[ExI] turing test, was: RE: ai class at stanford

spike spike66 at att.net
Tue Aug 30 01:35:36 UTC 2011

>... On Behalf Of Adrian Tymes
Subject: Re: [ExI] ai class at stanford

On Mon, Aug 29, 2011 at 3:53 PM, Mike Dougherty <msd001 at gmail.com> wrote:
>> ... but where/when do you call it a person and grant it all the 
>> rights that people currently hold?

>...The current standard, so far as there is one, is when it passes the
Turing Test.

That notion is oversimplified.  In a limited sense, an Eliza derivative,
which is really merely a lookup table, has already passed the Turing test.
A few years ago a guy created a teen-speak version of Eliza and turned it
loose in a teen chat site.  Granted it was a limited version of the Turing
test, since the teens did not generally know there is a chat simulator, but
some of them fell hard for it.  One conversation went on for over fifty
minutes.  Even then, the reason the twit caught on is by noting that the
answers were coming back too fast to have been done by a human.

In the recent chess tournament, the opponent of the cheater were initially
fooled, thinking they were playing a human when they were actually playing a
computer.  This limited Turing test was initially passed, then subsequently
failed, after the arbiters became suspicious that an untitled player was
scalping experts and masters.  In that case, the Turing test failure was
only uncovered by results, not the "dialog" of the chess game.

When a computer passes a rigorous version of the Turing test, we may
discover that human intelligence is for the most part an enormous lookup
table, or is far simpler in common practice than we always assumed.  Eliza
is a far more interesting conversationalist than at least some humans


More information about the extropy-chat mailing list