[net.followup] Turing the interior of Chinese boxes

trc (11/24/82)

It is true that the person in the box doesn't understand Chinese. 
However, the blind spot in the argument is obvious - the person is NOT
alone in the box!  There is a book of Chinese language knowledge in the
box too!  In order for this book to provide correct answers to questions
in Chinese characters, it must contain information about the relationships
of those characters.  In order to truly pass the Turing test, the information
must be of a level at least equivalent to that in a system that understands
the meaning of those characters.  Together with the information processing
capabilities of a human (or a computer for that matter, since the rules
are presumed to be formal ones), the SYSTEM can be claimed to be intelligent
if it is able to pass the Turing test.

As to looking at the hardware to determine if the system is intelligent,
this sounds suspiciously like a code phrase for "if it doesn't look like
the brain, it cant be intelligent".  And if it is intended in a literal
sense, it is worth noting that any piece of hardware can be simulated
with a computer and an appropriate program - which is what Turing was 
getting at with his "Turing machine".

Actually, the Turing test (as commonly conceived) might have one flaw  -
it relies upon human judgement of intelligence.  However, human beings
can be fooled.  An intelligent alien, with alien concepts, but
having a good grasp of Chinese, might still fail the test. A non-intelligent
computer, with a tricky program, such as Eliza, might fool humans for
a while into believing it is intelligent.  Unfortunately, there doesn't seem
to be any straight-forward objective test for intelligence, so human
judgement is the best alternative.

The solution to this difficulty seems to be "the test of time".  For
as long as a system demonstrates intelligence, it will be given the
benefit of the doubt. If it seems to fail the test, it will be tested
extensively to make sure it was a true failure, not merely a mistake
or mis-understanding.  Actually, I believe that Turing may have said
this, but did not emphasize it, believing it to be obvious.  Does
anyone have a copy of his original proposal handy?


				Tom Craver
				houti!trc