.apokrisis

    What does that mean? I confess to having lived a dissipated life but I don't think that's what you mean.

    And that is a framework which biology and neurology slot straight into.apokrisis

    Sure. No disagreement.

    Well I would say this shows you don't have an appropriate general metaphysical framework.apokrisis

    To my eternal shame. Must one have a general metaphysical framework in order to have an opinion about whether minds are algorithms? I don't know much about philosophy but I do know a lot about algorithms, and I know that algorithms are literally dumber than rocks. A rock sitting there on the ground has more claim to be a thinking entity than an algorithm does. I don't need to have a general metaphysical framework to know that.


    It has to be a central issue if you are arguing either for or against artificial life and mind.apokrisis

    If you are arguing that before I can express an opinion I need to know more about philosophy, this forum would be a lonely place if you applied that standard to everyone. But I do take your point. I'm ignorant of philosophy. I've stated that many times. I don't pretend to know what I don't know.

    That is why I urged you to read that Pattee paper.apokrisis

    Sorry I must have missed that. Link again please?

    Whatever mind is, it's not a computation.
    — fishfry

    That's a hand-waving statement, so not much use in a serious debate here.
    apokrisis

    It's a summary of my strongly held belief, based on a lifetime of having a mind (and occasionally losing it) and an adult lifetime of working with algorithms. Algorithms are much dumber than minds.

    At the moment I have
    no clue what you even mean by "mind". I get the impression it is probably the standard dualistic substance ontology - a sensing stuff, a bunch of "feels".apokrisis

    Well if you're going to take me down the rabbit hole of the philosophy of mind, I'm sure you know more than I do. But I don't think that my subjective experiences and ability to be in the world as an intelligent human being are the result of an algorithmic process.

    So we wouldn't even be on the same page for a serious discussion in terms of a comparison of neurological processes and computational mechanisms. You are likely already convinced that there is no physicalist understanding of what brains do.apokrisis

    How can you say that? Of course there is some physicalist understanding of what brains do, even though our current state of knowledge is quite limited. And since I've said repeatedly that mind [whatever it is] is a function of brain/biochemistry, it follows that there may someday be understanding of it. Why would you think I've said the opposite?

    You seem to entirely miss the point.apokrisis

    I answered the literal question you asked. You should ask a more precise question then.

    You appear to believe that TMs completely define all possible notions of computation, information ...apokrisis

    That's my understanding of the Church-Turing thesis. If you have a different idea I'd be interested to hear it.

    and semiosis.apokrisis

    I confess I don't understand the meaning of that word, and looking it up from time to time doesn't seem to help.


    And so any question about "information processes" or "processing architecture" gets immediately translated into a TM view.apokrisis

    That's Church-Turing, and nobody has figured out how to defined a model of computation or information processing that violates it by not being a TM. You are very casually ignoring this point and I'd like to you to comment on it.


    But just maybe TMs are a very tiny fragment of a much larger landscape.apokrisis

    YES!! But not according to contemporary theory, because we know of no such larger computational landscape. That's exactly why I think we need a revolution in physics that shows us how to go past TMs into some mode of computation that is more powerful than a TM.

    Of course, there is something immensely powerful about TMs in being (almost) pure syntax/no semantics. In short, they are (near) perfect machines. They represent a completely constrained and rule-bound universe. And so they leave out all the "messiness" of the physical and biological world.apokrisis

    Yes. You are eloquently expressing my very point.


    They leave out, in fact, information as traditionally understood - ie: information as meaning.apokrisis

    Aha. Here we have a point of divergence in our worldviews. Information processing is what a TM does, by definition. Flipping bits according to an algorithm.

    When you say "information is meaning," that's something I absolutely deny by my definition of information. A bitstring carries information but it does not carry meaning. Only a human can say what the computation "means." Searle's intensionality. The Chinese room manipulates information but it has no notion of what any of it means. That's the entire point. One with which you sometimes seem to agree.

    I don't think you can claim that information is meaning. Information is meaningless. Humans give meaning to information. Isn't that true? If I say I saw a "cat," the symbols by themselves convey know meaning. It's humans, English-speaking ones at that, who say that the word cat stands for a furry domesticated mammal that's not a dog.

    It is like the syntax of Boolean logic. To reconnect to the OP, there is something "unreasonably effective" about reaching the limits on a de-semanticised view of reality - one where we just model reality in terms of its simplest syntactical rules.apokrisis

    Well you can't model all of reality with algorithms. Only certain aspects of it. The map is not the territory and the model is not the thing itself.

    So TMs and Boolean logic idealise reality. They abstract away the materiality or particularity of physicalist semantics to arrive at the simplest, sparest, syntactical forms.apokrisis

    Yes. We're 100% in agreement here. Algorithms abstract certain aspects of reality.

    Great. Defining the ultimate limits of reality is what it is all about. But maybe there is such a thing as over-simplification.apokrisis

    Like claiming that because NNs can do clever tricks like playing Go, that the human mind must be an NN. Now THAT is over simplification.


    Machines are rule-bound artificial systems.apokrisis

    If by machines you mean computations. If I am making a physicalist argument I have to claim that humans are machines but not computations.

    And so they can't construct themselves. They can't give themselves purposes , they don't have autonomy. Machines are useful to us humans as it is we who get to design the machines, build them to serve some purpose.apokrisis

    Yes. Humans provide the meaning. Information by itself has no meaning. You are agreeing with me again. Been happening a lot lately!!

    However organisms are systems with evolved designs and purposes. They have an irreducible causal complexity. And that is their "secret". There is always semantics - or semiosis - involved.apokrisis

    YES!!!!!!!!! That's what I'm saying. Minds go way beyond algorithms. You are totally agreeing with me.

    So the whole mechanical paradigm of nature is flawed at root if it excludes the basic causal complexity of real living and minding creatures.apokrisis

    Ah. No.The computational paradigm of nature is flawed. We need a breakthrough in figuring out physical explanations how meat machines like us can have minds yet not be computations. We need a revolution in physics, we need to break through the Church-Turing constraint, we need to figure out the nature of a machine that does more than computing.

    We can see that TMs and Boolean logic leave out formal and final cause. Well they leave out material cause as well. All they are is pure syntax. They can be used - by an organism with a purpose and a design - to represent a formal system of entailment. They can capture the description of a syntactic structure. But being such a rarified representation of reality, the computational patterns that result have an extreme real-world brittleness.apokrisis

    Right. Humans aren't computations. After all this you're agreeing with me.

    But then you say well yes humans aren't TMs but they are NN's. And you won't come to terms with the fact that NNs are a special case of TMs. NNs are algorithms. So you aren't gaining anything by claiming that humans are NNs and not TMs. We keep going over this point.

    In practice, any computer program or computer circuit is incredibly prone to bugs. Just one broken link and the whole finite state automata grinds to a halt.apokrisis

    Humans too. Disease, death. All machines are imperfect. The solar system probably isn't stable. [Open question at the moment]. The universe will someday collapse or else expand into eternal cold. Reality is highly imperfect. What point are you making?

    Organisms by contrasts not only thrive on physical instability, their very existence depends on it. Life and mind arise on the "edge of chaos" as where things are perched on the verge of falling apart, that is where the slightest extra informational nudge can push them instead into falling together.apokrisis

    Well you're arguing that humans aren't computations, which I've been saying for several posts now. You agree with me. I'm gratified.

    But if you want to claim that humans are NN's, you have no argument because NN's are TMs.

    So life and mind thrive on material dynamism. TMs and other machines only flourish where all the uncertainties of the real world have been managed out of existence by their human designers. Mindless routine following becomes possible where minds have made that a safe thing to do.apokrisis

    So minds aren't algorithms. You are making my point for me.

    Anyway, my point is that any biologist or neurologist would understand that computers and organisms are different in this fundamental way. There is a reason why TMs are both such "universal" machines, and also the most biologically helpless of physical structures.apokrisis

    Right. Right. Right. Right. So the question is: What exactly are we doing that goes beyond mere algorithms? That's the question. You haven't got an answer. I haven't either. But we agree that humans are not computations. And that NNs are computations. So humans aren't NNs either. We don't understand the mechanism by which humans operate in the world.


    There is a general metaphysical paradigm that accounts for why brains aren't computers, and yet also, we could build computers that start to have some of that biological realism designed into them.apokrisis

    Sure, the chess playing and Go playing and automobile driving and face recognizing algorithms are very impressive these days. That doesn't tell us anything about minds.

    A "true" NN has to learn for itself. That's both its advantage and disadvantage. It is essentially a black box to its human owner.apokrisis

    So you are distinguishing between the fake NN's that are merely reorganized ways of implementing TMs' and "true" NNs that are some theoretical construct that are NOT the NNs of current theory. That's it right there. You admit that you are not talking about NNs as currently understood. You are using "NN" to mean whatever it is that humans do, that's not a computation.

    I know a "mad genius" who has developed one of the currently most advanced neural network computers in the world. It runs his company for him. But he has no clue how it works inside. It grew its own "programme". And if it failed, he couldn't transfer its software to another hardware rack. He can't even do a memory back-up as such.apokrisis

    I call bs on that. Not that you don't know some guy, but that he can't back up his system. If it's built out of processors and memory devices then he can back them up just fine with perfectly conventional techniques. There is no magic hardware paradigm. There are only different ways of organizing the bit-flipping activity.

    But because the memory doesn't work like a traditional TM device, and instead is more like a brain, that is not such a problem as it has natural fault tolerance. The failure of individual links can't corrupt the whole system.apokrisis

    I spent years working on fault-tolerant systems. I think your friend is messing with you. Error-correcting codes, fault-tolerance via hardware redundancy, via voting, via consensus algorithms, are all well-studied ideas since the 1970's. Cryptocurrencies fall into that space, they're a brilliant solution to the problem of distributed consensus on adversarial networks.

    Your friend doesn't have a computer that's "more like a brain." He's just taking advantage of his philosophically trained friend who thinks computers are magic, and funnin' ya.

    So yep, the whole NN issue isn't clear-cut. But the field has a history now. Computer science has been exploring the degree to which neurologically realistic architectures can lead to a more organismic notion of a machine.apokrisis

    Absolutely. But they have not broken the Church-Turing barrier. They have not implemented a mind. And they have not by any stretch of the imagination proved that minds are NNs, any more than a 747 proves that a bird is like an airplane.

    We already have a mathematical definition of the most non-organismic one - a TM/Boolean one - as the theoretical limit of a machine that is all syntax, no semantics. So the next question for the engineers is how to start building back in some useful biological realism. And that in turn demands a general metaphysical theory about how to define "semantic processing", or semiosis.apokrisis

    Ok, you're arguing that the AI community needs to learn more philosophy. A traditional argument. But it still doesn't mean that minds are algorithms or NNs. It only means that NNs are better at building thinking machines than earlier computing paradigms.


    I do hope you agree that building artificial machines that exhibit "thinking" in constrained domains is one thing; and that claiming that the human mind works that same way is quite another.
  • apokrisis
    4.5k
    But isn't it true that just because we can cleverly simulate an approximation of certain aspects of the human mind, that this does not necessarily mean that this is literally how the human mind works?

    In other words we've invented flying machines; but that doesn't mean we've discovered the mechanism by which birds fly.
    fishfry

    This ignores the fact that the flying machine designers quickly gave up trying to copy the flapping wings of birds and instead focused on a non-bird model of flying machines. The flapping did not prove "unreasonably effective".

    Whereas the opposite is the case with NNs. Having got programmable computers, it was the case that even just emulating biologically-inspired information processing architectures was "unreasonably effective" for certain tasks, like pattern matching.

    So that is a particularly inapt comparison with which to make your case.

    If I'm making a physicalist (but not computationalist) argument, then I must admit that we are machines.fishfry

    So an organism is a machine? You seem out of touch with biology.

    Sorry I must have missed that. Link again please?fishfry

    Artificial Life Needs a Real Epistemology - H. H. Pattee
    http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.18.1316&rep=rep1&type=pdf

    Of course there is some physicalist understanding of what brains do, even though our current state of knowledge is quite limited. And since I've said repeatedly that mind [whatever it is] is a function of brain/biochemistry, it follows that there may someday be understanding of it. Why would you think I've said the opposite?fishfry

    So on the one hand you can't even define what you might mean by mind. On the other, you can make confident claims about neuroscience having a quite limited understanding. And you keep reverting to talk of "brain biochemistry" when the question is about cognitive functions.

    Don't you see the inconsistency of one minute admitting to knowing little, the next to be making a sweeping judgement of the whole field?

    That's my understanding of the Church-Turing thesis. If you have a different idea I'd be interested to hear it.fishfry

    That defines computation in the general limit ... if you are computing number theoretic functions.

    So perhaps brains might not be that kind of "computer". Maybe there is not a single arithmetic operation involved in their neural processes. Maybe even "summing weights" is just an analogy for the integrative processes of brain cells. Church-Turing may have zilch to do with neurology. And yet it is still wrong to then attribute neural information processes to "biochemistry".

    And how could you have a view either way without a little more neuroscience to inform your opinion?

    That's exactly why I think we need a revolution in physics that shows us how to go past TMs into some mode of computation that is more powerful than a TM.fishfry

    Given that TMs require no more physics than a gate that can read, write and erase a symbol on an infinite tape, why the heck would we expect new physics to make a difference to Turing universal computation?

    The power of Turing machines is that they need the least physics we can imagine. What more do you want - time travel, Hilbert space, quantum teleportation? That's back to front. It is the virtual elimination of any complicated physics which is the guarantee of the computational universality.

    When you say "information is meaning," that's something I absolutely deny by my definition of information.fishfry

    Who could win an argument against your private definitions?

    So let's stick to the real world of science, maths and philosophy. If you want to talk about Shannon entropy, fine. But then we all know that is based on counting meaningless bits. If we understood the pattern to mean something, then each successive bit would fail to be such a surprise.

    If I know you are transmitting the digits of pi, I could stop you right after you said "3".

    I don't think you can claim that information is meaning. Information is meaningless. Humans give meaning to information. Isn't that true?fishfry

    You don't get it. Information theory defines a baseline where the meaning of a bit string is maximally uncertain. Each bit says nothing about the following bit. Then from that baseline, you can start to quantify the semantics. You can derive measures such as mutual information that speak to the information content.

    I don't think you can claim that information is meaning. Information is meaningless. Humans give meaning to information. Isn't that true? If I say I saw a "cat," the symbols by themselves convey know meaning. It's humans, English-speaking ones at that, who say that the word cat stands for a furry domesticated mammal that's not a dog.fishfry

    That's one of the advantages of a semiotic approach to the whole issue. It recognises that there is a modelling relation involved. A symbol has meaning due to a habit of interpretation. That habit is tied to action in the world. So the informational side of the equation is causally connected to the material side. There is only meaning in relation to the material consequences of any beliefs.

    Again, read Pattee - http://www.academia.edu/3144895/The_Necessity_of_Biosemiotics_Matter-Symbol_Complementarity

    But then you say well yes humans aren't TMs but they are NN's. And you won't come to terms with the fact that NNs are a special case of TMs. NNs are algorithms. So you aren't gaining anything by claiming that humans are NNs and not TMs. We keep going over this point.fishfry

    You keep misrepresenting my argument.

    The significance of an NN would be that it captures something important about brain cognition. That is different from claiming the brain is literally just an NN.

    And you seem confused about algorithms. They are rules for making calculations. So they are something we think it meaningful for a TM to do. They are not the barest syntax of rule following we can imagine. They are semantic actions performed on a machine.

    So already we are into the real world where computation carries extra semantic baggage. The algorithms are intended to represent some actual informational process. This could be just handling a company's payroll or driving a video display. Or it could be an attempt to mimic the connective behaviour of neural circuits.

    A TM is just a universal algorithm runner. How we then exploit that is down to the kind of information processing we think might be meaningful. We have to write an algorithm that seems to perform the task we have in mind. That could be representing brain functions. It could be representing accounting functions or moving image functions. Universal Turing machines have zilch to say about whether we humans are choosing to run usefully realistic routines or just scrambled garbage randomly concocted.

    You are confusing yourself in jumping so interchangeably between talk of TMs, information, computation and algorithms.

    What exactly are we doing that goes beyond mere algorithms?fishfry

    Again, we write the algorithms. They have zilch to do with the universality of TMs. So you can't claim them as "mere". They are intended to represent some meaningful relation expressed as some mathematical operation. They have to perform a function we find useful. Thus they could model a company's payroll, or model the cognitive operations of a brain.

    A payroll model is probably pretty ho hum. But a workable brain model?

    Yes, the map is not then the territory. As someone pushing semiosis - a modelling relations view of "information processing" - you don't have to explain that to me. It is what I've been saying.

    You admit that you are not talking about NNs as currently understood. You are using "NN" to mean whatever it is that humans do, that's not a computation.fishfry

    You are convincing me of your utter unfamiliarity with neural networks in practice. Or even in theory.

    I call bs on that. Not that you don't know some guy, but that he can't back up his system. If it's built out of processors and memory devices then he can back them up just fine with perfectly conventional techniquesfishfry

    In fact it is completely custom hardware. It is not a simulation of a neural net on conventional technology. It is a direct hardware implementation of a neural network.

    I do hope you agree that building artificial machines that exhibit "thinking" in constrained domains is one thing; and that claiming that the human mind works that same way is quite another.fishfry

    Yes, I've spent 40 years being critical of the over-blown claims of computer science. So I am basically skeptical of the usual talk of getting close to building "a conscious machine". I know enough about the biology of brains to see how far off any computer system still is.

    Indeed, I would like it if there was an in principle argument for why no mechanical device could ever simulate the necessary biological processes. It would suit my prejudices. So I am just being honest when I confess that there isn't an absolute argument. The effectiveness of NNs suggests that some level of mind-like technology - as good as cockroaches and ants - may be feasible.

    And remember where this started - your claim that abstract thoughts are biochemical processes. You followed that howler by jumping the other way - saying the mind was in no way the product of informational processes.

    This second misstep was based on your very narrow conception of information processing - one rooted in TMs.

    The reason for the unreasonable effectiveness of TMs is that they are the theoretical limit on semiotic encoding. Semiosis depends on symbols. A TM is the conceptually simplest machine for handling symbol strings.

    A DNA strand can code for a pretty vast array of protein molecules, but that’s it really. Human language can code for a vast array of ideas. That's really powerful as we know. But a TM can implement mathematical algorithms. It can articulate any mathematically-constructable pattern. That is a whole other level of semiosis.

    So yes. TMs are really basic. They represent pure syntactic potential, stripped of all physical constraints as well as all semantic.

    But then we do have to build back the semantics - add the algorithmic structures - to make TM-based technology do actually useful things. Much like DNA has to code for the kind of neural connectivity that can do actually useful things for organisms.

    Semiosis recognises the essential continuity here. It sees the ontological difference that codes or syntax makes, the new "unphysical" possibilities they create.

    Maybe that's the "physics revolution" you are talking about. I certainly think that it is myself. It explains the information theoretic and thermodynamic turn now happening in fundamental physics I would argue.
  • Janus
    6.1k
    But isn't that the problem? The way you phrase it suggests that you have certain beliefs about the nature of fundamentality.apokrisis

    I don't believe that intelligibility can extend to fundamentality. So, whatever names we use to denote it: substance, God, the Real, Firstness, the noumenal, the Will, the Apeiron, Buddha Nature and so on, will, with all their associations and connotations, be tools to relate them to our various systematic understandings in the intelligible world, the 'World as Idea' as Schopenhauer calls it.

    So, the idea of tychism is really just a dialectical negation of the idea of regularity, stability, concreteness; in short of 'being something'. Spinoza's substance was not thought by him to be "anything", but more like being everything and nothing, inasmuch as to be anything is to be a mode of substance. Hegel similarly said that pure being is close to being pure nothingness.We find apophatic notions of God or Buddha Nature that can be traced back thousands of years. So we can say of Tychism, as Hegel says in another context, that it is the "same old stew reheated".

    This then means reality arises by a restriction on a fundamental anythingness. So Peirce has a metaphysics we can recognise from Anaximander. And one that also is now straight out of modern quantum physics and thermodynamics.apokrisis

    Yes, or we can say that reality, 'being something', is an expression of a blind striving that is in itself not yet anything, or, on the other hand, that it is the emanation of an unfathomable, infinite intelligence, or that the 'being something' of temporality is the other face of the 'being everything and nothing' of eternity, and so on. The point of this is that the emergence of concrete somethingness as a process cannot be intelligibly traced back into firstness, because that is where intelligibility ends. We cannot say what is the symmetry of firstness that is broken to produce secondness, unless we impute an intelligence (albeit of an unfathomable order) to firstness, an intelligence of which our intelligence is a temporal reflection. That is the limitation of Schopenhauer's system; it is inexplicable that an ordered Cosmos can be the expression of a blind will. The same goes for any system that thinks firstness as a blind chaos.
bold
italic
underline
strike
code
quote
ulist
image
url
mention
reveal
youtube
tweet
Add a Comment

Welcome to The Philosophy Forum!

Get involved in philosophical discussions about knowledge, truth, language, consciousness, science, politics, religion, logic and mathematics, art, history, and lots more. No ads, no clutter, and very little agreement — just fascinating conversations.