Loading ...
Sorry, an error occurred while loading the content.
 

Re: Smith's essay in "Computationalism: New Directions"

Expand Messages
  • thoughtfuldavid
    Hi everyone, I have been a silent observer of this group for a while, but not really had the time (or the background) to contribute to the discussion so far.
    Message 1 of 26 , Aug 1 5:17 AM
      Hi everyone, I have been a silent observer of this group for a
      while, but not really had the time (or the background) to contribute
      to the discussion so far. However, I am now on "vacation" and (as
      Eray knows) this particular topic is very dear to my heart, so, yes
      please Josh, I would love to spend a while discussing it.

      I should say that I was actually priviledged to attend (and present
      a paper at) the conference-workshop from which the works in this
      book were derived -except for Aaron Sloman's. Unfortunately, it was
      5 years ago so I am not sure how much I can recall now! Anyway,
      below are a few initial comments on what you and Eray have posted so
      far. I look forward to further discussions.


      > Modestly titled, "Foundations of Computing", the basic idea is that
      > in order to understand how computation can underlay cognition, we
      > need to first understand computation. Smith basically tells us that
      > everything we knew about computation was wrong.

      I am not sure Smith is saying that what we know is wrong, rather we
      do not have a clear understanding of what computation actually is.
      Different people think of it in different ways as outlined in the
      six (7, 10, 11?) construals that he mentions. His goal has been to
      identify a single neat "Theory of computation" but by his own
      admission, he has been unable to do so.

      Computationalism is saying (very loosely) something to the effect
      that the "mind is a computer". It was always tacitly assumed that we
      knew what a computer was. Suddenly however, it seems we are less
      than clear about what a computer/computation actually is, which
      makes the project of understanding the mind in computational terms
      all the more difficult.


      > With respect to Smith and Sloman, I think they underrate the
      > importance of Turing's (and others') theoretical contributions.
      > The Turing Machine is not just a machine, it is a model of
      > computation.

      The significance of Turing's work to computation is beyond doubt.
      Surely, what Smith is saying is that there is more to computation in-
      the-wild than Turing's theoretical contributions indicate (sort of
      like knowing all about atoms does not tell you everything about
      water -for instance, that it is mixed with coffee and drunk by
      humans.) As for Sloman, it seems to me that he is merely pointing
      out that knowing these theoretical aspects of computing will not
      help in the quest to understand mind. Again, Turing's theory is so
      far abstracted from real-world considerations as to make it largely
      irrevelant to our present concerns -AI & Cognitive Science. [Eray,
      surely the Turing Machine is not really a machine since it cannot be
      built!]

      Perhaps I should stop at this point and ask whether that helped
      clarify things a bit or merely served to confuse them more?

      Eray- you question Smith's first construal, the idea of FSM (Formal
      Symbol manipulation). It seems reasonable to me. Can you perhaps say
      exactly do you think is wrong with it?
    • Jim Whitescarver
      While the Turing machine is not a good practical model for many problem types it is most certainly a general theoretical model of computing. The implication
      Message 2 of 26 , Aug 1 7:10 AM

        While the Turing machine is not a good practical model for many problem types it is most certainly a general theoretical model of computing.   The implication that is is somehow incomplete is pure quackery.  While a taxidermy of computational models is a worthwhile endeavor that can be enlightening he seems to be mixing analytical techniques with differing degrees of generality.

        On the issue of programming the mind in the laboratory, I do not agree that any new paradigm of computation is essential.  Below I have copied my largely subjective view of the issue, based on a model of the mind developed through meditation, suggesting how is might be emulated.   There is also some discussion and links which may be of interest here. 

        Jim
        From http://wikiworld.com/Consciousness

        I think consciousness will one day be created in the laboratory. Not next week, or next year. Perhaps in a thousand years. At that time we will know much about it, but it will not answer the ultimate question of why anything at all exists.

        It would be created as an active interconnection of universal self aware (dynamic model of self and environment) entities. Based on todays technology it might be built as neural network simulation running on thousands of machines across the Internet. It might result from the study of life and copying it. It may emerge accidental in the quest for artificial intelligence. It will evolve.

        Self-awareness only requires that an entity include a working model of itself and its environment. Consciousness I believe is self-awareness on a social/collective/composite level. But indeed, the subjective "feeling" of being that goes with consciousness is a mystery that future robots may contemplate but never resolve objectively.  Most "feeling" is chemical.  Somehow the "feeling" of conscious being is something else.

        I believe, being and existence are one. We can make a candle and it comes to be, but though we burn it we cannot erase its being. Consciousness to me is collective being with self awareness.

        Being is existence, collective being is composite connected interacting beings. As we can make or define things, we make beings, we do not put the being in them, it is manifest by virtue of existence.

        I have found that consciousness is like a set of radio programs on many different frequencies. We can tune in at different frequencies through meditation. Most often these channels of consciousness are not aware of each other and have no direct communication but collaborate on building and executing our model of reality. These intelligences within us also collaborate with external intelligences manifesting the spirit of life, and ultimately the spirit of being.

        Usually one frequency, usually somewhere in the beta range, 20-40 cycles per second, dominates the external body action while others maintain and regulate various internal body functions and pursue lower priority mental activities. All but the dominant one are forbidden control and we are not directly aware of them. When a new one takes control, its history becomes our conscious memory and the former moves to the background undetected by us in most cases.

        Q: This figure of 20-40 cycles per second, is it an estimate or a measured frequency, and what is the significance of different frequencies? Are higher frequencies 'more conscious' and lower ones less so? Does this model mean we all have multiple personalities?

        measured (and subjective meditation experience):

        • 0-6 cycles/second delta level - fish and reptiles - heart beat, spirit world, God
        • 7-12 cycles/second alpha level - mammals and dinosaurs - most human bodily functions
        • 13-25 cycles/second beta level - human - vision, movement and speech, spirit, God
        • 25-50000 sound - or brain really humming :)
        • radio
        • heat
        • radar
        • light
        • UV
        • Xray
        • gamma ray

        Complexity and ultimately consciousness can exist at any frequency, lower frequencies seems timeless and all knowing, high frequencies really hum. Indeed there are multiple trains of thought that we are normally unaware of, they are sort of different personalities, but really just parts of your one personality unless their is a conflict between them.

        What we call habit behavior is really conscious but we have no reason to be aware of it or if of what we think we are thinking. Only one train of thought is remembered at one time.

        Intuition and Beliefs

        Difficult problems by definition are non-linear. They do not succumb to linear solutions. They are solvable by techniques such as neural network simulations and EvolutionaryProgramming. It take a non-linear solution to solve a non-linear problem.

        With a linear solution, you can explain the answer, You can say because x and y and z blah blah blah. Non-linear solutions, however, only provide an answer. If the problem is defined correctly the answer will be correct.

        It used to frustrate me immensely that the woman in my life would not listen to reason and refused to give a good reason for their conclusions yet very often turned out to be right.

        Finally I figured out how it works. They build an analogy to the problem in their mind that most men see as irrelevant and they run the model to get the answer and as the model runs and gets better answers they change their mind in what seems to be an arbitrary manner.

        It seems obvious to me now that most all "leaps of intuition" work in a similar manner. These sort of solutions also have the property that they can be easily distributed among separate processors creating a collective intelligence.

        Non-linear models involve feedback mechanisms. various solutions are tried randomly and by heuristic and reinforced or dampened by feedback as the problem runs. There is always a best solution, but it can be overthrown at any time by new data and suddenly the old solution is replaced by the new. The fact that separate parts of the model can run independently as long as feedback is provided, in one mind or many, is independent of the sudden "leap".

        HumanBrain


        Page originally had the spelling of Consciousness, so for the history of this page prior to this Version 1, check there. ---StarPilot, playing SnuckerBug.


        "Consciousness to me is collective being with self awareness." "Self-awareness only requires that an entity include a working model of itself and its environment" - HumanBrain


        I think this article will enlighten you on the difficulty of reducing consciousness to physics:

        Facing Up to the Problem of Consciousness http://jamaica.u.arizona.edu/~chalmers/papers/facing.html

        -- GetReal

        Any system that processes information has some level of conscious experience. <CROSS JOIN> There is no distinction between processing information and manipulating matter/energy.

        Tell me what this means. Did I miss the point of the article? --DavidSiegel


        The connection between information processing and experience: (1) has not been established. A computer that manipulates a bunch of symbols according to a rule book is not necessarily experiencing anything, no more than an abacus being manipulated - is experiencing anything. (2) Even if there are cases where information processing entails experience - that connection is a further fact that needs to be explained.

        That experience would accompany some subset of physical processes instead of those physical processes just working in the dark i.e. without any such experience - is a further fact that needs explaining.

        One could adopt the belief that every physical thing experiences or that every physical interaction entails experience - but that would be an additional proposition to add to physics.

        The great task of physics, chemistry, biology and brain science - is to show how experience arises from mere physical processes, not merely to accept it as an unexplained brute fact.

        -- GetReal


        Science can answer how the machinery of consciousness works. What I've discovered subjectively how is is composed of streams of associations (thoughts/consciousness) on every frequency or channel, and organically left, right, front, back, all brain within a brain within a brain. I expect science can show that this model corresponds to neural network activity of the brain.

        I descended in meditation down to the consciousness of much lower than one cycle per second where beating the heart is conscious. My experience is that nothing is really unconscious and consciousness is the whole myriad of thought streams both currently controlling our body, by direct awareness and by "feeling" their effect on whatever they do control. The stream that seems most important generally takes control and all others continue in the background without our direct awareness as confusion results if we try to give control to more than one.

        If we want to understand this beyond the physical, only the science of memes, so far, is instructive. So that we can differentiate between the understanding of the mechanism and the non-physical, I call the non-physical part "Being", as distinct from consciousness or self awareness, which can be tested for in behavior and explained by physical processes. My understanding is the "being" is social or collective, as it is a distributed meme process.

        The treatment of information in Chalmers link is excellent.

        To borrow a phrase from Bateson (1972), physical information is a difference that makes a difference.

        I would also point out that Bateson defined knowledge as a difference that makes a difference that makes a difference. This is very cool stuff. It shows how fractals of domains of information are layered into information about information, or meta information. In a similar sense, conscious being is the awareness of self awareness. Like a higher mammal recognizing itself in a mirror. It is having a center of being outside oneself.

        The physical world also is composed of information channels on every frequency in the quantum and in every fractal level of organization in information ecologies where by the simple rules of EvolutionaryGameTheory intelligence ultimately emerges dominant. But this intelligence too is collective, not individual, and is seems to me we are part of that collection, the cosmic consciousness.

        --JimScarver

      • Jim Whitescarver
        I should have mentioned that in my view the object model is a complete and simple model of computation that makes the issue of methodology mute. It have two
        Message 3 of 26 , Aug 1 9:12 AM
          I should have mentioned that in my view the object model is a complete and simple model of computation that makes the issue of methodology mute.  It have two forms:

          Postfix: object message
          Prefix message object

          Where the message content determines the method and outputs if any.

          This is unrestricted computing which can use arbitrary methodologies for method definitions.  It is a general model of communicating entities, directly communicating only when the sender is defined in the receiver.

          The non-physical is the organization that evolves among the participating objects, manifesting being outside the collection of the objects (cells) thus organized, synthesizing new objects.

          Jim

          Jim Whitescarver wrote:

          While the Turing machine is not a good practical model for many problem types it is most certainly a general theoretical model of computing.   The implication that is is somehow incomplete is pure quackery.  While a taxidermy of computational models is a worthwhile endeavor that can be enlightening he seems to be mixing analytical techniques with differing degrees of generality.

          On the issue of programming the mind in the laboratory, I do not agree that any new paradigm of computation is essential.  Below I have copied my largely subjective view of the issue, based on a model of the mind developed through meditation, suggesting how is might be emulated.   There is also some discussion and links which may be of interest here. 

          Jim
          >From http://wikiworld.com/Consciousness

          I think consciousness will one day be created in the laboratory. Not next week, or next year. Perhaps in a thousand years. At that time we will know much about it, but it will not answer the ultimate question of why anything at all exists.

          It would be created as an active interconnection of universal self aware (dynamic model of self and environment) entities. Based on todays technology it might be built as neural network simulation running on thousands of machines across the Internet. It might result from the study of life and copying it. It may emerge accidental in the quest for artificial intelligence. It will evolve.

          Self-awareness only requires that an entity include a working model of itself and its environment. Consciousness I believe is self-awareness on a social/collective/composite level. But indeed, the subjective "feeling" of being that goes with consciousness is a mystery that future robots may contemplate but never resolve objectively.  Most "feeling" is chemical.  Somehow the "feeling" of conscious being is something else.

          I believe, being and existence are one. We can make a candle and it comes to be, but though we burn it we cannot erase its being. Consciousness to me is collective being with self awareness.

          Being is existence, collective being is composite connected interacting beings. As we can make or define things, we make beings, we do not put the being in them, it is manifest by virtue of existence.

          I have found that consciousness is like a set of radio programs on many different frequencies. We can tune in at different frequencies through meditation. Most often these channels of consciousness are not aware of each other and have no direct communication but collaborate on building and executing our model of reality. These intelligences within us also collaborate with external intelligences manifesting the spirit of life, and ultimately the spirit of being.

          Usually one frequency, usually somewhere in the beta range, 20-40 cycles per second, dominates the external body action while others maintain and regulate various internal body functions and pursue lower priority mental activities. All but the dominant one are forbidden control and we are not directly aware of them. When a new one takes control, its history becomes our conscious memory and the former moves to the background undetected by us in most cases.

          Q: This figure of 20-40 cycles per second, is it an estimate or a measured frequency, and what is the significance of different frequencies? Are higher frequencies 'more conscious' and lower ones less so? Does this model mean we all have multiple personalities?

          measured (and subjective meditation experience):

          • 0-6 cycles/second delta level - fish and reptiles - heart beat, spirit world, God
          • 7-12 cycles/second alpha level - mammals and dinosaurs - most human bodily functions
          • 13-25 cycles/second beta level - human - vision, movement and speech, spirit, God
          • 25-50000 sound - or brain really humming :)
          • radio
          • heat
          • radar
          • light
          • UV
          • Xray
          • gamma ray

          Complexity and ultimately consciousness can exist at any frequency, lower frequencies seems timeless and all knowing, high frequencies really hum. Indeed there are multiple trains of thought that we are normally unaware of, they are sort of different personalities, but really just parts of your one personality unless their is a conflict between them.

          What we call habit behavior is really conscious but we have no reason to be aware of it or if of what we think we are thinking. Only one train of thought is remembered at one time.

          Intuition and Beliefs

          Difficult problems by definition are non-linear. They do not succumb to linear solutions. They are solvable by techniques such as neural network simulations and EvolutionaryProgramming. It take a non-linear solution to solve a non-linear problem.

          With a linear solution, you can explain the answer, You can say because x and y and z blah blah blah. Non-linear solutions, however, only provide an answer. If the problem is defined correctly the answer will be correct.

          It used to frustrate me immensely that the woman in my life would not listen to reason and refused to give a good reason for their conclusions yet very often turned out to be right.

          Finally I figured out how it works. They build an analogy to the problem in their mind that most men see as irrelevant and they run the model to get the answer and as the model runs and gets better answers they change their mind in what seems to be an arbitrary manner.

          It seems obvious to me now that most all "leaps of intuition" work in a similar manner. These sort of solutions also have the property that they can be easily distributed among separate processors creating a collective intelligence.

          Non-linear models involve feedback mechanisms. various solutions are tried randomly and by heuristic and reinforced or dampened by feedback as the problem runs. There is always a best solution, but it can be overthrown at any time by new data and suddenly the old solution is replaced by the new. The fact that separate parts of the model can run independently as long as feedback is provided, in one mind or many, is independent of the sudden "leap".

          HumanBrain


          Page originally had the spelling of Consciousness, so for the history of this page prior to this Version 1, check there. ---StarPilot, playing SnuckerBug.


          "Consciousness to me is collective being with self awareness." "Self-awareness only requires that an entity include a working model of itself and its environment" - HumanBrain


          I think this article will enlighten you on the difficulty of reducing consciousness to physics:

          Facing Up to the Problem of Consciousness http://jamaica.u.arizona.edu/~chalmers/papers/facing.html

          -- GetReal

          Any system that processes information has some level of conscious experience. <CROSS JOIN> There is no distinction between processing information and manipulating matter/energy.

          Tell me what this means. Did I miss the point of the article? --DavidSiegel


          The connection between information processing and experience: (1) has not been established. A computer that manipulates a bunch of symbols according to a rule book is not necessarily experiencing anything, no more than an abacus being manipulated - is experiencing anything. (2) Even if there are cases where information processing entails experience - that connection is a further fact that needs to be explained.

          That experience would accompany some subset of physical processes instead of those physical processes just working in the dark i.e. without any such experience - is a further fact that needs explaining.

          One could adopt the belief that every physical thing experiences or that every physical interaction entails experience - but that would be an additional proposition to add to physics.

          The great task of physics, chemistry, biology and brain science - is to show how experience arises from mere physical processes, not merely to accept it as an unexplained brute fact.

          -- GetReal


          Science can answer how the machinery of consciousness works. What I've discovered subjectively how is is composed of streams of associations (thoughts/consciousness) on every frequency or channel, and organically left, right, front, back, all brain within a brain within a brain. I expect science can show that this model corresponds to neural network activity of the brain.

          I descended in meditation down to the consciousness of much lower than one cycle per second where beating the heart is conscious. My experience is that nothing is really unconscious and consciousness is the whole myriad of thought streams both currently controlling our body, by direct awareness and by "feeling" their effect on whatever they do control. The stream that seems most important generally takes control and all others continue in the background without our direct awareness as confusion results if we try to give control to more than one.

          If we want to understand this beyond the physical, only the science of memes, so far, is instructive. So that we can differentiate between the understanding of the mechanism and the non-physical, I call the non-physical part "Being", as distinct from consciousness or self awareness, which can be tested for in behavior and explained by physical processes. My understanding is the "being" is social or collective, as it is a distributed meme process.

          The treatment of information in Chalmers link is excellent.

          To borrow a phrase from Bateson (1972), physical information is a difference that makes a difference.

          I would also point out that Bateson defined knowledge as a difference that makes a difference that makes a difference. This is very cool stuff. It shows how fractals of domains of information are layered into information about information, or meta information. In a similar sense, conscious being is the awareness of self awareness. Like a higher mammal recognizing itself in a mirror. It is having a center of being outside oneself.

          The physical world also is composed of information channels on every frequency in the quantum and in every fractal level of organization in information ecologies where by the simple rules of EvolutionaryGameTheory intelligence ultimately emerges dominant. But this intelligence too is collective, not individual, and is seems to me we are part of that collection, the cosmic consciousness.

          --JimScarver


        • jrstern
          ... I disagree with this entirely. I say that computation is about how we can build nice, little machines that can be filled with bits, cranked, and somehow
          Message 4 of 26 , Aug 1 10:16 AM
            --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...> wrote:
            > I, too, think that a revised view of computation is necessary.
            > However, I think along the lines of Marvin Minsky's www.edge.org
            > interview: computation is not only about electronic computers, it
            > is the science and practice of complex systems.

            I disagree with this entirely.

            I say that computation is about how we can build nice, little
            machines that can be filled with bits, cranked, and somehow
            produce new bit patterns that we can interpret interestingly.

            Just what processes are involved in generating sunspots is simply
            not in the same field of study as wondering how it is that we,
            as intelligent agents, can *talk* about sunspots. This disconnect
            between methodologically solipsistic computation and distal,
            ontological truths is crucial to having any conversation at all
            about computation Wittgenstein said so, and it's the rightest thing
            he ever gave us! (That was the "linguistic turn" (LT). As it worked
            out, the *reasons* he gave for LT were basically positivistic and
            skeptical, and, well, for the most part, wrong. When Turing came
            along and talked to Wittgenstein circa 1937, Wittgenstein had every
            opportunity to see how computation offers an entirely different
            rationale for LT, but Wittgenstein rejected it. If only old Ludwig
            could spend a little time with us here in 2004, surrounded by
            computers, I think he would have to change his tune!)

            > Thus, it applies equally to natural systems as well as
            > man-made. For an example of the elegance of computational approach
            > for studying biological systems, a nice example is the recent
            > theoretical trends in Amorphous Computing and Cellular Automata.
            > Needless to say, computational sciences have shown remarkable
            > utility in fields quite distinct from computing itself, and is
            > continuing to become a major application.

            There are probably many links between computation and biology, as
            all living systems are encoded in easily observed strings of nucleic
            acids and such. Please note this is a true physical encoding, and
            not just a stance we take! There is much we can learn about
            computation by looking at biological systems.

            Yet, I take a tendentious methodological
            stance on all of this. I say that if we choose to limit ourselves
            just to GOFAI, basically symbolic computation as done on electronic
            digital computers, we lose nothing in generality in any conclusions
            we reach. Further, I limit the discussion to matters of "cognition",
            ignoring qualia, consciousness, free will, etc. I have speculations
            concerning these other matters, but do not wish to engage them.
            Please note that anything I post here, will be from this perspective.
            My only project is analyzing, correcting, and rehabilitating a form
            of strong AI that is, to all intents and purposes, GOFAI,
            Release 2.0

            > With respect to Smith and Sloman, I think they underrate the
            > importance of Turing's (and others') theoretical contributions. The
            > Turing Machine is not just a machine, it is a model of computation.
            > Smith says there can be no theory of computation, but in reality we
            > have quite powerful theories of computation, and indeed, we have
            > theories of fundamental concepts *based* on computation such as
            > Algorithmic Information Theory and Algorithmic Probability Theory.

            I am in great sympathy with everything that Smith and Sloman are
            attempting, but they make a mistake in dismissing Turing computation,
            which needs to be better understood. I suggest a historical study of
            Hilbert, Brouwer, Frege, Wittgenstein, Tarski, and a few others,
            shines new light on what Turing did. Seeing it requires the kind
            of 20-20 hindsight we can have today, with our empirical knowledge
            of computers-in-the-wild, as Smith says.

            > There are so many claims in Smith's article that it would take four
            > times the length of his article to refute them. His philosophy is,
            > in my opinon, of the sort "Look, we don't exist" kind of
            > unnecessary skepticism....

            Yes, and that is exactly why eliminativism about computation is
            wrong, and even Minsky's claim that computation = science is wrong.
            Computation is a separable domain of study. In fact, I believe that
            it is a domain with roots over two thousand years old - computation
            is, in my reading, a superset of the domain of the study
            of language. That is, I am very sympathetic, again methodologically
            (at least), with the "language of thought" (LOT) hypothesis.
            My gosh, if digital electronic computers do not instantiate LOT,
            then what the heck do they do? (I know that is not quite as easy
            a question as it looks like, but it's a starting point).

            In fact, Minksy posted a message a few months back regarding his
            meetings as a grad student with Quine, and how Minsky, back then,
            was trying (unsuccessfully) to convince Quine that computation was
            a process not implicit, inherent, or immanent in logical or physical
            systems. That's my probably overbroad interpretation of what
            Minsky posted, but I still see it as in conflict with what he said
            in that Edge piece, if you cite it correctly.

            > For instance, let's take the "Seven Construals of Computation" in
            > Smith's article. It seems that Smith tries to distinguish
            > conceptions of computation, attempting to define them, and then
            > criticize them based on his own definitions, which I believe
            > do not reflect the true spirit of "old" computationalism.

            I am very sympathetic to these lists of important features. I have
            a couple of my own. The best one has about fifteen entries. Mine
            are much more basic. My #1 is that computation is a physical
            process. I know that right there I go into conflict with the
            majority current view, but it can be made to work (in fact, it is
            the ONLY thing that can be made to work, as I hope to eventually
            show!), but I don't have to invent anything major to say so, I just
            have to subscribe to a particularist form of nominalism to make
            this case, and to reject any hint of platonism. In fact, noting
            and remembering that computation is physical is the best tool I have
            found for detecting and eschewing platonism.

            > Let's have a look at the first such "construal":
            > 1. formal symbol manipulation (fsm): the idea, derivative from a
            > century's work in formal logic and metamathematics, of a machine
            > manipulating symbolic or (at least potentially) meaningful
            > expressions without regard to their interpretation or semantic
            > content;
            >
            > Is this really the case?

            No!

            I endorse the "symbol" part, I would qualify the "manipulation" part,
            and question or reject the "formal" part. Smith here is referring
            to what Fodor (and I) like to refer to as methodological solipsism,
            but I for one prefer to view this in causal terms, not formal ones.
            Fodor with his syntactic essentialism does subscribe to something
            like the formal view and it ultimately fails him, or at least he
            thinks it does!

            In any case, as above, work on logic is not prior to work on
            language.

            > And I am very suspicious of Smith's expressions like the following:
            >
            > Turing machines, notions of "effective computability", and the
            > like--fails as a theory of computing, in spite of its name and
            > its popularity. It is simultaneously too broad, in applying to more
            > things than computers, and too narrow, in that it fails to apply
            > to some things that are computers. More seriously, what it is
            > a theory of, is not computing.
            >
            > Here Smith turns this serious subject into a talk show. How is
            > it too broad? Was not one of the original things we wanted
            > to show was how general computation is in real life?

            Yes, exactly!!!

            > How is it too narrow? The theories Smith refers to are all
            > theories of discrete computation, and there is no aspect of
            > discrete computation that they fail to address.

            Oh, I suppose I could give a sympathetic reading to what Smith
            is saying, but it still comes out wrong in the end.

            I'm sure that Smith and Sloman and I are all groping the same
            elephant, (you all know the parable of the blind men and the
            elephant, right?) but I think I have the better view!
            Not that I've published it yet.

            > Sloman also seems out of focus in this issue. He seems to think
            > that things like interrupts are a very crucial part of computation
            > that the theory does not address. I think it is a bit premature
            > to say that interrupt handling facility (both system and user
            > program side) is not just a small program.

            Again, what I see in this point from Sloman is a desire to show the
            embedding of computational systems in the world, to show that they
            are NOT purely formal and NOT entirely solipsistic. But he doesn't
            unpack the ideas far enough to make his case.

            > Have I successfully incited a discussion?

            Too soon to tell. :)

            Josh
          • jrstern
            ... Outstanding! Five year ago, ... well, I guess I was early to the vague idea that the underlying process of computation needed reexamination, but somewhat
            Message 5 of 26 , Aug 1 10:45 AM
              --- In ai-philosophy@yahoogroups.com, "thoughtfuldavid" <david@b...>
              wrote:
              > I should say that I was actually priviledged to attend (and present
              > a paper at) the conference-workshop from which the works in this
              > book were derived -except for Aaron Sloman's. Unfortunately, it was
              > 5 years ago so I am not sure how much I can recall now! Anyway,
              > below are a few initial comments on what you and Eray have posted
              > so far. I look forward to further discussions.

              Outstanding!

              Five year ago, ... well, I guess I was early to the vague idea that
              the underlying process of computation needed reexamination, but
              somewhat late in clarifying the problems. We'll just have to see
              if I can be first over the line with some new answers!

              > > Modestly titled, "Foundations of Computing", the basic idea is
              > > that in order to understand how computation can underlay
              > > cognition, we need to first understand computation. Smith
              > > basically tells us that everything we knew about computation
              > > was wrong.
              >
              > I am not sure Smith is saying that what we know is wrong, rather we
              > do not have a clear understanding of what computation actually is.

              In the interests of brevity I engaged in a bit of hyperbole - but
              only a bit, since Smith goes on to trash Turing computation and
              propose an eliminative and universalistic view of computation which
              indeed *does* conflict both with what I was taught in school thirty
              years ago and with what I believe today.

              > Different people think of it in different ways as outlined in the
              > six (7, 10, 11?) construals that he mentions. His goal has been to
              > identify a single neat "Theory of computation" but by his own
              > admission, he has been unable to do so.

              If he requires it to be neat, he may be disappointed. That does
              not exclude the possibility that there is a clear and singular
              theory, but with some minimal level of complexity. It turns out
              that quantum physics has, um, however many basic particles as it
              has, but that number does not seem to be one, and nothing about it
              is all that simple!

              > Computationalism is saying (very loosely) something to the effect
              > that the "mind is a computer". It was always tacitly assumed that
              > we knew what a computer was. Suddenly however, it seems we are
              > less than clear about what a computer/computation actually is,
              > which makes the project of understanding the mind in computational
              > terms all the more difficult.

              Yes, but at the same time it provides new hope! Perhaps once
              we understand what even basic (small 'b') computation is really
              all about, it will disolve some old problems of computationalism.

              > > With respect to Smith and Sloman, I think they underrate the
              > > importance of Turing's (and others') theoretical contributions.
              > > The Turing Machine is not just a machine, it is a model of
              > > computation.
              >
              > The significance of Turing's work to computation is beyond doubt.

              Looks like you and Eray and I pretty much agree on that,
              contra Smith and Sloman.

              > Surely, what Smith is saying is that there is more to computation
              > in-the-wild than Turing's theoretical contributions indicate (sort
              > of like knowing all about atoms does not tell you everything about
              > water -for instance, that it is mixed with coffee and drunk by
              > humans.)

              It is not at all clear to me that Smith leaves any significant room
              for Turing's work. And, y'know, don't call me Shirley.

              > As for Sloman, it seems to me that he is merely pointing
              > out that knowing these theoretical aspects of computing will not
              > help in the quest to understand mind.

              No, I think that is not a correct reading.

              Remember, Sloman here is attempting a revisionist explanation
              of mundane (my favorite term for it) computing,
              NOT computationalism!

              > Again, Turing's theory is so
              > far abstracted from real-world considerations as to make it
              > largely irrevelant to our present concerns -
              > AI & Cognitive Science.
              > [Eray, surely the Turing Machine is not really a machine since
              > it cannot be built!]

              Please see my response to Eray. I cannot speak for him, he may
              agree with you here, but I do not. For me, a Turing machine is just
              as instantiatable as a Buick. In fact, even more so, for one can
              talk of Buicks in ideal terms, and I do not believe that anything
              interesting about computation can be determined without examining
              physically realizable (in principle) systems.

              That's a constructivist viewpoint. If it means I have to deflate
              the importance of the halting problem, so be it. As with Sloman,
              I don't see any relevance of that to the problems of mundane
              computation or exalted AI. Only a platonist would care, and
              I am not one of those.

              I do realize how far from the mainstream this position puts me,
              but it is a much more modest position simply to recognize Turing
              machines as physically realizable. Remember, the basic Turing
              machine can be quite simple, is not even a UTM, and while it may
              have an infinite tape in principle, large classes of instances need
              finite and very modest amounts of "tape".

              Finally, I do not really understand how you (or others) can
              criticize Turing's theory as "abstracted". Do you believe that
              the binary structure of a Pentium microprocessor is so abstracted
              from real-world considerations that it is theoretically irrelevant?
              I don't. Remember, I'm a raving physicalist. This makes me a
              charter member of the reductionist club, too. I need to explain
              just how meaningfulness comes to relate to bits. And, I suggest,
              so does everyone who cares to talk about computation
              or computationalism.

              And maybe you shouldn't call Eray "Shirley", either.

              --

              If anyone likes, I think the most fun I could have out of this
              discussion would be to cite and criticize the later parts of
              Smith's paper, in which I believe he completely flies
              off the tracks!

              Josh
            • thoughtfuldavid
              Josh, you wrote, ... Turing Machines have -by definition- infinite tapes and thus (so far as I am aware) are impossible to implement, unlike Buicks. ... I find
              Message 6 of 26 , Aug 1 2:43 PM
                Josh, you wrote,

                > For me, a Turing machine is just
                > as instantiatable as a Buick.

                Turing Machines have -by definition- infinite tapes and thus (so far
                as I am aware) are impossible to implement, unlike Buicks.

                > In fact, even more so, for one can
                > talk of Buicks in ideal terms, and I do not believe that anything
                > interesting about computation can be determined without examining
                > physically realizable (in principle) systems.

                I find it difficult to square this with your earlier statement about
                the value of Turing's (purely theoretical) work.

                > I don't see any relevance of that to the problems of mundane
                > computation or exalted AI.
                > ...
                > I do realize how far from the mainstream this position puts me,
                > but it is a much more modest position simply to recognize Turing
                > machines as physically realizable. Remember, the basic Turing
                > machine can be quite simple, is not even a UTM, and while it may
                > have an infinite tape in principle, large classes of instances need
                > finite and very modest amounts of "tape".
                >
                > Finally, I do not really understand how you (or others) can
                > criticize Turing's theory as "abstracted". Do you believe that
                > the binary structure of a Pentium microprocessor is so abstracted
                > from real-world considerations that it is theoretically irrelevant?
                > I don't. Remember, I'm a raving physicalist. This makes me a
                > charter member of the reductionist club, too. I need to explain
                > just how meaningfulness comes to relate to bits. And, I suggest,
                > so does everyone who cares to talk about computation
                > or computationalism.
                >

                Again, as I read Smith and Sloman, they are both saying (in effect)
                that there is much more to computation than Turing Machine
                formalisms alone would appear to suggest. When I refered to Turing's
                theory being "abstracted from real world considerations" this was
                not intended as a criticism, but a statement of fact. The formalism
                says nothing of power constraints, timing, etc., for example. As I
                see it we are all saying the same thing (just not understanding each
                other!)

                I too certainly believe the mind has a physical basis and, yes, we
                need to explain how certain parts of the world can meaningfully
                refer to other parts. In particular, we need to understand the
                difference between cognition and computation. The suggestion (or
                mine anyway) is that this distinction can be made based on the forms
                of reference they involve.
              • jrstern
                ... This is a common and pernicious misinterpretation of Turing s work. It is my view that Turing needs to be better understood. If Smith and Sloman agree
                Message 7 of 26 , Aug 1 5:35 PM
                  --- In ai-philosophy@yahoogroups.com, "thoughtfuldavid" <david@b...>
                  wrote:
                  > Josh, you wrote,
                  >
                  > > For me, a Turing machine is just
                  > > as instantiatable as a Buick.
                  >
                  > Turing Machines have -by definition- infinite tapes and thus
                  > (so far as I am aware) are impossible to implement, unlike Buicks.

                  This is a common and pernicious misinterpretation of Turing's work.
                  It is my view that Turing needs to be better understood. If Smith
                  and Sloman agree that computation needs to be better understood,
                  and I agree, I appear to differ from them by suggesting this is
                  best begun by reexamining and better understanding Turing.

                  The significance of Turing's 1936 paper was not after all in what
                  it told us about computable numbers, but in the structural framework
                  Turing invented to make his argument. Maybe he was preceeded by
                  Babbage, and maybe one can argue to the point by other means, but
                  the basic idea was that a very large class, possibly universal,
                  of all computational functions, can be reduced to bits.

                  "The machine is supplied with a "tape", (the analogue of paper)
                  running through it, and divided into sections
                  (called "squares") each capable of bearing a "symbol"."

                  No mention of "infinite".

                  I suggest that all interesting computations can be completed
                  with a finite tape. How finite? Well, my current workstation
                  has 1.6 gigabits of RAM and 640 gigabits of disk. Give me
                  something ten or a hundred or a trillion times the scale and
                  we have a finite machine of quite significant capabilities.
                  In fact, give me something three orders of magnitude *smaller*,
                  and we have something of sufficient magnitude to cover all
                  interesting arguments about cognition.

                  I have long taken the position that computational AI is not even
                  a matter of scale. I strongly suspect that once we have figured
                  out the science and philosophy and mechanics, human-scale AI's
                  will be realizable on workstations a tenth of the size and a tenth
                  of the speed of the one I'm typing on right now.

                  A "Turing machine" is any machine, even with a few states and a
                  finite tape, that performs the sequential operations of a state
                  machine. While infinities *may* be involved, there is certainly
                  no requirement for them. As a matter of definition, the claim that
                  Turing machines cannot be built is just wrong.

                  The motivation for the claim is more interesting, it is a remnant
                  and rampant urge to idealism, which I assert will always prove to
                  be both a needless assumption and the source of theoretic failures.

                  I mentioned this on c.a.p. a couple of months ago, but Searle has
                  a new essay in a new book, "Views Into the Chinese Room". In this
                  essay, Searle takes a bit of a revisionary view of his own work,
                  and asserts this very point, that computation defined as an ideal,
                  builds into itself the ineliminable assumption that it is simply
                  the wrong kind of thing out of which to build intelligence for a
                  material being. It is as if people have an urge to preserve and
                  repeat the dualistic problems of Cartesianism. I completely agree
                  with Searle on this point, but I disagree with his premise -- MY
                  computation is not ideal, it is particular and physical. And I do
                  not see that Turing made this mistake, no matter how many of his
                  modern followers do.

                  For that matter, let us not insult the Buick. What, after all,
                  is the Buick? It is no doubt that which instantiates the full
                  wonderfulness of Buick. The problem, as Berkeley noted, is that
                  there are essential differences between any type and its tokens.
                  My red Buick does not capture all the wonderfulness of your blue
                  Buick. Therefore, Buicks cannot really be instantiated. QED,
                  not. This logic works the same for Buicks and Turing machines.

                  > > In fact, even more so, for one can
                  > > talk of Buicks in ideal terms, and I do not believe that
                  > > anything interesting about computation can be determined
                  > > without examining physically realizable (in principle) systems.
                  >
                  > I find it difficult to square this with your earlier statement
                  > about the value of Turing's (purely theoretical) work.

                  Perhaps the above will help.

                  > Again, as I read Smith and Sloman, they are both saying (in effect)
                  > that there is much more to computation than Turing Machine
                  > formalisms alone would appear to suggest.

                  Well certainly, I do not think even Turing could disagree with that!
                  But I see what Turing did do as foundational to virtually any
                  discussion of these enlarged theories of computation, while they
                  seem to think he missed some basic essential feature.

                  I guess I believe there is a form of reduction that takes everything
                  back down to bits, and I can see no real distance between any bit
                  theory and what Turing has already given us.

                  >When I refered to Turing's
                  > theory being "abstracted from real world considerations" this was
                  > not intended as a criticism, but a statement of fact. The
                  > formalism says nothing of power constraints, timing, etc., for
                  > example. As I see it we are all saying the same thing (just not
                  > understanding each other!)

                  I am being coy about this, but I really do mean to say something
                  else. I take an entirely physicalist, particularist, and
                  INSTRUMENTALIST view of computation (these all being options
                  within the realm of nominalism).

                  You said:
                  > Again, Turing's theory is so
                  > far abstracted from real-world considerations as to make it
                  > largely irrevelant to our present concerns -
                  > AI & Cognitive Science.

                  I would say something like:

                  No theory of computation is complete or even interesting,
                  until and unless it provides a reduction down to the level
                  of bits and states, as does Turing's. In fact, Turing has
                  already done this work for us, so let us understand it fully
                  and work to build a complete theory on this foundation.

                  Reduction is not abstraction, I guess that's what I'm saying.
                  How does that strike you?

                  > I too certainly believe the mind has a physical basis and, yes,
                  > we need to explain how certain parts of the world can
                  > meaningfully refer to other parts. In particular, we need to
                  > understand the difference between cognition and computation.
                  > The suggestion (or mine anyway) is that this distinction can be
                  > made based on the forms of reference they involve.

                  Mmmmm, I don't know.

                  I've come to hold YET ANOTHER tendentious position, which is
                  that while there may be some differences between computation and
                  cognition, there is also some significant overlap (picture a Venn
                  diagram of two circles overlapping, I suggest the shared area
                  is quite large). The question of the day is about what
                  computation is, short of cognition. We have to answer that before
                  we can draw the diagram accurately.

                  Question: does computation, short of cognition, involve reference?

                  I suggest it does, but in a rather weak and perhaps theoretically
                  novel way. We write programs in high-level languages that certainly
                  look as if they involve reference, insofar as one wants to say that
                  any linguistic forms involve reference. This is one of those
                  aspects of computing in-the-wild that we want to account for.
                  But, if mundane computation involves this reference, then what,
                  in terms of reference, remains to comprise cognition?

                  I don't think there is any short answer to the above, but there
                  may be a long answer!

                  Josh
                • Jim Whitescarver
                  ... Buicks use an indeterminant amount of gas, just as any problem can only use a finite finite tape in finite time. You can in principle add tape as needed
                  Message 8 of 26 , Aug 1 5:41 PM
                    thoughtfuldavid wrote:

                    >Turing Machines have -by definition- infinite tapes and thus (so far
                    >as I am aware) are impossible to implement, unlike Buicks.
                    >
                    Buicks use an indeterminant amount of gas, just as any problem can only
                    use a finite finite tape in finite time. You can in principle add tape
                    as needed as I believe Turing pointed out.

                    But you are correct that is is not intended to be a practical machine,
                    just a general one. There are no more general machines despite a lot of
                    recent garbage suggesting that there is without providing any
                    theoretical foundation for that claim.

                    Jim
                  • Eray Ozkural
                    ... The instantaneous description (ID) of a TM is finite at all times. The infinite part of the tape has constant algorithmic complexity (it is all blank
                    Message 9 of 26 , Aug 1 6:02 PM
                      On Monday 02 August 2004 00:43, thoughtfuldavid wrote:
                      > Josh, you wrote,
                      >
                      > > For me, a Turing machine is just
                      > > as instantiatable as a Buick.
                      >
                      > Turing Machines have -by definition- infinite tapes and thus (so far
                      > as I am aware) are impossible to implement, unlike Buicks.

                      The instantaneous description (ID) of a TM is finite at all times. The
                      "infinite" part of the tape has constant algorithmic complexity (it is all
                      blank symbols), therefore it does not have to be taken into account in a
                      constructivist interpretation.

                      It's somewhat elegant to say that the blank portion of the tape, after the
                      last symbol described in ID is "infinite", but the ID itself is finite, and
                      is sufficient to implement a TM in the physical world.

                      In theory of computation texts, it is argued for this point of view rather
                      than the other alternative which supposes that TMs exist only in theory and
                      not in practice. I can give an exact reference if needed :)

                      I think Aaron Sloman might disagree with this point of view, but I believe my
                      above explanation consists of mathematical statements, instead of
                      philosophical ones.

                      Regards,

                      --
                      Eray Ozkural (exa) <erayo@...>
                      Comp. Sci. Dept., Bilkent University, Ankara KDE Project: http://www.kde.org
                      http://www.cs.bilkent.edu.tr/~erayo Malfunction: http://malfunct.iuma.com
                      GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C
                    • jrstern
                      ... Please do, it wouldn t hurt if you have two or three! ... I m not certain I know the difference, but it seems to me that the mathematicians I run into, or
                      Message 10 of 26 , Aug 1 6:57 PM
                        --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...> wrote:
                        > In theory of computation texts, it is argued for this point of view
                        > rather than the other alternative which supposes that TMs exist
                        > only in theory and not in practice. I can give an exact reference
                        > if needed :)

                        Please do, it wouldn't hurt if you have two or three!

                        > I think Aaron Sloman might disagree with this point of view, but I
                        > believe my above explanation consists of mathematical statements,
                        > instead of philosophical ones.

                        I'm not certain I know the difference, but it seems to me that the
                        mathematicians I run into, or people who think they are
                        mathematicians, tend to the abstract viewpoint on TMs, which is odd,
                        since they also tend to platonism about mathematics - and the world.

                        I believe one or two of the SEP entries on and about computation
                        refer to the abstract position, and I've had some discussions with
                        them about that.

                        Again, Searle found the abstract theory so widespread, he was able to
                        base his Chinese Room and Wordstar parables on just that point.

                        Again, Sloman's paper gives an argument why the halting problem does
                        not matter for AI, and while I might prefer some slightly different
                        arguments, I share the motivation and something like the conclusion.
                        I'd say the only need for infinite (non-constructive) arguments re
                        Turing machines is just that, the halting problem - or any other
                        problem you want to reduce to computational, which involves infinite
                        sets.

                        Josh
                      • thoughtfuldavid
                        Josh, you wrote that, ... framework ... My understanding is that the intellectual significance of Turing s work lies precisely in what it says about computable
                        Message 11 of 26 , Aug 2 12:11 AM
                          Josh, you wrote that,

                          > The significance of Turing's 1936 paper was not after all in what
                          > it told us about computable numbers, but in the structural
                          framework
                          > Turing invented to make his argument. Maybe he was preceeded by
                          > Babbage, and maybe one can argue to the point by other means, but
                          > the basic idea was that a very large class, possibly universal,
                          > of all computational functions, can be reduced to bits.

                          My understanding is that the intellectual significance of Turing's
                          work lies precisely in what it says about computable numbers. It is
                          seen as an equivalent formulation of proofs by Godel and Church(?)
                          concerning recursive number theory. The "machine" has to have an
                          infinite tape otherwise there is no proof. In essence, the Turing
                          Machine defines an important class of computational devices distinct
                          from the FSA (Finite State Automata -of which our PC's are prime
                          examples however much memory they may have!) It is simply not true
                          that all computational functions can be reduced to bits (the value
                          of pi being an obvious example.) [Note: The former may be mistaken
                          since I am decidedly not a mathematician or expert on this topic.]

                          Later in the same post you say,

                          > Reduction is not abstraction, I guess that's what I'm saying.
                          > How does that strike you?

                          True, essentially one is bottom-up, the other top-down. Obviously
                          there are other differences too, but let's not argue that now. Much
                          more interesting is your following statement,


                          > I've come to hold YET ANOTHER tendentious position, which is
                          > that while there may be some differences between computation and
                          > cognition, there is also some significant overlap (picture a Venn
                          > diagram of two circles overlapping, I suggest the shared area
                          > is quite large). The question of the day is about what
                          > computation is, short of cognition. We have to answer that before
                          > we can draw the diagram accurately.

                          Smith and I drew the Venn diagram with the cogitive systems circle
                          (set) completely enclosed by the computational systems circle,
                          rather than simply intersecting with it. That is to say that all
                          cognitive systems are computational. This is the computationalist
                          claim (I think.)

                          > Question: does computation, short of cognition, involve reference?

                          Yes (and no!) Computational devices involve only indirect reference
                          obtained via a cognitive agent. Without the agent it would not
                          involve reference (or even be a computer!) I don't want to muddy the
                          waters, but I think this is what Searle sees as observer-relative. A
                          rock is a chair only if someone so uses it. Similarly, the orbits of
                          the planets are a computer only if someone so uses them (which, of
                          course, does not detract from their other functions, such as
                          providing us somewhere to live!)
                        • Jim Whitescarver
                          Turing s machine must be considered independent of the rest of Turing s work if it is the generality of the machine that is at issue. Certainly any associative
                          Message 12 of 26 , Aug 2 8:25 AM
                            Turing's machine must be considered independent of the rest of Turing's
                            work if it is the generality of the machine that is at issue.

                            Certainly any associative memory or references scheme that may be
                            conceived of can be modeled and run on a Turing machine. It is a
                            universal machine on which we may construct anything logically
                            constructible since it can run any program.

                            The limit of the Turing machine is that it cannot make an infinite
                            construction in finite time.

                            Thus the issue of completness becomes a question of whether infinite
                            constructions preexist since they certainly not have been constructed.

                            It would seem to me that if you define what preexists and add a
                            universal machine to define all that may be constructed from it you can
                            achieve no greater universality.

                            As far as halting goes, that is death to an information system. The
                            halting problem is uninteresting because most interesting algorithm give
                            better answer over time and never halt. I contend that the halting
                            problem is itself self-referential and unanswerable, by analogy to
                            Russel's paradox . In this respect, Turing was wrong to apply proof by
                            induction since not true does not mean false for an unanswerable
                            proposition. All he showed, in my view, is that Platonic analysis of
                            universal systems will yield Platonic results. It does not illuminate
                            the nature of universal machines or information systems in general.

                            The issue is software, not hardware, it is the program, not the machine
                            that is at issue here. While the software must ultimately be manifest
                            physically, it is not necessarily the machine. The machine may be
                            conceived of as any universal machine.

                            Cognition, in so far as it can be described, may be expressed as
                            software, and the software can run on any universal computer. The
                            significance of the Turing machine is that we can show universality by
                            equivalence to it. That is it's only useful function and only significance.

                            Jim
                            thoughtfuldavid wrote:

                            >Josh, you wrote that,
                            >
                            >
                            >
                            >>The significance of Turing's 1936 paper was not after all in what
                            >>it told us about computable numbers, but in the structural
                            >>
                            >>
                            >framework
                            >
                            >
                            >>Turing invented to make his argument. Maybe he was preceeded by
                            >>Babbage, and maybe one can argue to the point by other means, but
                            >>the basic idea was that a very large class, possibly universal,
                            >>of all computational functions, can be reduced to bits.
                            >>
                            >>
                            >
                            >My understanding is that the intellectual significance of Turing's
                            >work lies precisely in what it says about computable numbers. It is
                            >seen as an equivalent formulation of proofs by Godel and Church(?)
                            >concerning recursive number theory. The "machine" has to have an
                            >infinite tape otherwise there is no proof. In essence, the Turing
                            >Machine defines an important class of computational devices distinct
                            >from the FSA (Finite State Automata -of which our PC's are prime
                            >examples however much memory they may have!) It is simply not true
                            >that all computational functions can be reduced to bits (the value
                            >of pi being an obvious example.) [Note: The former may be mistaken
                            >since I am decidedly not a mathematician or expert on this topic.]
                            >
                            >Later in the same post you say,
                            >
                            >
                            >
                            >>Reduction is not abstraction, I guess that's what I'm saying.
                            >>How does that strike you?
                            >>
                            >>
                            >
                            >True, essentially one is bottom-up, the other top-down. Obviously
                            >there are other differences too, but let's not argue that now. Much
                            >more interesting is your following statement,
                            >
                            >
                            >
                            >
                            >>I've come to hold YET ANOTHER tendentious position, which is
                            >>that while there may be some differences between computation and
                            >>cognition, there is also some significant overlap (picture a Venn
                            >>diagram of two circles overlapping, I suggest the shared area
                            >>is quite large). The question of the day is about what
                            >>computation is, short of cognition. We have to answer that before
                            >>we can draw the diagram accurately.
                            >>
                            >>
                            >
                            >Smith and I drew the Venn diagram with the cogitive systems circle
                            >(set) completely enclosed by the computational systems circle,
                            >rather than simply intersecting with it. That is to say that all
                            >cognitive systems are computational. This is the computationalist
                            >claim (I think.)
                            >
                            >
                            >
                            >>Question: does computation, short of cognition, involve reference?
                            >>
                            >>
                            >
                            >Yes (and no!) Computational devices involve only indirect reference
                            >obtained via a cognitive agent. Without the agent it would not
                            >involve reference (or even be a computer!) I don't want to muddy the
                            >waters, but I think this is what Searle sees as observer-relative. A
                            >rock is a chair only if someone so uses it. Similarly, the orbits of
                            >the planets are a computer only if someone so uses them (which, of
                            >course, does not detract from their other functions, such as
                            >providing us somewhere to live!)
                            >
                            >
                            >
                          • Eray Ozkural
                            Hello, David. Sorry for responding a little late. I hope the heat of discussion has not dropped below the reply threshold, yet. ... Thanks for participating in
                            Message 13 of 26 , Aug 2 9:04 AM
                              Hello, David.

                              Sorry for responding a little late. I hope the heat of discussion has not
                              dropped below the reply threshold, yet.

                              On Sunday 01 August 2004 15:17, thoughtfuldavid wrote:
                              > Hi everyone, I have been a silent observer of this group for a
                              > while, but not really had the time (or the background) to contribute
                              > to the discussion so far. However, I am now on "vacation" and (as
                              > Eray knows) this particular topic is very dear to my heart, so, yes
                              > please Josh, I would love to spend a while discussing it.
                              >
                              > I should say that I was actually priviledged to attend (and present
                              > a paper at) the conference-workshop from which the works in this
                              > book were derived -except for Aaron Sloman's. Unfortunately, it was
                              > 5 years ago so I am not sure how much I can recall now! Anyway,
                              > below are a few initial comments on what you and Eray have posted so
                              > far. I look forward to further discussions.

                              Thanks for participating in the discussion. I am pleased that I've been able
                              to bring together some of the most interested people on the net.

                              > > Modestly titled, "Foundations of Computing", the basic idea is that
                              > > in order to understand how computation can underlay cognition, we
                              > > need to first understand computation. Smith basically tells us that
                              > > everything we knew about computation was wrong.
                              >
                              > I am not sure Smith is saying that what we know is wrong, rather we
                              > do not have a clear understanding of what computation actually is.
                              > Different people think of it in different ways as outlined in the
                              > six (7, 10, 11?) construals that he mentions. His goal has been to
                              > identify a single neat "Theory of computation" but by his own
                              > admission, he has been unable to do so.
                              >
                              > Computationalism is saying (very loosely) something to the effect
                              > that the "mind is a computer". It was always tacitly assumed that we
                              > knew what a computer was. Suddenly however, it seems we are less
                              > than clear about what a computer/computation actually is, which
                              > makes the project of understanding the mind in computational terms
                              > all the more difficult.

                              Yes, this seems to be one of Smith's criteria for a theory of computation:
                              that it performs as a foundation for a computationalist view of the mind.

                              His criteria (empirical, conceptual, cognitive) seem indeed necessary for a
                              theory of computation, but I think his analysis is (very) flawed. We really
                              should get down to specific points for discussing that, of course.

                              > > With respect to Smith and Sloman, I think they underrate the
                              > > importance of Turing's (and others') theoretical contributions.
                              > > The Turing Machine is not just a machine, it is a model of
                              > > computation.
                              >
                              > The significance of Turing's work to computation is beyond doubt.
                              > Surely, what Smith is saying is that there is more to computation in-
                              > the-wild than Turing's theoretical contributions indicate (sort of
                              > like knowing all about atoms does not tell you everything about
                              > water -for instance, that it is mixed with coffee and drunk by
                              > humans.)

                              Yes, but that seems to be a social theory of water, rather than a foundational
                              theory. In foundational terms, quantum chemistry can explain (in principle)
                              why it is possible at all to mix coffee with water, and why it is consumable
                              by humans.

                              Likewise, I think, theory of computation can explain, why Microsoft Word works
                              at all. That is the relevant foundational question, in my opinion. (It meets
                              Smith's empirical criteria, for instance.)

                              Which foundational question would you suspect is left out by our current
                              theory of computation? Is there any conceptual loop-hole?

                              > As for Sloman, it seems to me that he is merely pointing
                              > out that knowing these theoretical aspects of computing will not
                              > help in the quest to understand mind. Again, Turing's theory is so
                              > far abstracted from real-world considerations as to make it largely
                              > irrevelant to our present concerns -AI & Cognitive Science. [Eray,
                              > surely the Turing Machine is not really a machine since it cannot be
                              > built!]

                              Well, I think it can be built insofar any machine blueprint can be built. It's
                              just not a particularly efficient architecture, or an easy way of
                              programming. TM is a very low-level machine code, not very fitting for humans
                              to deal with (at least at a conscious level...), thus the intuitive
                              "resistance" to the view that it is a real machine like a screwdriver.

                              Defining the screwdriver, for instance, I could say that, it can be in theory
                              used by a human to drive an unbounded number of screws, one by one. Which
                              might look neat on a textbook. Does that make the screwdriver an impossible
                              machine? (Hint: it's all in the head!)

                              In theory texts, for instance, it is said that the RAM model of computation is
                              equivalent in computational power to the Turing Machine. Do you disagree with
                              that point of view?

                              > Perhaps I should stop at this point and ask whether that helped
                              > clarify things a bit or merely served to confuse them more?

                              I think it makes the questions clearer.

                              > Eray- you question Smith's first construal, the idea of FSM (Formal
                              > Symbol manipulation). It seems reasonable to me. Can you perhaps say
                              > exactly do you think is wrong with it?

                              I think the first part is reasonable, that computation is manipulation of
                              symbols via formal rules. That is exactly what a TM is mathematically.
                              However, the part about "a machine manipulating at least potentially
                              meaningful expressions without regard to their interpretation or semantic
                              content" is misleading. This seems to be Searle's interpretation of digital
                              computation, that it does not have semantics by itself: ie. semantics is
                              somewhere else to be found.

                              So, Smith seems to give us the opinion of an AI opponent, as a commonly
                              accepted definition (construal) of computation. I don't think that's a good
                              way of arguing. If you accept that, you are going to conclude that this view
                              does not meet "conceptual" and "cognitive" criteria because it disregards
                              semantics by definition. (as Smith himself does)

                              On the other hand, Smith is being inconsistent. First, he declares "semantics"
                              subject as recalcitrant. He also says that it is not possible to claim that
                              semantics arises from computation because it will be either circular or
                              assuming some biases like mind-body dualism. And *then* criticizes FSM view
                              because it doesn't talk about semantics.

                              I think that is very muddy thinking. No wonder why it took him 30 years to
                              come up with a, well, completely misled account of computation. He seems to
                              be more like a net.kook than a philosopher by my standards of brevity,
                              clarity, consistency and rationality.

                              Let's see more of what I think is a wrong view of FSM concept.

                              Smith first says: "It (FSM) explicitly characterizes computing in terms of a
                              semantic or intentional aspect, if for no other reason than that without some
                              such intentional character there would be no warrant in calling it symbol
                              manipulation", and goes on to defend his confused thinking in the footnote.

                              First, "symbol" does *not* mean intentionality. A symbol is just that, a sign.
                              It may refer to other things, but those references are resolved in *minds*,
                              there is no such thing as the reference, or intentionality of a symbol
                              without a mind to resolve it. (We can think of it as pointer dereferencing)
                              That is, when the wind blows over the symbol, or sun shines on top of it, it
                              does not meet a reference! Only the symbol! (Reference is *constructed*) So,
                              it's quite confused to think that FSM defines computation in terms of
                              intentions. Quite the opposite. It defines computation in objective terms.
                              The symbols in the computation could mean something to an observer, or not.
                              That is not the problem. (Example: how much do you understand when you look
                              at the hexadecimal memory dump of a binary computer program?) The problem is
                              whether the symbols can ever have inherent intentionality from the point of
                              view of *computation* itself. (And I think the answer to that question is an
                              echoing YES)

                              Smith then says:
                              "The fsm construal has a distinctly antisemantical flavor, owing to its claim
                              that computation is the "manipulation of symbols independent of their
                              semantics." On analysis, it turns out to be motivated by two entirely
                              different, ultimately incompatible, independence intuitions. The first
                              motivation is at the level of the theory, and is reminiscent of a
                              reductionist desire for a "semantics-free" account. It takes the fsm thesis
                              to be a claim that computation can be described or analyzed in a
                              semantics-free way. If that were true, so the argument goes, that would go
                              some distance towards naturalizing intentionality. (As Haugeland says "... if
                              you take care of the syntax, the semantics will take care of itself" (1981a,
                              23); see also Haugeland ( 1985 ).

                              There is a second motivating intuition, different in character, that holds at
                              the level of the phenomenon. Here the idea is simply the familiar observation
                              that intentional phenomena, such as reasoning, hoping, or dreaming, carry on
                              in relative independence of their subject matters or referents. Reference and
                              truth, it is recognized, are just not the sorts of properties that can play a
                              causal role in engendering behavior--essentially because they involve some
                              sort of relational coordination with things that are too far away (in some
                              relevant respect) to make a difference. This relational characteristic of
                              intentionality--something I call semantic disconnection--is such a deep
                              aspect of intentional phenomena that it is hard to imagine its being false.
                              Without it, falsity would cease to exist, but so too would hypotheticals;
                              fantasy lives would be metaphysically banned; you would not be able to think
                              about continental drift without bringing the tectonic plates along with you."

                              Now, FSM has become antisemantical, whereas it was a completely intentional
                              theory in the preceding quote. Another inconsistency. (Note also the poor
                              explanation about thoughts of tectonic plates)

                              FSM does not mean that we manipulate symbols independent of their semantics.
                              The meaning of a symbol is exactly what the program specifies it to be. It
                              can, for instance, depend on causal interactions with the larger world, or it
                              can be a temporary structure, completely fictional, such as a new program
                              that the machine has generated and stored in somewhere, to run it, and then
                              delete.

                              In my opinion, it is wrong to call the mechanism of the program just "syntax"
                              and define "semantics" to be something else. What we should be interested in
                              is the physical operation of the machine, and the causal history it will
                              generate according to its program.

                              Let's not forget Smith's ultimate refutation of FSM, of course:

                              "The notion of a transducer must be split in two. In order to retain an
                              antisemantical (fsm) construal of computing, someone interested in
                              transducers would have to distinguish:

                              1. physical transducers , for operations or modules that cross or mediate
                              between the inside and outside of a system; and

                              2. semantic transducers , for operations or modules that mediate or "cross"
                              between symbols and their referents.

                              And it is this bifurcation, finally, that irrevocably defeats the
                              antisemantical formalists' claim. For the only remotely plausible notion of
                              transducer, in practice, is the physical one. That is what we think of when
                              we imagine vision, touch, smell, articulation, wheels, muscles, and the like:
                              systems that mediate between the internals of a system and the "outside"
                              world. Transducers, that is, at least in informal imagination of
                              practitioners, are for connecting systems to their (physical) environments.
                              28 What poses a challenge to the formal (antisemantical) symbol manipulation
                              construal of computation, on the other hand, are semantic transducers: those
                              aspects of a system that involve trading between occurrent states of affairs,
                              on the one hand, and representations of them, on the other. Antisemantics is
                              challenged as much by disquotation as by driving around. "

                              The distinction, is not only vacuous, I think it is ignorant. According to the
                              usual FSM definition, there is no distinction for the computer to start a
                              motor (by producing the appropriate signal=symbol), or to add two numbers (by
                              producing the correct sequence of symbols).

                              In fact, what Smith calls "semantic transduction" is the *very* basis of
                              computing. Without counting, we do not get effective computability. The
                              physical transducers are just I/O conventions from the POV of the computer.
                              The computation is already physical. (FSM does *not* say it is not physical!)

                              Of course, Smith might not appreciate the distinction between the POV of the
                              computer, and the POV of a philosopher who does not know much about computing
                              except elementary programming and word processing. How else could the
                              philosopher exclaim "There can be no theory of computation!"?

                              I think if we looked carefully we would find all seven of his construals to be
                              poor interpretations of computation and in utter error. That is how he
                              reaches his grand (?) conclusions. He first distorts the theoretical
                              discourse, and then makes it contradict in some rather nonsensical ways, and
                              then assert conclusions. His method of manipulating facts reminds me of some
                              famous philosophers in 20th century, but I was not impressed.

                              Regards,

                              --
                              Eray Ozkural (exa) <erayo@...>
                              Comp. Sci. Dept., Bilkent University, Ankara KDE Project: http://www.kde.org
                              http://www.cs.bilkent.edu.tr/~erayo Malfunction: http://malfunct.iuma.com
                              GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C
                            • jrstern
                              ... I think what I said it non-controversial, that the invention of the Turing machine itself has turned out to be much more important than any particular
                              Message 14 of 26 , Aug 2 9:10 AM
                                --- In ai-philosophy@yahoogroups.com, "thoughtfuldavid" <david@b...>
                                wrote:
                                > Josh, you wrote that,
                                >
                                > > The significance of Turing's 1936 paper was not after all in what
                                > > it told us about computable numbers, but in the structural
                                > > framework Turing invented to make his argument.
                                >
                                > My understanding is that the intellectual significance of Turing's
                                > work lies precisely in what it says about computable numbers.

                                I think what I said it non-controversial, that the invention of
                                the Turing machine itself has turned out to be much more important
                                than any particular calculation done with it.

                                > > I've come to hold YET ANOTHER tendentious position, which is
                                > > that while there may be some differences between computation and
                                > > cognition, there is also some significant overlap (picture a
                                > > Venn diagram of two circles overlapping, I suggest the shared
                                > > area is quite large). The question of the day is about what
                                > > computation is, short of cognition. We have to answer that
                                > > before we can draw the diagram accurately.
                                >
                                > Smith and I drew the Venn diagram with the cogitive systems circle
                                > (set) completely enclosed by the computational systems circle,
                                > rather than simply intersecting with it. That is to say that all
                                > cognitive systems are computational. This is the computationalist
                                > claim (I think.)

                                Yyyyyes, .... thank you for this. I've been drawing that simple
                                little diagram, and you remind me to be clearer about it.

                                Yes I believe that all cognition is computational, there are no
                                parts of it that are not, no other science or theology is required.
                                In fact, perhaps the circles are coincident, the two domains
                                identical, but I am trying to find an excuse to keep them distinct!
                                So let me ask, what do you think there might be, which is
                                computational but *not* cognitive? What I was thinking of in the
                                way of things which might be cognitive but "not computational"
                                were aspects of externalism, context, embeddedness.

                                > > Question: does computation, short of cognition, involve
                                > > reference?
                                >
                                > Yes (and no!) Computational devices involve only indirect
                                > reference obtained via a cognitive agent. Without the agent it
                                > would not involve reference (or even be a computer!) I don't
                                > want to muddy the waters, but I think this is what Searle sees
                                > as observer-relative. A rock is a chair only if someone so
                                > uses it. Similarly, the orbits of the planets are a computer
                                > only if someone so uses them (which, of course, does not
                                > detract from their other functions, such as providing us
                                > somewhere to live!)

                                Searle is the author (I think) of the issue of original vs derived
                                intentionality. I no longer think this is a coherent argument.
                                If you read his new essay, I'm not sure Searle really thinks this
                                is still a coherent argument. I'm not entirely certain Searle
                                *ever* thought it was a coherent argument, but it was tricky in
                                its fallacies and he threw it at the AI crowd, and it flustered
                                them (us) in an amusing way.

                                Sure, even a chair is a chair only if someone says it is,
                                even if nobody ever sits in it, but the thing is, the chair
                                has the property of being stable under a 200kg load, even if
                                nobody ever says it is a chair. And if it isn't stable under
                                load, and somebody sits in it (or doesn't), then is it really
                                a chair after all?

                                The orbits of a planet are never a computer, and the paint on
                                Searle's wall is never running Wordstar, btw.

                                The marks on paper, however, do not comprise a word until
                                they are properly interpreted. So there are some things in
                                the universe that are involved in relations that contribute
                                to cognition, but not everything can participate in all
                                relations in a meaningful or significant manner. One would
                                think this statement obvious, but Searle's parables work by
                                ignoring it.

                                This issue of whether mundane computation involves reference,
                                is my improved handle on the issue that Smith and Sloman are
                                writing about here. Think about it a little further, if you will.

                                That is, I must disgree with your phrasing of the matter,
                                "Computational devices involve only indirect reference
                                obtained via a cognitive agent." It would seem such an
                                assumption would make AI impossible ab initio by granting
                                some special privilege to something called a "cognitive agent".
                                Well, but that is the very question, whether a computational
                                system can *be* a cognitive agent. Neither Searle nor Dennett
                                nor you nor me is privileged to make such determinations by fiat,
                                just as neither Searle's ass nor mine is privileged to determine
                                just what constitutes a chair. The issue is just much more
                                involved than that.

                                Josh
                              • jrstern
                                ... I think what you see is a computer science guy rather light on philosophical background and style, and not keeping the logic straight, either, as you point
                                Message 15 of 26 , Aug 2 9:51 AM
                                  --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...> wrote:
                                  > ... Smith is being inconsistent. First, he declares "semantics"
                                  > subject as recalcitrant. He also says that it is not possible to
                                  > claim that semantics arises from computation because it will be
                                  > either circular or assuming some biases like mind-body dualism.
                                  > And *then* criticizes FSM view because it doesn't talk about
                                  > semantics.
                                  >
                                  > I think that is very muddy thinking. No wonder why it took him
                                  > 30 years to come up with a, well, completely misled account of
                                  > computation. He seems to be more like a net.kook than a
                                  > philosopher by my standards of brevity, clarity, consistency
                                  > and rationality.

                                  I think what you see is a computer science guy rather light on
                                  philosophical background and style, and not keeping the logic
                                  straight, either, as you point out below. In his defense he
                                  was working on other topics, most of that 30 years!

                                  > Let's see more of what I think is a wrong view of FSM concept.
                                  >
                                  > Smith first says: "It (FSM) explicitly characterizes computing in
                                  > terms of a semantic or intentional aspect, if for no other reason
                                  > than that without some such intentional character there would be
                                  > no warrant in calling it symbol manipulation", and goes on
                                  > to defend his confused thinking in the footnote.
                                  >
                                  > First, "symbol" does *not* mean intentionality.
                                  ...

                                  Eray, I agree with pretty much all of your analysis. Smith makes a
                                  complete hash of things as he goes, bad assumptions, contradictory
                                  explanations, etc.

                                  You say many interesting things here that I'm going to pass over for
                                  the moment, maybe try to come back to later today.

                                  > Now, FSM has become antisemantical, whereas it was a completely
                                  > intentional theory in the preceding quote. Another inconsistency.

                                  I like the antisemantical instinct here. I respect Smith for
                                  getting his hands on both horns of the dilemma - how *do* we
                                  account on the one hand for intentionality, and on the other
                                  for the antisemantical notions of methodological solipsism?

                                  But he eventually follows in the steps of Winograd, of having
                                  a computer scientist completely throw up his hands and surrender
                                  the field to the first philosopher to challenge his metaphysics.
                                  He wants to equate computation to any and all complex problems,
                                  rather than have it be a separate study - though he also has
                                  sentences that allow that there *might* still be a theory of
                                  computation. Completely inconsistent. I was going to say the
                                  book editor should have cleaned this up -- but then, I haven't
                                  even seen the book yet, just the draft, maybe it *was* cleaned
                                  up in the book (I should have a copy in a few days).

                                  > "The notion of a transducer must be split in two. In order
                                  > to retain an antisemantical (fsm) construal of computing,
                                  > someone interested in transducers would have to distinguish:
                                  >
                                  > 1. physical transducers , for operations or modules that
                                  > cross or mediate between the inside and outside of a system; and
                                  >
                                  > 2. semantic transducers , for operations or modules that mediate
                                  > or "cross" between symbols and their referents.
                                  >
                                  > And it is this bifurcation, finally, that irrevocably defeats
                                  > the antisemantical formalists' claim.
                                  ...

                                  He has descended into gibberish here, and I salute your strength
                                  in trying to untangle it, I'd rather pass. But then, I've never
                                  had any enthusiasm for trying to deconstruct issues of sensory data
                                  versus concepts versus externalism. I have my own formulation, which
                                  I would simply put forward as an alternative, and let a generation
                                  of grad students do the compare and contrast.

                                  > Of course, Smith might not appreciate the distinction between
                                  > the POV of the computer, and the POV of a philosopher who does
                                  > not know much about computing except elementary programming
                                  > and word processing. How else could the
                                  > philosopher exclaim "There can be no theory of computation!"?

                                  Smith has been a heavy-duty programmer in his time, has he not?

                                  > I think if we looked carefully we would find all seven of his
                                  > construals to be poor interpretations of computation and
                                  > in utter error.

                                  I think a little later I will list the seven and respond to them
                                  in turn.

                                  Josh
                                • Eray Ozkural
                                  ... I think I got his background wrong. Sorry for the stupid assumptions above. -- Eray Ozkural (exa) Comp. Sci. Dept., Bilkent
                                  Message 16 of 26 , Aug 2 10:01 AM
                                    On Monday 02 August 2004 19:51, jrstern wrote:
                                    > > Of course, Smith might not appreciate the distinction between
                                    > > the POV of the computer, and the POV of a philosopher who does
                                    > > not know much about computing except elementary programming
                                    > > and word processing. How else could the
                                    > > philosopher exclaim "There can be no theory of computation!"?
                                    >
                                    > Smith has been a heavy-duty programmer in his time, has he not?

                                    I think I got his background wrong. Sorry for the stupid assumptions above.

                                    --
                                    Eray Ozkural (exa) <erayo@...>
                                    Comp. Sci. Dept., Bilkent University, Ankara KDE Project: http://www.kde.org
                                    http://www.cs.bilkent.edu.tr/~erayo Malfunction: http://malfunct.iuma.com
                                    GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C
                                  • Eray Ozkural
                                    ... Okay, this is not unlike the traditional view, but then, I think you don t consider such a thing as replication of DNA as computation. I do, because I
                                    Message 17 of 26 , Aug 2 10:27 AM
                                      On Sunday 01 August 2004 20:16, jrstern wrote:
                                      > --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...> wrote:
                                      > > I, too, think that a revised view of computation is necessary.
                                      > > However, I think along the lines of Marvin Minsky's www.edge.org
                                      > > interview: computation is not only about electronic computers, it
                                      > > is the science and practice of complex systems.
                                      >
                                      > I disagree with this entirely.
                                      >
                                      > I say that computation is about how we can build nice, little
                                      > machines that can be filled with bits, cranked, and somehow
                                      > produce new bit patterns that we can interpret interestingly.

                                      Okay, this is not unlike the traditional view, but then, I think you don't
                                      consider such a thing as replication of DNA as computation. I do, because I
                                      think it would be inconsistent to say that
                                      1) We can build a computer out of DNAs, proteins, cellular machinery (proven
                                      as you know)
                                      2) Protein transcription, DNA replication, etc. can be simulated, and indeed
                                      their computational properties help us to develop genetic engineering.
                                      3) Replication of the DNA is not computational

                                      That is, why I find it hard to say that computation is by definition those
                                      things that we build. Then, it would automatically follow that our brains are
                                      not computers, and anything we build will be categorically different than
                                      brains. There would be no need for ai-philosophy, either :) [IMO, that is so,
                                      but I reckon this view can be contested]

                                      > Just what processes are involved in generating sunspots is simply
                                      > not in the same field of study as wondering how it is that we,
                                      > as intelligent agents, can *talk* about sunspots.

                                      I disagree. :) To be able to talk about sunspots scientifically, you have to
                                      understand the processes underlying them (ie. you need a scientific
                                      *theory*). Otherwise, you could ascertain that sunspots have a metaphysical
                                      (usually means magical) linkage to the ratio of skin cancer on earth, and
                                      that the sunspots are projections of demons on the surface of sun. That, too,
                                      is talk.

                                      In my opinion, the necessary scientific theory can be analyzed in
                                      computational terms (as in Algorithmic Information Theory), and therefore,
                                      our talk about sunspots is in the same field of study (it's called science,
                                      not philosophy!) as "what processes are involved in generating sunspots".

                                      Talk about computation, however, may be somewhat different, because some of
                                      the talk about computation is evidently philosophical, rather than merely
                                      scientific as shown in your historical remarks below. (as well as in the last
                                      three paragraphs of mine)

                                      > This disconnect
                                      > between methodologically solipsistic computation and distal,
                                      > ontological truths is crucial to having any conversation at all
                                      > about computation Wittgenstein said so, and it's the rightest thing
                                      > he ever gave us! (That was the "linguistic turn" (LT). As it worked
                                      > out, the *reasons* he gave for LT were basically positivistic and
                                      > skeptical, and, well, for the most part, wrong. When Turing came
                                      > along and talked to Wittgenstein circa 1937, Wittgenstein had every
                                      > opportunity to see how computation offers an entirely different
                                      > rationale for LT, but Wittgenstein rejected it. If only old Ludwig
                                      > could spend a little time with us here in 2004, surrounded by
                                      > computers, I think he would have to change his tune!)

                                      Yes, but I think Wittgenstein's true reasons might be different (such as
                                      theological), we cannot know at the moment.

                                      I am not certain what Turing said about Linguistic Turn. Do you have any
                                      references for that? (Preferably online)

                                      > > Thus, it applies equally to natural systems as well as
                                      > > man-made. For an example of the elegance of computational approach
                                      > > for studying biological systems, a nice example is the recent
                                      > > theoretical trends in Amorphous Computing and Cellular Automata.
                                      > > Needless to say, computational sciences have shown remarkable
                                      > > utility in fields quite distinct from computing itself, and is
                                      > > continuing to become a major application.
                                      >
                                      > There are probably many links between computation and biology, as
                                      > all living systems are encoded in easily observed strings of nucleic
                                      > acids and such. Please note this is a true physical encoding, and
                                      > not just a stance we take! There is much we can learn about
                                      > computation by looking at biological systems.

                                      Yes, it is a true physical encoding. Which brings us to the question: do not
                                      the bits on your hard-drive have a true physical encoding? If they are not
                                      physical, how can the disk-head, a purely physical entity, read this
                                      encoding?

                                      Another question: how easier is it to observe "strings of nucleic acids", than
                                      the "arrangements of magnetic properties on the surface of a disk"? Why did
                                      it take decades to "crack" the human genome, then?

                                      I think you might need a stronger argument for the necessity of methodological
                                      solipsism.

                                      >
                                      > Yet, I take a tendentious methodological
                                      > stance on all of this. I say that if we choose to limit ourselves
                                      > just to GOFAI, basically symbolic computation as done on electronic
                                      > digital computers, we lose nothing in generality in any conclusions
                                      > we reach. Further, I limit the discussion to matters of "cognition",
                                      > ignoring qualia, consciousness, free will, etc. I have speculations
                                      > concerning these other matters, but do not wish to engage them.
                                      > Please note that anything I post here, will be from this perspective.
                                      > My only project is analyzing, correcting, and rehabilitating a form
                                      > of strong AI that is, to all intents and purposes, GOFAI,
                                      > Release 2.0

                                      I see. But the term GOFAI may be misleading. It is said that GOFAI is a term
                                      used only by people who are not familiar with AI. Do any AI proponents
                                      actually use that (somewhat satirical) term?

                                      > > With respect to Smith and Sloman, I think they underrate the
                                      > > importance of Turing's (and others') theoretical contributions. The
                                      > > Turing Machine is not just a machine, it is a model of computation.
                                      > > Smith says there can be no theory of computation, but in reality we
                                      > > have quite powerful theories of computation, and indeed, we have
                                      > > theories of fundamental concepts *based* on computation such as
                                      > > Algorithmic Information Theory and Algorithmic Probability Theory.
                                      >
                                      > I am in great sympathy with everything that Smith and Sloman are
                                      > attempting, but they make a mistake in dismissing Turing computation,
                                      > which needs to be better understood. I suggest a historical study of
                                      > Hilbert, Brouwer, Frege, Wittgenstein, Tarski, and a few others,
                                      > shines new light on what Turing did. Seeing it requires the kind
                                      > of 20-20 hindsight we can have today, with our empirical knowledge
                                      > of computers-in-the-wild, as Smith says.

                                      I agree. But I think we must also take into account the work of later
                                      mathematicians.

                                      > > There are so many claims in Smith's article that it would take four
                                      > > times the length of his article to refute them. His philosophy is,
                                      > > in my opinon, of the sort "Look, we don't exist" kind of
                                      > > unnecessary skepticism....
                                      >
                                      > Yes, and that is exactly why eliminativism about computation is
                                      > wrong, and even Minsky's claim that computation = science is wrong.
                                      > Computation is a separable domain of study. In fact, I believe that
                                      > it is a domain with roots over two thousand years old - computation
                                      > is, in my reading, a superset of the domain of the study
                                      > of language. That is, I am very sympathetic, again methodologically
                                      > (at least), with the "language of thought" (LOT) hypothesis.
                                      > My gosh, if digital electronic computers do not instantiate LOT,
                                      > then what the heck do they do? (I know that is not quite as easy
                                      > a question as it looks like, but it's a starting point).

                                      If you remember my (somewhat strange) arguments for wide linguistic idealism,
                                      I had argued that computation is indeed the generalized form of linguistic
                                      analysis. And that computably enumerable computation is the most fitting
                                      candidate for LoT.

                                      > In fact, Minksy posted a message a few months back regarding his
                                      > meetings as a grad student with Quine, and how Minsky, back then,
                                      > was trying (unsuccessfully) to convince Quine that computation was
                                      > a process not implicit, inherent, or immanent in logical or physical
                                      > systems. That's my probably overbroad interpretation of what
                                      > Minsky posted, but I still see it as in conflict with what he said
                                      > in that Edge piece, if you cite it correctly.

                                      I think I cited it correctly, but of course it may be fun to watch it for
                                      yourself. I am not sure if Minsky told Quine that, though. If it is not
                                      logical or physical, then what is it? :)

                                      > > For instance, let's take the "Seven Construals of Computation" in
                                      > > Smith's article. It seems that Smith tries to distinguish
                                      > > conceptions of computation, attempting to define them, and then
                                      > > criticize them based on his own definitions, which I believe
                                      > > do not reflect the true spirit of "old" computationalism.
                                      >
                                      > I am very sympathetic to these lists of important features. I have
                                      > a couple of my own. The best one has about fifteen entries. Mine
                                      > are much more basic. My #1 is that computation is a physical
                                      > process. I know that right there I go into conflict with the
                                      > majority current view, but it can be made to work (in fact, it is
                                      > the ONLY thing that can be made to work, as I hope to eventually
                                      > show!), but I don't have to invent anything major to say so, I just
                                      > have to subscribe to a particularist form of nominalism to make
                                      > this case, and to reject any hint of platonism. In fact, noting
                                      > and remembering that computation is physical is the best tool I have
                                      > found for detecting and eschewing platonism.

                                      I am opposed to the widespread Platonist interpretations of computationalism
                                      myself. I think accepting a Platonist stance or not makes a big difference
                                      philosophically. I'm trying to construct an opposing view at least for the
                                      sake of an alternative. Digital physicists are making these cryptic Platonist
                                      remarks. How do they know it is true? :)

                                      In addition, I think that computation is purely physical. I am also
                                      sympathetic with nominalism, but our views with you seem to differ
                                      considerably. (Maybe because I am not truly a nominalist?)

                                      However, I'm not a solipsist by any measure. I think there is one objective
                                      reality, and our minds (and kidneys) are a very physical part of that
                                      reality. What we call subjective, in my opinion, is simply variety and
                                      complexity of existence (And therefore of thought). I think I'd mentioned
                                      that I'm a monist.

                                      > > Let's have a look at the first such "construal":
                                      > > 1. formal symbol manipulation (fsm): the idea, derivative from a
                                      > > century's work in formal logic and metamathematics, of a machine
                                      > > manipulating symbolic or (at least potentially) meaningful
                                      > > expressions without regard to their interpretation or semantic
                                      > > content;
                                      > >
                                      > > Is this really the case?
                                      >
                                      > No!
                                      >
                                      > I endorse the "symbol" part, I would qualify the "manipulation" part,
                                      > and question or reject the "formal" part. Smith here is referring
                                      > to what Fodor (and I) like to refer to as methodological solipsism,
                                      > but I for one prefer to view this in causal terms, not formal ones.
                                      > Fodor with his syntactic essentialism does subscribe to something
                                      > like the formal view and it ultimately fails him, or at least he
                                      > thinks it does!

                                      By formal, do we not simply mean that the manipulations are governed by clear
                                      mechanical rules?

                                      > In any case, as above, work on logic is not prior to work on
                                      > language.

                                      I could agree with that. I find that the logicist views of language and
                                      mathematics can be quite misleading.

                                      > > And I am very suspicious of Smith's expressions like the following:
                                      > >
                                      > > Turing machines, notions of "effective computability", and the
                                      > > like--fails as a theory of computing, in spite of its name and
                                      > > its popularity. It is simultaneously too broad, in applying to more
                                      > > things than computers, and too narrow, in that it fails to apply
                                      > > to some things that are computers. More seriously, what it is
                                      > > a theory of, is not computing.
                                      > >
                                      > > Here Smith turns this serious subject into a talk show. How is
                                      > > it too broad? Was not one of the original things we wanted
                                      > > to show was how general computation is in real life?
                                      >
                                      > Yes, exactly!!!

                                      Yes, it was, and I think the traditional view does this excellently from a
                                      postmodern perspective. It shows us that many things in nature are in fact
                                      computational (A point that you would not endorse :), that they can be
                                      understood, analyzed and engineered in computational terms. (like say,
                                      weather forecast)

                                      > > How is it too narrow? The theories Smith refers to are all
                                      > > theories of discrete computation, and there is no aspect of
                                      > > discrete computation that they fail to address.
                                      >
                                      > Oh, I suppose I could give a sympathetic reading to what Smith
                                      > is saying, but it still comes out wrong in the end.
                                      >
                                      > I'm sure that Smith and Sloman and I are all groping the same
                                      > elephant, (you all know the parable of the blind men and the
                                      > elephant, right?) but I think I have the better view!
                                      > Not that I've published it yet.

                                      I think you have some interesting arguments, and the capacity for a thorough
                                      exposition of them. I wish you luck in confronting the philosophical
                                      journals. (I don't have the courage myself)

                                      > > Sloman also seems out of focus in this issue. He seems to think
                                      > > that things like interrupts are a very crucial part of computation
                                      > > that the theory does not address. I think it is a bit premature
                                      > > to say that interrupt handling facility (both system and user
                                      > > program side) is not just a small program.
                                      >
                                      > Again, what I see in this point from Sloman is a desire to show the
                                      > embedding of computational systems in the world, to show that they
                                      > are NOT purely formal and NOT entirely solipsistic. But he doesn't
                                      > unpack the ideas far enough to make his case.

                                      I see. But I fail to understand how a computer could be made "informal" by
                                      embedding computational systems in the world. There is some conceptual
                                      fuzziness when talking about "embedding" computers in the world. Is there a
                                      computer that is not embedded in the real world? That seems a little absurd.
                                      If it does not come out of heaven instantly, its program will have a total
                                      causal connection to the rest of the world (in that every bit of the program
                                      string will have to come from somewhere), except in singularity (like Big
                                      Bang) where the concept of causality may be irrelevant.

                                      The degree of further interaction may vary, just like for any other machine,
                                      but how does that make a computer different from any other machine in terms
                                      of interaction? (Maybe because we presuppose minds are different than bodies,
                                      and computers are minds?? :)

                                      On the other hand, I think it's wrong to view computation as a purely
                                      abstract, idealized mathematical concept that has nothing to do with physical
                                      reality, and I agree with Sloman in his desire. Computation is best
                                      understood as a physical activity first. However, I think the current theory
                                      already shows that natural complexity is best understood in terms of
                                      computation. In fact, that is how personal computers are best analyzed: with
                                      the theory of computation, without which we would be unable to construct
                                      things like advanced programming languages, for instance. If we left
                                      computing to a bunch of bit cowboys, we would hardly get things like
                                      databases or adaptive systems right :)

                                      Maybe, digital philosophy does give a good answer to Sloman's concerns, in
                                      that the very substance of physical existence can be computational (bridging
                                      the gap between the formal and informal, solipsistic and not-so-solipsistic),
                                      however, the philosophical arguments are rather weak so far. (That's what I'm
                                      working on, mainly. If you look at the archives of sci.physics.discrete,
                                      there have been quite interesting discussions)

                                      Regards,

                                      --
                                      Eray Ozkural (exa) <erayo@...>
                                      Comp. Sci. Dept., Bilkent University, Ankara KDE Project: http://www.kde.org
                                      http://www.cs.bilkent.edu.tr/~erayo Malfunction: http://malfunct.iuma.com
                                      GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C
                                    • thoughtfuldavid
                                      Eray... You may not think very highly of Smith s arguments, but given his background (see http://www.counterbalance.net/bio/csmith- body.html for a short
                                      Message 18 of 26 , Aug 2 11:24 AM
                                        Eray... You may not think very highly of Smith's arguments, but
                                        given his background (see http://www.counterbalance.net/bio/csmith-
                                        body.html for a short biography) you should certainly assume that he
                                        has a reasonable idea of what he is discussing!

                                        You begin your critique by saying,

                                        > First, "symbol" does *not* mean intentionality. A symbol is just
                                        that, a sign.

                                        I confess to thinking exactly the same way when I was first exposed
                                        to this subject. BUT unfortunately this is simply not (exactly) what
                                        is understood by the term symbol in much of the philosophical
                                        literature. I am not sure I can precisely say what it is, but I
                                        believe I am correct in saying that it is taken to have a semantic
                                        element to it and that Smith does use the term appropriately,
                                        accurately and consistently in his argument. This then renders most
                                        of your analysis vacuous. Sorry.


                                        I assume that the FSM construal originates (or is at least related)
                                        with Newell & Simon's hypothesis that a physical symbol system is a
                                        necessary and sufficient condition for intelligent action. A symbol
                                        system comprises symbols and rules for manipulating them. The
                                        symbols are open to a systematic semantic interpretation (i.e., a
                                        symbol always stands for the same thing), whilst the rules are
                                        defined relative to the "form or shape" of the symbols they
                                        manipulate, rather than relative to the things they might stand for.

                                        > [Smith] "The fsm construal has a distinctly antisemantical flavor,
                                        owing to its claim
                                        > that computation is the "manipulation of symbols independent of
                                        their semantics." "

                                        > [Haugeland] "... if you take care of the syntax, the semantics
                                        will take care of itself"

                                        > [Eray] "FSM does not mean that we manipulate symbols independent
                                        of their semantics.
                                        > The meaning of a symbol is exactly what the program specifies it
                                        to be."

                                        The key is to arrange for the rules (equivalently the physical
                                        mechanism that implements them) to somehow respect (follow) the
                                        semantics of the symbols.

                                        > [Eray] It may refer to other things, but those references are
                                        resolved in *minds*,
                                        > there is no such thing as the reference, or intentionality of a
                                        symbol without a
                                        > mind to resolve it. ... The symbols in the computation could mean
                                        something to an
                                        > observer, or not. That is not the problem. ... The problem is
                                        whether the symbols
                                        > can ever have inherent intentionality from the point of view of
                                        *computation*
                                        > itself. (And I think the answer to that question is an echoing YES)

                                        YES, if we (cognitive agents) are computational, otherwise it is not
                                        entirely obvious -at least to me.
                                      • jrstern
                                        ... You can make much of a sign versus a symbol, but I propose that the philosophical, philological, and semiotic literature of the last 2,500 years does not
                                        Message 19 of 26 , Aug 2 12:08 PM
                                          --- In ai-philosophy@yahoogroups.com, "thoughtfuldavid" <david@b...>
                                          wrote:
                                          > > First, "symbol" does *not* mean intentionality. A symbol is just
                                          > > that, a sign.
                                          >
                                          > I confess to thinking exactly the same way when I was first exposed
                                          > to this subject. BUT unfortunately this is simply not (exactly)
                                          > what is understood by the term symbol in much of the philosophical
                                          > literature.

                                          You can make much of a sign versus a symbol, but I propose that the
                                          philosophical, philological, and semiotic literature of the last
                                          2,500 years does not resolve the issue.

                                          If there is no such thing as intentionality in the first place, you
                                          cannot define symbol in intentional terms. At best, Smith (and any
                                          others) simply interdefines symbol and intentionality here. In the
                                          last twenty years or so, nobody has really honored the
                                          symbol/intentionality tradition, seeing it trumped by externalist
                                          arguments. Those who do honor it today seem to be neo-Fregeans like
                                          Kripke, who justify it not on intentional grounds but on
                                          essentialist, normative, modal, and platonistic grounds -- all of
                                          which basically suck. If you believe *any* of these things, not only
                                          do you lose the basis for computationalism, you cannot define even
                                          mundane computation, except in the hand-waving, derived-
                                          intentionality way that Searle was deriding for the last twenty years.

                                          Please excuse me if I dismiss these without due consideration. I
                                          realize you still seem happy to recite these as solutions, or at
                                          least conventional dogma. We could take them one at a time, but each
                                          is an argument in itself. The point is, that Smith and Sloman are
                                          seen as rejecting them (Smith only cites these en passant in order to
                                          dismiss them anyway!), and I think both Eray and I are pretty well
                                          conversant with these established theories, and we're not impressed,
                                          either.

                                          > I am not sure I can precisely say what it is, but I
                                          > believe I am correct in saying that it is taken to have a semantic
                                          > element to it and that Smith does use the term appropriately,
                                          > accurately and consistently in his argument.

                                          No he does not, he uses it loosely and inconsistently at best, even
                                          if you were simply grading him on a test of the ideas.

                                          It is almost (if not quite) entirely unknown to see a guy with a
                                          computer science background use classic philosophy and linguistic
                                          ideas correctly. This refers to me, too, btw. The problem is,
                                          without speaking for Smith, I do not *want* to use them "correctly",
                                          as reconciling the two traditions is an immense struggle, and in
                                          order to put out whatever new ideas we want to propose, we take many
                                          shortcuts in referring to the opposition.

                                          So I don't really knock Smith for getting the recitation wrong, I
                                          agree with him that the old explanations don't work anyway, I
                                          consider his negative point sufficiently established, and try to
                                          focus on whatever new ideas he proposes instead.

                                          Josh
                                        • thoughtfuldavid
                                          Earlier Josh wrote, ... You ask What is computational but not cognitive? Am I missing something? How about your PC or even a pocket calculator? I presume you
                                          Message 20 of 26 , Aug 2 3:12 PM
                                            Earlier Josh wrote,

                                            > Yes I believe that all cognition is computational, there are no
                                            > parts of it that are not, no other science or theology is required.
                                            > In fact, perhaps the circles are coincident, the two domains
                                            > identical, but I am trying to find an excuse to keep them distinct!
                                            > So let me ask, what do you think there might be, which is
                                            > computational but *not* cognitive? What I was thinking of in the
                                            > way of things which might be cognitive but "not computational"
                                            > were aspects of externalism, context, embeddedness.

                                            You ask "What is computational but not cognitive?" Am I missing
                                            something? How about your PC or even a pocket calculator? I presume
                                            you would not want to ascribe cognitive abilities to them. In the
                                            last sentence I assume you mean cognitive -AND computational- since
                                            we agreed (in the first sentence) that "all cognition is
                                            computational". As to what additional characteristics systems must
                                            possess to be classed as cognitive, I would say simply that they
                                            must have direct (original, as opposed to derived) intentionality -
                                            that is, the symbols employed by the system must have meaning for it
                                            without the need for any other cognitive system.

                                            I should also say that I do not view the boundary of the cognitive
                                            systems circle (set) as being a sharp one. Rather, I believe
                                            cognition is a matter of degree, from the lowly pocket caculator -no
                                            cognition, through locusts -some cognition, to human beings -
                                            generally cognitive!

                                            On the other points you raise,

                                            > Sure, even a chair is a chair only if someone says it is,
                                            > even if nobody ever sits in it, but the thing is, the chair
                                            > has the property of being stable under a 200kg load, even if
                                            > nobody ever says it is a chair. And if it isn't stable under
                                            > load, and somebody sits in it (or doesn't), then is it really
                                            > a chair after all?

                                            You are correct, of course, the chair is a chair if someone
                                            considers it so, irrespective of whether anyone actually sits on it.
                                            If someone does sit on it and it collases it may still be viewed as
                                            a chair, just a broken one!

                                            > The orbits of a planet are never a computer, and the paint on
                                            > Searle's wall is never running Wordstar, btw.

                                            Sorry, but you will have to explain this one to me. The planet would
                                            not be a general purpose computer, but it could most surely function
                                            as a special purpose one. Searle's wall running Wordstar is another
                                            matter altogether. I would agree that it is not a computer of any
                                            kind.


                                            > This issue of whether mundane computation involves reference,
                                            > is my improved handle on the issue that Smith and Sloman are
                                            > writing about here. Think about it a little further, if you will.
                                            >
                                            > That is, I must disgree with your phrasing of the matter,
                                            > "Computational devices involve only indirect reference
                                            > obtained via a cognitive agent." It would seem such an
                                            > assumption would make AI impossible ab initio by granting
                                            > some special privilege to something called a "cognitive agent".
                                            > Well, but that is the very question, whether a computational
                                            > system can *be* a cognitive agent. Neither Searle nor Dennett
                                            > nor you nor me is privileged to make such determinations by fiat,
                                            > just as neither Searle's ass nor mine is privileged to determine
                                            > just what constitutes a chair. The issue is just much more
                                            > involved than that.

                                            I don't understand why you say that this position would make AI
                                            impossible. We have agreed (I think) that "ALL cognitive systems are
                                            computational" and that "There (can) exist computational systems
                                            that are NOT cognitive." This allows for us to be cognitive and our
                                            PC's to be a purely (mundane) computational systems. It also allows
                                            that by adding the "missing" ingredient to our PC we might make it
                                            cognitive. This is at least part of what AI is attempting to do,
                                            isn't it?

                                            The chair may be a bad example here, because it seems to me that I
                                            can in fact personally decide what constitutes a chair -for example
                                            that rock over there becomes a chair the moment I choose to sit down
                                            on it to rest. The case of cognition and computation may not be
                                            quite so straightforward but in essence I see no real reason that we
                                            cannot decide what we consider to be cognitive or whatever. These
                                            notions are not yet written in stone, so provided that we do justice
                                            to the common conceptions and are coherent/consistent then others
                                            may come to accept our views too.
                                          • jrstern
                                            ... Maybe, but I want arguments, not assertions. Remember, McCarthy (in)famously claims that a bimetallic strip is intelligent , just not *very* intelligent.
                                            Message 21 of 26 , Aug 2 4:27 PM
                                              --- In ai-philosophy@yahoogroups.com, "thoughtfuldavid" <david@b...>
                                              wrote:
                                              >
                                              > Earlier Josh wrote,
                                              >
                                              > > Yes I believe that all cognition is computational, there are no
                                              > > parts of it that are not, no other science or theology is
                                              > > required. In fact, perhaps the circles are coincident, the two
                                              > > domains identical, but I am trying to find an excuse to keep
                                              > > them distinct! So let me ask, what do you think there might be,
                                              > > which is computational but *not* cognitive? What I was thinking
                                              > > of in the way of things which might be cognitive but "not
                                              > > computational" were aspects of externalism, context,
                                              > > embeddedness.
                                              >
                                              > You ask "What is computational but not cognitive?" Am I missing
                                              > something? How about your PC or even a pocket calculator? I
                                              > presume you would not want to ascribe cognitive abilities to them.

                                              Maybe, but I want arguments, not assertions.

                                              Remember, McCarthy (in)famously claims that a bimetallic strip is
                                              "intelligent", just not *very* intelligent. You seem to be willing
                                              to claim that planets are computational, and while you did not
                                              recite your argument, such arguments are generally convertible by
                                              simple word replacement from arguments that they are following
                                              rules (which is incorrect, but that's the flavor I guess you need
                                              to assert to say that planets are computational), to arguments
                                              that following rules is ipso facto cognition.

                                              Sorry to outrage you, but many of these traditional arguments
                                              are inconsistent and incomplete (in my analysis), besides being
                                              mistaken.

                                              > In the
                                              > last sentence I assume you mean cognitive -AND computational-
                                              > since we agreed (in the first sentence) that "all cognition is
                                              > computational".

                                              No, I mean what I said, I want to reserve a category of something
                                              like computational, which cognitive systems would not involve.
                                              The category may be empty, but I think not. Since I guess you've
                                              given me a response, I should specify mine. What I have in mind
                                              here might include some brute facts about hardware, which may be
                                              necessary for computation but not specifiable in the description
                                              of a cognitive system.

                                              I should have clarified, call the fact
                                              that ultimately cognition is entirely computational,
                                              computational(1). Perhaps I should choose another term for one
                                              or the other, but let us call the (hypothetical) aspects which
                                              are cognitive but not computational, somehow excluded from a
                                              computational(2). My examples of these were externalism and
                                              context. How do you describe context as computational?

                                              (The diagram here is two overlapping circles, one is cognition
                                              the other is computation(2). They are both enclosed in a larger
                                              circle which represents computation(1).)

                                              (btw, thank you for making me spell it out, ... though I realize
                                              I might still have some way to go. I may be right or wrong,
                                              but I want the issue at least to be clear!)

                                              > As to what additional characteristics systems must
                                              > possess to be classed as cognitive, I would say simply that they
                                              > must have direct (original, as opposed to derived)
                                              > intentionality - that is, the symbols employed by the system must
                                              > have meaning for it without the need for any other cognitive
                                              > system.

                                              I am very leary of the entire idea of intentionality, but if I
                                              were going to allow it, I would define it roughly as you say.
                                              But tell me, in a mundane computer program, if I have a variable
                                              "nChairs" or pMyChair, do these have "meaning" "for the program",
                                              or not? I find no easy answer for this question, and my recent
                                              urge is to argue that they *do* have "meaning" "for the program".
                                              One can use up a lot of scare quotes this way!

                                              > I should also say that I do not view the boundary of the cognitive
                                              > systems circle (set) as being a sharp one. Rather, I believe
                                              > cognition is a matter of degree, from the lowly pocket caculator
                                              > -no cognition, through locusts -some cognition, to human beings -
                                              > generally cognitive!

                                              Agreed, but simply as a rhetorical point, I would draw a line
                                              somewhere, and limit my conversation to that range. Planets
                                              fall outside that arbitrary range.

                                              (Actually I lied here, I see it as more than a rhetorical point,
                                              but let's just go with that for now.)

                                              > On the other points you raise,
                                              >
                                              > > Sure, even a chair is a chair only if someone says it is,
                                              > > even if nobody ever sits in it, but the thing is, the chair
                                              > > has the property of being stable under a 200kg load, even if
                                              > > nobody ever says it is a chair. And if it isn't stable under
                                              > > load, and somebody sits in it (or doesn't), then is it really
                                              > > a chair after all?
                                              >
                                              > You are correct, of course, the chair is a chair if someone
                                              > considers it so, irrespective of whether anyone actually sits on
                                              > it. If someone does sit on it and it collases it may still be
                                              > viewed as a chair, just a broken one!

                                              Well, I don't know. Perhaps it was only an imitation chair
                                              all along. Perhaps my attribution was mistaken due to drugs
                                              or bad light or whatever.

                                              We get into issues of type and token, category, natural type,
                                              platonism versus nominalism, normativity, attribution, reference,
                                              representation, ontology, and commitment here. No easy answers.
                                              No short ones, anyway.

                                              > > The orbits of a planet are never a computer, and the paint on
                                              > > Searle's wall is never running Wordstar, btw.
                                              >
                                              > Sorry, but you will have to explain this one to me. The planet
                                              > would not be a general purpose computer, but it could most surely
                                              > function as a special purpose one. Searle's wall running Wordstar
                                              > is another matter altogether. I would agree that it is not a
                                              > computer of any kind.

                                              These are ancient arguments about rule-following and lawful systems,
                                              and again, I think we want to avoid borderline cases for the moment.
                                              I do allow that a good theory may be intertranslatable with one that
                                              attributes computation to some unlikely systems, but considering
                                              them so up front obscures all reason. "The universe is its own
                                              most efficient emulator", but nothing is learned by asserting this
                                              (putative) general rule in broad or in particular.

                                              > > This issue of whether mundane computation involves reference,
                                              > > is my improved handle on the issue that Smith and Sloman are
                                              > > writing about here. Think about it a little further, if you
                                              > > will.
                                              > >
                                              > > That is, I must disgree with your phrasing of the matter,
                                              > > "Computational devices involve only indirect reference
                                              > > obtained via a cognitive agent." It would seem such an
                                              > > assumption would make AI impossible ab initio by granting
                                              > > some special privilege to something called a "cognitive agent".
                                              > > Well, but that is the very question, whether a computational
                                              > > system can *be* a cognitive agent. Neither Searle nor Dennett
                                              > > nor you nor me is privileged to make such determinations by fiat,
                                              > > just as neither Searle's ass nor mine is privileged to determine
                                              > > just what constitutes a chair. The issue is just much more
                                              > > involved than that.

                                              > I don't understand why you say that this position would make AI
                                              > impossible. We have agreed (I think) that "ALL cognitive systems
                                              > are computational" and that "There (can) exist computational
                                              > systems that are NOT cognitive." This allows for us to be
                                              > cognitive and our PC's to be a purely (mundane) computational
                                              > systems. It also allows that by adding the "missing" ingredient
                                              > to our PC we might make it cognitive. This is at least part
                                              > of what AI is attempting to do, isn't it?

                                              Maybe, but this is asserting the conclusions. Again, the point of
                                              Smith and Sloman (and in part, me) is that we need to understand
                                              better what the basic computation comprises before I can say what
                                              is and what isn't cognitive, computational(1) and computational(2).

                                              > The chair may be a bad example here, because it seems to me that
                                              > I can in fact personally decide what constitutes a chair -for
                                              > example that rock over there becomes a chair the moment I choose
                                              > to sit down on it to rest. The case of cognition and computation
                                              > may not be quite so straightforward but in essence I see no real
                                              > reason that we cannot decide what we consider to be cognitive or
                                              > whatever. These notions are not yet written in stone, so provided
                                              > that we do justice to the common conceptions and are
                                              > coherent/consistent then others may come to accept our views too.

                                              Indeed, we need to come up with some objective way of determining
                                              these things, perhaps even chairs. If you say it's a chair, and I
                                              say it's not, what is "the case"? Is everything you sit on a chair?
                                              I think not. Lord knows not everything I sit on is!

                                              More to the point, can a person designate a chair without sitting
                                              in it? We seem to agree that, at least to a first approximation,
                                              they can. Now, for extra points, can a computer system (program
                                              plus hardware plus execution) then designate a chair, even though
                                              it never will sit in it? Can a static program do so? Can a
                                              sentence of English language do so, interpreted by a person as
                                              may be necessary? If so, does the designation come from the
                                              person or the sentence? In any case, can then a program be the
                                              interpreter? ... and more stuff like that.

                                              Josh

                                              ps - I do not think that any of these questions has a concise
                                              and complete answer.
                                            • jrstern
                                              ... Jim, I think we re in general agreement on a lot of these things, thanks for your posts. To pick a nit here, I would not want to say that cognition is
                                              Message 22 of 26 , Aug 2 4:35 PM
                                                --- In ai-philosophy@yahoogroups.com, Jim Whitescarver <jim@x> wrote:

                                                > Cognition, in so far as it can be described, may be expressed as
                                                > software, and the software can run on any universal computer. The
                                                > significance of the Turing machine is that we can show universality
                                                > by equivalence to it. That is it's only useful function and only
                                                > significance.
                                                >
                                                > Jim

                                                Jim, I think we're in general agreement on a lot of these things,
                                                thanks for your posts.

                                                To pick a nit here, I would not want to say that cognition is
                                                expressed *as* software, as the cognition itself is more of a
                                                process, and software qua software is static.

                                                To pick another nit, it is in principle possible (easy) to freeze
                                                software into the hardware (state definitions) of a single-purpose
                                                Turing machine, so whatever role a piece of software might have in
                                                producing a cognitive process, it need not involve issues of the
                                                generality of Turing machines. We might know about this generality,
                                                but it simply isn't an issue in a particular case. This is the
                                                inverse argument to the kind one usually sees from Penrose and the
                                                like. I'm not sure Sloman ever quite says it this concisely, but I
                                                think it is what he is trying to say!

                                                Josh
                                              • Jim Whitescarver
                                                Interesting, I agree. The brain is not just a computer is it a highly complex dynamical information system with atoms as unwitting computational and
                                                Message 23 of 26 , Aug 2 5:53 PM
                                                  Interesting, I agree. The brain is not just a computer is it a highly
                                                  complex dynamical information system with atoms as unwitting
                                                  computational and structural units. It is a machine. A machine with
                                                  highly specialized components (organs). Its activity can be modeled
                                                  generally as an information system. Cognition could be no less of an
                                                  information system.

                                                  I do agree with much of Sloman but the Turing machine is the wrong
                                                  target in my book. It is universal and is not intended to be
                                                  practical. Being theoretical its clock rate is as fast as you choose
                                                  and have an unlimited amount of information running any conceivable
                                                  system or universe. Instead he should simply point out that the only
                                                  value of a Turing machine is in proving universality by equivalence just
                                                  because it is simple, not because it it practical. I apologize for not
                                                  reading Sloman carefully yet <shame>, my comments may have been
                                                  misdirected a bit. It is quite reasonable, and a bit daring, to
                                                  speculate on the higher order logic cognition may imply is manifest.
                                                  But if it is describable it can be implemented in universal logic, hence
                                                  the Turing machine, or equivalently a network of nand-delay logical
                                                  circuit components, or any other universal machine.

                                                  Thank you. I didn't mean to beat a dead horse here, just trying to be
                                                  clear. I am happy you could relate to my comments.

                                                  Jim
                                                  jrstern wrote:

                                                  >--- In ai-philosophy@yahoogroups.com, Jim Whitescarver <jim@x> wrote:
                                                  >
                                                  >
                                                  >
                                                  >>Cognition, in so far as it can be described, may be expressed as
                                                  >>software, and the software can run on any universal computer. The
                                                  >>significance of the Turing machine is that we can show universality
                                                  >>by equivalence to it. That is it's only useful function and only
                                                  >>significance.
                                                  >>
                                                  >>Jim
                                                  >>
                                                  >>
                                                  >
                                                  >Jim, I think we're in general agreement on a lot of these things,
                                                  >thanks for your posts.
                                                  >
                                                  >To pick a nit here, I would not want to say that cognition is
                                                  >expressed *as* software, as the cognition itself is more of a
                                                  >process, and software qua software is static.
                                                  >
                                                  >To pick another nit, it is in principle possible (easy) to freeze
                                                  >software into the hardware (state definitions) of a single-purpose
                                                  >Turing machine, so whatever role a piece of software might have in
                                                  >producing a cognitive process, it need not involve issues of the
                                                  >generality of Turing machines. We might know about this generality,
                                                  >but it simply isn't an issue in a particular case. This is the
                                                  >inverse argument to the kind one usually sees from Penrose and the
                                                  >like. I'm not sure Sloman ever quite says it this concisely, but I
                                                  >think it is what he is trying to say!
                                                  >
                                                  >Josh
                                                  >
                                                  >
                                                • jrstern
                                                  ... it ... Ach, I ve been unclear again. It s not the building that I really meant to focus on, but what it is the things we build must do, to be
                                                  Message 24 of 26 , Aug 2 6:11 PM
                                                    --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...> wrote:
                                                    > On Sunday 01 August 2004 20:16, jrstern wrote:
                                                    > > --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@c...>
                                                    wrote:
                                                    > > > I, too, think that a revised view of computation is necessary.
                                                    > > > However, I think along the lines of Marvin Minsky's www.edge.org
                                                    > > > interview: computation is not only about electronic computers,
                                                    it
                                                    > > > is the science and practice of complex systems.
                                                    > >
                                                    > > I disagree with this entirely.
                                                    > >
                                                    > > I say that computation is about how we can build nice, little
                                                    > > machines that can be filled with bits, cranked, and somehow
                                                    > > produce new bit patterns that we can interpret interestingly.
                                                    >
                                                    > Okay, this is not unlike the traditional view, but then, I think
                                                    > you don't consider such a thing as replication of DNA as
                                                    > computation.
                                                    > I do, because I think it would be inconsistent to say that
                                                    > 1) We can build a computer out of DNAs, proteins, cellular
                                                    > machinery (proven as you know)
                                                    > 2) Protein transcription, DNA replication, etc. can be simulated,
                                                    > and indeed
                                                    > their computational properties help us to develop genetic
                                                    > engineering.
                                                    > 3) Replication of the DNA is not computational

                                                    Ach, I've been unclear again. It's not the building that I
                                                    really meant to focus on, but what it is the things we build
                                                    must do, to be computational.

                                                    Is the replication of DNA "computational"? Very hard to say.
                                                    Let's try to find clearer examples.

                                                    But certainly, computational things can evolve, they do not
                                                    have to be purpose-built. I need to be clear on this, because
                                                    of the natural systems (eg, planets) I want to exclude, and because
                                                    there are teleosemanticists out there who put a lot of weight on
                                                    the historical provenance of things, as I absolutely do NOT!

                                                    > That is, why I find it hard to say that computation is by
                                                    > definition those things that we build. Then, it would
                                                    > automatically follow that our brains are not computers, and
                                                    > anything we build will be categorically different than brains.
                                                    > There would be no need for ai-philosophy, either :) [IMO, that
                                                    > is so, but I reckon this view can be contested]

                                                    Yes, yes, I apologize for making you spell this out.

                                                    > > Just what processes are involved in generating sunspots is simply
                                                    > > not in the same field of study as wondering how it is that we,
                                                    > > as intelligent agents, can *talk* about sunspots.
                                                    >
                                                    > I disagree. :)

                                                    Well, I'll just put you down temporarily as "undecided". :)

                                                    > To be able to talk about sunspots scientifically, you have to
                                                    > understand the processes underlying them (ie. you need a
                                                    > scientific *theory*). Otherwise, you could ascertain that
                                                    > sunspots have a metaphysical (usually means magical) linkage
                                                    > to the ratio of skin cancer on earth, and that the sunspots
                                                    > are projections of demons on the surface of sun. That, too,
                                                    > is talk.

                                                    Quite. So we now have *three* things on the table, scientific
                                                    theories, unscientific theories, and talk. I'm interested only
                                                    in the talk. I might build a robot which is entirely mistaken
                                                    about the facts anytime it converses about science, and which yet
                                                    passes the Turing test. In fact, that is almost advisable! :)

                                                    > > This disconnect
                                                    > > between methodologically solipsistic computation and distal,
                                                    > > ontological truths is crucial to having any conversation at all
                                                    > > about computation Wittgenstein said so, and it's the rightest
                                                    > > thing he ever gave us! (That was the "linguistic turn" (LT).
                                                    > > As it worked out, the *reasons* he gave for LT were basically
                                                    > > positivistic and skeptical, and, well, for the most part, wrong.
                                                    > > When Turing came along and talked to Wittgenstein circa 1937,
                                                    > > Wittgenstein had every opportunity to see how computation offers
                                                    > > an entirely different rationale for LT, but Wittgenstein rejected
                                                    > > it. If only old Ludwig could spend a little time with us here
                                                    > > in 2004, surrounded by computers, I think he would have
                                                    > > to change his tune!)
                                                    >
                                                    > Yes, but I think Wittgenstein's true reasons might be different
                                                    > (such as theological), we cannot know at the moment.

                                                    Metaphysical, perhaps, but not theological, as such, afaik.

                                                    > I am not certain what Turing said about Linguistic Turn. Do you
                                                    > have any references for that? (Preferably online)

                                                    Turing? I do not believe he ever used the term. He eschewed it
                                                    entirely, and to the best of my knowledge, never gave any explicit
                                                    credit to any philosophical theory. He was, after all, a
                                                    mathematician, or something of the sort. One has to do quite a
                                                    bit of historical reconstruction to see what was current when he
                                                    wrote, and how his writings fit in. I believe LT was used to
                                                    describe Wittgenstein's work, I'm not sure Wittgenstein ever used
                                                    it about himself. I see Turing as basing his 1936 paper very
                                                    heavily on Wittgenstein's work circa TLP, and Turing's 1950 paper
                                                    very much on Wittgenstein's then unpublished later works, now
                                                    available under a dozen titles including PI.

                                                    > Yes, it is a true physical encoding. Which brings us to the
                                                    > question: do not the bits on your hard-drive have a true physical
                                                    > encoding? If they are not physical, how can the disk-head, a
                                                    > purely physical entity, read this encoding?

                                                    Yes indeed the bits have a physical encoding! I believe this is a
                                                    crucially important point.

                                                    > Another question: how easier is it to observe "strings of
                                                    > nucleic acids", than the "arrangements of magnetic properties
                                                    > on the surface of a disk"? Why did it take decades to "crack"
                                                    > the human genome, then?
                                                    >
                                                    > I think you might need a stronger argument for the necessity
                                                    > of methodological solipsism.

                                                    You've lost me with these questions. I'd say that with today's
                                                    technology, observing the magnetic domains on a modern disk drive
                                                    is just marginally simpler than decoding a string of DNA. Let's
                                                    call them equal. Then what? It took decades to crack the human
                                                    genome because the data rate of decoding in 1980 was very slow.
                                                    Today, they do it much faster (it's not entirely accurate, you know,
                                                    but it is very fast!) Within a few years, they will
                                                    probably be able to decode your personal billion base code within
                                                    an hour or so. Doc McCoy's medical triquarter can probably do
                                                    it in realtime. However, I don't see how the time constant bears
                                                    on the question of methodological solipsism (MS). Every time my
                                                    Pentium executes an instruction, it does so methodologically
                                                    solipsistically (is that the right English construction???).
                                                    Perhaps MS is the right way to look at DNA replication - my
                                                    individual cells do not know that they are producing a liver cell,
                                                    they just go about their business as if there was nothing else
                                                    going on in the world.

                                                    I wonder if I've addressed your point.

                                                    > I see. But the term GOFAI may be misleading. It is said that GOFAI
                                                    > is a term used only by people who are not familiar with AI. Do any
                                                    > AI proponents actually use that (somewhat satirical) term?

                                                    I advocate AI, yet I'm afraid that some satire about early AI
                                                    efforts is well-earned.

                                                    > > I am in great sympathy with everything that Smith and Sloman are
                                                    > > attempting, but they make a mistake in dismissing Turing
                                                    > > computation, which needs to be better understood. I suggest
                                                    > > a historical study of Hilbert, Brouwer, Frege, Wittgenstein,
                                                    > > Tarski, and a few others, shines new light on what Turing did.
                                                    > > Seeing it requires the kind of 20-20 hindsight we can have
                                                    > > today, with our empirical knowledge of computers-in-the-wild,
                                                    > > as Smith says.
                                                    >
                                                    > I agree. But I think we must also take into account the work
                                                    > of later mathematicians.

                                                    You seem much more conversant than I with such modern work, but in
                                                    my opinion the important matters are more qualitative and not
                                                    exactly mathematical -- I think the matter of computation needs
                                                    to be pursued as a philosophical project, with math coming later.

                                                    > If you remember my (somewhat strange) arguments for wide
                                                    > linguistic idealism, I had argued that computation is indeed the
                                                    > generalized form of linguistic analysis. And that computably
                                                    > enumerable computation is the most fitting candidate for LoT.

                                                    I may have missed that, but it sounds like we'd be in general
                                                    agreement. Or not -- it may be one of those cases where we
                                                    use conflicting terms to describe a single concept. I avoid
                                                    any kind of "idealism" like the plague. And my basis for
                                                    discussion almost always starts with the solipsistic side
                                                    of things. Aha, perhaps that's why you questioned my use
                                                    of MS above!

                                                    > I think I cited it correctly, but of course it may be fun to
                                                    > watch it for yourself. I am not sure if Minsky told Quine
                                                    > that, though. If it is not logical or physical,
                                                    > then what is it? :)

                                                    * Sui generis.
                                                    * Computational.
                                                    * Process.

                                                    I always refer to Hacker's book, "Wittgenstein's Place in
                                                    Twentieth Century Analytic Philosophy", in which he discusses
                                                    Quine's "apostasy" to the linguistic turn in celebrating science,
                                                    with its ontological commitments, as true philosophy. Another
                                                    version of this same tidal turn, occurring circa 1970 plus or
                                                    minus fifteen years, is available in Suppe's "The Structure of
                                                    Scientific Theories" Unfortunately, neither is the least bit
                                                    concise. And neither talked about computation, as such, but
                                                    they at least locate Wittgenstein and Quine for us contra the
                                                    modern philosophical and meta-scientific theories which
                                                    prevail today.

                                                    > I am opposed to the widespread Platonist interpretations
                                                    > of computationalism myself. I think accepting a Platonist stance
                                                    > or not makes a big difference philosophically. I'm trying
                                                    > to construct an opposing view at least for the sake of an
                                                    > alternative. Digital physicists are making these cryptic Platonist
                                                    > remarks. How do they know it is true? :)

                                                    They don't. Any physicist may pretend to be platonist, but
                                                    really they don't care, they've all been instrumentalist since
                                                    quantum theory, whatever they pretend.

                                                    > In addition, I think that computation is purely physical.
                                                    > I am also sympathetic with nominalism, but our views with
                                                    > you seem to differ considerably. (Maybe because I am not truly
                                                    > a nominalist?)

                                                    Well, nominalism is basically the alternative to platonism,
                                                    so I'm glad we're together at least that far. I think I'm
                                                    pushing a more extreme (and not fully elaborated, and certainly
                                                    not published) form of nominalism. If you really want to be
                                                    a physicalist, and want to naturalize computation to conventional
                                                    (and I mean well within Newtonian parameters) physics, I think
                                                    you will eventually have to join me out there!

                                                    > However, I'm not a solipsist by any measure. I think there is
                                                    > one objective reality, and our minds (and kidneys) are a very
                                                    > physical part of that reality. What we call subjective, in my
                                                    > opinion, is simply variety and complexity of existence
                                                    > (And therefore of thought). I think I'd mentioned
                                                    > that I'm a monist.

                                                    Perhaps I disagree on the subjectivity. You seem to see it as
                                                    some sort of epistemological limitation on rationality. I suppose
                                                    I've heard that sort of argument before. I certainly do not hold
                                                    it now. Of course any cognitive agency is going to be something
                                                    less than omniscient, but I do not think that that constitutes
                                                    subjectivity. It is rather a question of how a cognitive agent
                                                    knows _anything_ _at_ _all_, and right or wrong, I count that
                                                    as subjective!

                                                    I guess I should be explicit, that in my theory there are two
                                                    levels, with some form of what Fodor calls "non-strict reduction"
                                                    between them. The higher level is contentful. The lower level
                                                    is methodologically solipsistic. This is only Fodor's metaphysics,
                                                    more or less, which he derived from computation as it was known
                                                    and loved circa 1975. In this reading, if you're going to talk
                                                    about digital electronic computers as we know them, you are of
                                                    necessity going to be talking about MS. That is certainly not
                                                    the *whole* story, but there is no story, until and unless it
                                                    includes the mechanical level, which pretty nearly must be MS.

                                                    There may be just one objective reality, and I may believe it
                                                    and nothing but it, but I do not think I can prove it, and I do
                                                    not think I need to prove it, in fact, I do not think I can
                                                    assume it, and never need to assert it, in order to achieve
                                                    computation, much less cognition. That's just the LT: no
                                                    ontological commitments.

                                                    > By formal, do we not simply mean that the manipulations are
                                                    > governed by clear mechanical rules?

                                                    These are matters of definition and connotation, most of them
                                                    best avoided, but fwiw, it seems to me that "formal" tends to
                                                    be ideal, non-physical, synchronic (atemporal), while "mechanical"
                                                    and/or "rules" tend to imply physically realizable, causal, and
                                                    diachronic (temporal). I'm pretty much allergic to the term
                                                    "formal" these days.

                                                    > > In any case, as above, work on logic is not prior to work on
                                                    > > language.
                                                    >
                                                    > I could agree with that. I find that the logicist views of
                                                    > language and mathematics can be quite misleading.

                                                    Most people would argue with us, you know. It seems to me that
                                                    putting logic first goes with the ontological, scientific viewpoint.
                                                    If you want to put *anything* prior to logic, people are going to
                                                    look at you funny.

                                                    > Yes, it was, and I think the traditional view does this
                                                    > excellently from a postmodern perspective. It shows us that many
                                                    > things in nature are in fact computational (A point that you
                                                    > would not endorse :), that they can be understood, analyzed and
                                                    > engineered in computational terms. (like say, weather forecast)

                                                    Postmodern? You mean, it strikes you as gibberish? :)
                                                    I like Derrida as much as the next guy, but there are very,
                                                    very few people who can make it work the way he does, and he's
                                                    just a joker half the time, anyway.

                                                    Yes, sure, Scheutz and Agre both refer to critical theory
                                                    which makes them at least honorary postmodernists, but as I
                                                    said before, that's dangerous business!

                                                    "No ontological commitments", that's what I keep telling you,
                                                    that's the linguistic turn and what could be more postmodern!
                                                    I guess Wittgenstein was the first postmodernist, ... preceeded
                                                    only a little bit by the sophists! There *is* nothing new
                                                    under the sun.

                                                    Let's just stipulate that I disagree with such ontological
                                                    statements and move on.

                                                    (Note that my disagreement with ontological statements is NOT a
                                                    disagreement about any ontological fact!!!)

                                                    > I think you have some interesting arguments, and the capacity
                                                    > for a thorough exposition of them. I wish you luck in confronting
                                                    > the philosophical journals. (I don't have the courage myself)

                                                    We shall see. I'm relatively confident that if I think I've got a
                                                    publishable article, they will agree. But getting my own approval
                                                    on these things has proven very tough!

                                                    > I see. But I fail to understand how a computer could be made
                                                    > "informal" by embedding computational systems in the world.
                                                    ...

                                                    Well, formal versus informal is not really the issue. I think
                                                    we share some intuitions on this. For me, it's simple causality,
                                                    and that's it. The traditional problems with this are supposed
                                                    to be accounting for error, if everything is so simply connected.
                                                    I think that frankly, that's a ridiculous assertion. You earlier
                                                    asserted a simple argument against it, that "subjectivity" is in
                                                    the incomplete modelling of the universe in any agency. Error is
                                                    simply accounted for by the problems of using a (very!) finite
                                                    axiomatic base for a complex universe. Nothing more need be said!
                                                    Only, I don't think we even want to make such an argument! I have
                                                    some compositional, MS arguments I prefer instead. I think they
                                                    reach pretty much the same conclusions, but much more robustly.
                                                    Gawd, I have got to get this down in publishable form!

                                                    > Maybe, digital philosophy does give a good answer to Sloman's
                                                    > concerns, in that the very substance of physical existence can
                                                    > be computational (bridging the gap between the formal and
                                                    > informal, solipsistic and not-so-solipsistic), however, the
                                                    > philosophical arguments are rather weak so far. (That's what
                                                    > I'm working on, mainly. If you look at the archives
                                                    > of sci.physics.discrete,
                                                    > there have been quite interesting discussions)

                                                    I'm sure. I enjoyed Greg Bear's "Anvil of Stars" which involves
                                                    such speculative physics. It might even be true, for all I know,
                                                    but I do not see it as either necessary or sufficient
                                                    for cognition in even the best case. Do you think such wild
                                                    physics are necessary in order to explain why your Pentium
                                                    (or whatever) does what it does? Remember that the topic of
                                                    the day is how even mundane computation goes about its
                                                    business. Yes, now that you mention it, I do recall a few of
                                                    your postings along these lines, probably in c.a.p. Seems to
                                                    me an exercise in asserting the conclusion, a creatively new
                                                    form of teleological explanation, just short of theological,
                                                    or maybe not short of it at all. Perhaps best if I say nothing
                                                    more about it!

                                                    Josh
                                                  Your message has been successfully submitted and would be delivered to recipients shortly.