Re: 3 "Real" Conscious Machines [WAS Re: Singularity: A rock 'em, shock'em ending soon?]

From: Charles D Hixson (
Date: Fri Jan 20 2006 - 15:51:58 MST

On Wednesday 18 January 2006 06:56 am, Woody Long wrote:
> > [Original Message]
> > From: Ben Heaton <>
> > To: <>
> >
> > On 1/17/06, Woody Long <> wrote:
> > > ...
> > > ...
> > > purposes a conscious machine.
> >
> > So you're saying that for a machine to be conscious, it must
> > understand the input it receives. Do you have an idea for a test that
> > can be used to determine whether a particular machine meets that
> > requirement?
> >
> > -Ben Heaton
> Yes, for a machine to be conscious, it must (at least) understand
> (receive/process) language inputs, in the same way as human level
> consciousness understands (receives/processes) these language inputs.
> Ken Woody Long

This requirement makes it impossible to currently achieve. Currently no one
knows HOW human level consciousness understands. We don't even have decent
definitions of either consciousness or understands.

Propose an operational test, instead. Propose a test that can actually be
attempted by both people and machines. Then we can judge whether the test
reasonably proves what we consider the requirements of consciousness to
require. (The judgement should be done before we know the results of the
test, of course. People have a long history of cheating, here, by moving the
goal posts.)

Personally, I think that the fact that the Turing test is coming under fire
means that people are starting to think that computers might stand a chance
of passing it. When they clearly couldn't, nobody objected to it.

OTOH, these machines are showing up years before they are expected. Now I
personally believe that "human intelligence" is a much simpler and more
limited phenomena that is generally believed, but I don't expect to see human
level intelligence in computers within the next five years, perhaps a bit
longer. (And then it will need to *learn*!) I'm willing to be proven wrong,
but this year it will take quite a bit of proof.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:55 MDT