next up previous
Next: Coda: What went wrong? Up: Patrick J. Hayes: What Previous: Objective computerhood?

Reducible to absurdity?

Part of the motivation for finding a better account of being a computer, like the `magic-paper' account I sketch, is precisely to capture the intuition that the digestive system, Searle's office wall, etc., are not computers. (Or if they are, they are extremely trivial computers, rather as a rock is a computer with one symbol which means ``do nothing.'') I take this as simply obvious: my digestive system doesn't, in fact, do word processing or compute tables of prime numbers. If our existing accounts of computer and computation predict that it does, or even can be seen that way, there is something wrong with them. Rob Shufflebeam agrees:

[Stufflebeam:] Physical devices/systems don't suddenly become symbolic-digital computers because their behavior has a symbolic interpretation. The difference between as-if symbolic computation and real symbolic computation is a difference that makes a difference.

[Hartcastle:] Though I am sympathetic to Hayes' general approach (computations are things that computers do, so we need to define what a computer is first), I worry that the answer provided does not adequately counter the critic's claims.

First, regarding the specific suggestion that to be a computer requires a certain level of complexity: this does not seem to answer Searle's real challenge, viz., that his digestive system does everything that Hayes claims a computer does, yet we don't think of digestive systems as being computers.

To avoid triviality, however, a magic-paper computer has to be not just complex in the sense of having a large number of physical states, but also these states must exhibit tokens of a language with a productive syntax which can be interpreted as referring to the state-transitions which are the causal consequences of the state. A machine becomes recognisable as a computer precisely when this data-structure language is more complex than the trivial language obtained by assigning a unique token to each state, so that it can be accurately described as storing and processing information. This is a much more demanding criterion for any proposed Putnam- or Searle-type construction.

However, this is only a sketch, and details are not yet worked out; so I will join with Valerie Hartcastle in her last word:

[Hartcastle:] I raise more questions than I answer, and for that I also apologize.

next up previous
Next: Coda: What went wrong? Up: Patrick J. Hayes: What Previous: Objective computerhood?

Fri Jul 25 22:00:35 MEST 1997