Re: Implementation

From: Hans Moravec <hpm.domain.name.hidden>
Date: Tue, 27 Jul 1999 21:07:53 -0400

Russell Standish <R.Standish.domain.name.hidden>:
> With any HLUT algorithm, the machine will eventually fail the Turing
> test if we quiz it long enough.

Dying after a normal lifespan doesn't count as failing the Turing
test. In fact, it is a necessary condition for faithfully imitating
a human, who ages and dies.

> The situation is probably even better than that - the HLUT
> algorithm is most likely of higher computation complexity that the
> Turing test required to falsify it.

This is nonsense, and also irrelevant. The complexity is the smallest
program that could produce the table. Maybe a conventional AI
program, but so what?

> However, in the context of our Olympia story, we know the questions
> of the Turing test before hand. I think you will agree that that it
> is trivial to construct a nonconcious entity that will pass that
> particular Turing test. In retrospect, I was being a little vague in
> my argument.

I certainly do not agree. I impute consciousness to characters
in a movie as easily as to people I watch live through a security
camera.

>> An HLUT ...
>> deemed to "die", presumably finite on all branches.

> Untrue. This would imply that all computations halt.

The computation can halt when the HLUT-encoded entity
dies. Having a finite lifetime doesn't negate your
consciousness while you're alive.
Received on Tue Jul 27 1999 - 18:10:42 PDT