Tuesday, July 05, 2005

Different Views of Consciousness

The great game theorist Robert Aumann writes about consciousness.
Sometimes, people express perplexity as to the nature of the problem. They do not see anything mysterious about consciousness, and do not understand in what way it is different from other neurological functions like, say, the regulation of breathing. Asked whether a computer could in principle be conscious, they answer, "why not?"

We are dumbfounded by this reaction, and can only conjecture that these people are themselves not conscious. To me, it is evident that no combination of silicon chips and wires could conceivably "experience" in the sense that I do. Consciousness involves something beyond the merely physical and mechanical.

A bit of a different view than that of Manuel Blum.
The question whether an entity is CONSCS is a function of its algorithms, not the stuff (silicon or carbon) that implements those algorithms.
Why are great scientists like Blum and Aumann taking on consciousness late in their careers? One of the many possible research questions Blum threw out in his talk:
What happens when an entity stops being an entity?
So perhaps they study consciousness as a way to deal with their own mortality.

10 comments:

  1. So perhaps they study consciousness as a way to deal with their own mortality.

    That's a bit morbid. Hope they don't read this blog :)

    I saw Blum's talk at CCC05 and I didn't really buy it. If all we are are Turing machines with a model of the outside world (which in itself is a Turing machine with "unlimited resources", if I recall correctly), then at least in my mind that really limits human experiences--from a computability point of view there are experiences which we cannot have, which is reasonable; we are not gods. However, it effectively means our experiences are countable. I tend to think that the human experience is a bit more substanative; that we don't fall into some discrete domain. Further, it would mean that there is only a finite difference between any two person's experience. These, to me, seem like too severe of limitations.

    ReplyDelete
  2. The understanding of consciousness is fundamental to the question of the meaning of life which is the main question of philosophy. Every human being at some time or the other ponders this question and some among them try to think through it more than others. It is not surprising that people later in their careers feel
    more free to ponder these larger
    questions.

    ReplyDelete
  3. However, it [modelling experience as interacting Turing Machines] effectively means our experiences are countable. I tend to think that the human experience is a bit more substanative; that we don't fall into some discrete domain.

    Not necessarily. I didn't see M. Blum's talk, so I don't know exactly how he set it up. In general, though, you could use Turing Machines over the reals (or some other uncountable set?) to address your objection.

    There is, in fact, a theory of Turing Machines over the reals that has been proposed by L. Blum, Shub, and Smale. Here is an overview from Notices of the AMS:

    http://www-2.cs.cmu.edu/~lblum/PAPERS/TuringMeetsNewton.pdf

    The authors have, of course, many technical papers (and an entire book with Cucker) with more details. There are others working in this area as well.

    One of the questions that comes up which is relevant to your objection is the notion of a "transfer" - if I prove a result about a Turing Machine over a ring R1, when does that result "transfer" (or not) to a statement about Turing Machines over a ring R2? I regret I don't know this area well, but it seems to me that by setting the ring R1 to be whatever M. Blum did in his talk and the ring R2 to be the reals, we could get precise answers as to how severe or not the limitation you identify really is.

    -David Molnar

    ReplyDelete
  4. So perhaps they study consciousness as a way to deal with their own mortality.

    As someone who has been discussing this subject with Manuel for over three years, I suppose I cannot keep my big mouth shut.

    I have to say that the above psychoanalysis seems pretty far off. From my understanding, Manuel has always been interested in consciousness/intelligence. (E.g. When he received tenure at Berkeley, he took a week off and thought about consciousness in a nearby hotel.) So I think it has little to do with his age.

    Manuel has taken up this study at CMU for several reasons, and while I do not pretend to know them all, I can list some. (Note, if I thought he would read this blog, I would not attempt to answer in lieu of him.) One reason is simply that, in his position, he can do it if he wants-- and he's always wanted to. Another reason is that the idea of strong AI is a bit more kosher and prevalent at CMU than at other places (cf. Simon, Newell). Also, he found a grad student who seriously listens to his ideas about it. ;)

    --ryan williams

    ReplyDelete
  5. Asked whether a computer could in principle be conscious, they answer, "why not?"

    This is more easily blamed on the averaged person's lack of understanding of how a computer works than on their understanding of consciousness. (Although I agree most people have thought hard about what exactly consciousness is.)

    ReplyDelete
  6. Has any one else here read the Simulation Argument and been disturbed by it?

    If indeed it's easy to simulate the physical world on tiny computers, an advanced civilization could be running billions of different Earths right now (so, odds are, we are one of the simulations). The laws of physics in our own universe don't even have to be the same as the simulator's universe.

    ReplyDelete
  7. I think Karl Popper got it right in claiming that any scientific hypothesis should be falsifiable: It should be possible to put the hypothesis to an experimental test. So - is there any hypothesis about consciousness that can indeed be falsified by some hypothetical outcome of a physical experiment?

    ReplyDelete
  8. To me, it is evident that no combination of chemicals, electrical signals and flesh could conceivably "experience" in the sense that I do. Consciousness involves something beyond the merely physical and mechanical.

    (see also: "Learning to be me" in Axiomatic by Greg Egan)

    ReplyDelete
  9. is there any hypothesis about consciousness that can indeed be falsified by some hypothetical outcome of a physical experiment?

    Of course consciousness is problematic to define. But the cognitive psychologists have been doing experiments for quite a while that support (falsifiably so) theories about mental states that could reasonably correspond to informal notions of conciousness. The backlash against behaviorism from the sixties on is exactly that a philosophical program supported by falsifiable experiment.

    Mitch

    ReplyDelete
  10. "However, it effectively means our experiences are countable. I tend to think that the human experience is a bit more substanative; that we don't fall into some discrete domain."

    Natural numbers can be really large, you know.

    ReplyDelete