How many bits do you think the brain uses to reproduce such
"information"? Or does it simply know what salty is like? And might
such phenomenal abilities to "instantiate information" be one reason
for such "common sense" intelligence that abstract machines still
lack?
Wouldn't an honest and intelligent abstract machine be able to
recognize that it doesn't really know what salty is like, just as you
must admit that you don't know what salty is like for me, unless salty
is the same for me as it is for you?
"
I believe that your approach to this issue results from your view of
perception, and communication. You have taken a top down approach to
the problem of abstract sensations - "I know what salty is, even if I
can't explain it" - and seem to be implying that an A.I. should start
with this idea about what salty is, before it trys to experience. It
would seem to me that this problem requires a bottom up approach -
surely humans new how to taste salt before they could express this in
words?
So let it be for A.I.s. Using an electronic nose, or some other such
contrivance, they will have their own experience of the taste of salt -
based on the feedback they get from their input device. This experience
of the taste of salt will be no less valid to them as our experiences
are to us. You have made a good point though - that an A.I. would not
be able to compare it's experience to that of a non-A.I.
On the subject of the copy paradox in general, specifically the
difference between the original and the copy:
Surely at the point of creation - the instant of copying - and assuming
the copying is exact - the original and the copy are indistinguishable.
The copy "is" the original to all intents and purposes. A copying
process which is not exact, or one in which the copy has a different
experience of the copying process to the original, is not valid for this
thought experiment.
For another view of this subject, I recommend the first part of the
short story "Dust" by Greg Egan.
Robert Bricheno
bricher@msmsd2.hoechst.com