Putnam Minds and Machines

AssignedTopicsPutnamMindsandMachines

ResponsePapersPutnamMindsandMachines

-- ShaughanLavine - 13 Nov 2005

Putnam may have invented functionalism, but this paper is not about functionalism. That is, functionalism is generally thought of as claiming that minds bear some relevant resemblance to computer programs, but this article nowhere claims anything like that.

What he says is that Turing machines::computers is a legitimate answer to the analogy question minds::brains

That comparison does not require that minds be in any way like Turing machines. After all, yellow::color is a legitimate answer to leopard::feline.

When I say that the comparison does not require that minds be in any way like Turing machines, what I mean is that no theory of how the mind functions is any more susceptible to the analogy than any other. Turing machines, connectionism, ectoplasm, ... . The point is wholly independent of such categories.

I construe the argument in a deviant way. I take the argument to show that programs (Turing machines) are second-realm entities. (Remember Frege: three realms. First realm, physical; second realm, mental; third realm, purely formal, mathematics.)

First, let me define isomorphism between mathematical structures: A mathematical structure consists of some domain plus some interpreted vocabulary on the domain. Example: domain the natural numbers; vocabulary, 0,1,+,x. Two structures are isomorphic if there is a one-to-one correspondence between their domains that preserves the vocabulary. Example: suppose we had an infinite set of wooden number blocks. The two structures have different domains with different properties: one has wood things in its domain, the other abstract things, but, for mathematical purposes, it makes no difference which we use. That is why mathematics can be applied (both to physical systems and to other mathematics). As some put it, the same (isomorphic) mathematical structure has many realizations.

One popular view in the philosophy of mathematics today is called structuralism. Resnik wrote a book called Mathematics as a Science of Patterns. The idea is that only structural properties matter in mathematics, what we care about is independent of particular realizations. This amounts to, if you will, defining the third realm as the realm of structure (pattern).

Putnam's characterization of Turing machines makes them look a lot like third-realm entities. However, it is not true that anything isomorphic to a Turing machine (a program) counts as a realization of the program. To count as a realization, the "next state" relation must be a temporal relation. In addition, I think, to the extent that one endorses causal talk at all (and, as we know from Broad, if you don't, the analogy is of no interest) it must be that each state and "tape configuration" causes the realization to enter the appropriate next state and next "tape configuration."

The causal claim is suggested because we would not take accidental behavior in accordance with a program to be a realization of the program, just as we don't take the six monkeys typing the complete works of Shakespeare as ... .

The temporal claim is supported by making up a mathematical structure based on a Turing machine table: a "Turing structure" is a pair of functions from the natural numbers to (a) states and (b) "tape descriptions" such that the values of the functions at n are appropriate to the nth stage of the Turing machine computation. That is "isomorphic" to the Turing machine but is not a realization of it.

So, the first realm is physical, the third realm is "up to isomorphism", the second realm is up to some limited constraints, in my example, temporal, causal, and otherwise structural.

This is what I take Putnam to be claiming: that both programs and minds are, in my peculiar sense, second-realm entities, and that all the traditional problems of body and mind are problems for the second realm.

Let me now give two examples to drive my point home that the argument doesn't require that minds be like programs.

First, suppose we had a complete physical theory, and that the mind supervenes on the physical structure of the body. If we knew the complete physical configuration of someone's body, including the brain and the sense organs, muscles, ... . We could then write a program that simulated the complete physical system, with sensors and transducers that connect it to external inputs and outputs appropriately. Such a "robot" would function just like the person.

Second example, suppose we have a complete explicitly formulated theory of human psychology of any sort whatsoever (suppose it is realized in ectoplasm, has nondeterministic components, whatever). Since the theory is complete and explicitly formulated, we can infer its results using a program. Hence we can build, once more, a robot.