Re: Searle's Chinese Room Argument

From: Harnad, Stevan (harnad@cogsci.soton.ac.uk)
Date: Tue Feb 18 1997 - 17:46:05 GMT


> From: Smith, Damian M <dms93fm@soton.ac.uk>
>
> Regarding the Turing test:
>
> Isn't it a bit on the hard side? Why must the attribution of
> consciousness to a computer be limited to the situation where that
> computer flawlessly emulates a human being?

Because we don't know what's just a "flawed" consciousness and what's no
consciousness at all. Consider the Turing test to be "conservative."
Maybe less than that could still generate a mind; but requiring that the
candidate be indistinguishable from us in what it can DO makes it
least likely that we will wrongly assume something has a mind when it
doesn't.

Yes, the test is hard, and the way to passing it will be littered with
Nobel Laurels.

> Hypothetically, is it not possible to imagine a situation where one
> might form an impression of consciousness existing in a computer,
> without it necessarily having a property of being indistinguishably
> human in nature?

It's very easy to think a computer has a mind. Sometimes people
think computers that phone them with adverts are real people with real
minds. But the point of the Turing Test is not to fool some of the
people, some of the time. It's to make a candidate that really has all
of our capacities. Then, if mind is as mind does, it should have
a mind.

No guarantees, though. A Turing-Test-passing robot could fail to have a
mind and a thermostat could have one. This is just going with the odds.
The assumption is that whatever it was that generated minds in living
creatures, it must have had to do with their ability to survive and
reproduce. Both surviving and reproducing are things organisms DO.
So maybe if we can design a system that can do all that too, then it's
mental lights will go on too.

But of course we need to start with the amphioxis or even lower, but the
trouble is, we are not nearly as good at judging whether an amphioxis
has passed the amphioxis Turing Test as we are in judging the human
Turing Test. You're tempted to say "Fine, then test the robot amphioxis
with real amphioxises, to see whether they can ell the difference.
But unfortunately, though amphioxises are surely conscious, they are
not very thoughtful. SO it's easy to fool them; you don't even
need the T-Test. Just dab the right odours on a tephlon glove and
they'll have a go (either mating or combating, depending on what
odourant you use).



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:49 GMT