[Martin Taylor 920924 14:45]
(Rick Marken 920924.1000)
Rick, it's just a fluke that I saw your response so quickly. I'm still some
150 messages behind in my catching up. But since I did...
I agree that, as far as I know, the Turing Test has not been used in the
context of the PCT Test. I read your posting as saying that it intrinsically
would not be so used.
If you think about it in the light of applying the PCT Test, there is still
a problem, because in any human hierarchy there will be thousands of controlled
or controllable variables, and only at the top (my view, not Bill P's) there
will be the intrinsic variables. The computer has an entirely different set
of "intrinsic variables" that have been designed into its software. One
could conceive of distinguishing reliably between a silicon and a carbon
intelligence if one could do things that might affect these intrinsic variables
(one would observe the actions that related to their control, not changes
in them, of course, and these actions would be exactly what you were talking
about--the behavioural manifestations, not the CEV, which are unobservable
in the case of intrinsic variables).
Also, in the case of a linguistically communicating machine, the only percepts
that it could control are those it gets from the Turing Tester, so if it is
to control anything, it must do so by eliciting what it needs from the Tester.
As I understand the Harnad TTT, the subject has access to outside sensory
and motor connections (outside the dialogue, that is). Such a machine could
be observed to control variables that the Tester could disturb, so it would
be more likely to be able to show behaviour (i.e. different levels of control)
that would be hard to distinguish from a biological intelligence.
Martin