[FOM] The Lucas Penrose Thesis
wood0571 at tc.umn.edu
Mon Oct 2 21:33:27 EDT 2006
Eray Ozkural wrote:
> However, at any rate, we should be able to agree on that the
>meaning of a natural language term as ordinarily understood by us is
>nothing more than an aggregate of mental states (i.e. states of perception)
>that are in our memory.
>Thus, we can make the computer watch video tapes that tell the meaning
>of these word, and the computer will be able to associate the sensory
>data with the word, by using machine learning algorithms.
I'm not sure why we should agree with this. The picture of meaning
espoused here is a wildly controversial one. It's been a wildly
controversial one since it was first proposed by the British Empiricists
in the 17th century. Part of the implausibility of the picture comes
from the fact that finding an "aggregate of states of perception" which
can uniquely specify the meaning of any natural language term is nigh
impossible. Further, it would then seem as if it is possible (in fact,
likely) that the meaning of a term like "rabbit" is different for me
than it is for you (since we almost certainly possess different
aggregates of perceptual states).
With respect to the notion of mathematical meaning, Kripke's exposition
of the difficulties in specifying that we mean "plus" by "+" in "2+2=4"
and not "quus" where "quus" is some function like:
quus(x,y) = plus(x,y) unless x > 10^10^10 or y > 10^10^10, in which
case quus(x,y) = 5
come to mind. It's hard to imagine an aggregate of states of perception
could distinguish between "plus" and "quus". Note that I'm not saying
(necessarily) that the picture of meaning you present is mistaken, but
simply that it isn't obviously correct. Given the many problems that
arise for it, perhaps the claim that we should all be able to agree to
it is a mite bit hasty.
More information about the FOM