From: Dan Fabulich (daniel.fabulich@yale.edu)
Date: Mon Nov 29 1999 - 19:40:08 MST
'What is your name?' 'Eliezer S. Yudkowsky.' 'IT DOESN'T MATTER WHAT
YOUR NAME IS!!!':
> I think qualia are the result of non-Turing-computable processes that we
> probably can't duplicate artificially with today's technology, so why
> bother?
Having read your website, you hint at why you believe this, but, to the
best of my knowledge, you never specify why. You mention that it has
something to do with the problem of defining instantiation, but I have no
idea what you might be referring to.
In thinking about the problem only a little bit, it seems to me that what
you may be referring to is what philosophers of language refer to as the
problem of "translation" from one language into another. That is,
defining instantiation is easy: A instantiates B when A yields the same
output as B for any input. The problem is ascertaining whether the I/O
is, in fact, the same.
This might seem easy at first if the I/O are written on similar medium (if
they both use punch cards, say, or magnetic tape), but that doesn't go far
enough. Surely, for example, a universal Turing computer designed with
white/black paintbrushes writing on colored paper could instantiate a
Turing computer designed with magnetic tape. Right off, however, you find
the necessity for a translation rule: Maybe 0 = black, 1 = white.
This problem gets worse and worse as you start trying to test the outer
boundaries of the language. How weird a Turing computer can you build
that still instantiates the first one? What constitutes a well-defined
translation rule? (Hofstadterian considerations apply here.)
Am I on the right track here?
-Dan
-unless you love someone-
-nothing else makes any sense-
e.e. cummings
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:05:53 MST