[extropy-chat] Twenty Questions Twenty Questions - Theneural-net on the Internet

[spike]

> -----Original Message-----
> From: extropy-chat-bounces at lists.extropy.org [mailto:extropy-chat-
> bounces at lists.extropy.org] On Behalf Of Erik Starck
> Sent: Friday, February 18, 2005 10:02 AM
> To: ExI chat list
> Subject: Re: [extropy-chat] Twenty Questions Twenty Questions - Theneural-
> net on the Internet
> 
> At 12:38 2005-02-18 Jeff Medina wrote:
> >Erik: "Don't know if specific brands of objects are allowed in the game."
> >
> >20Q.net: "The object you think of should be something that most people
> >would know about, but, never a specific person, place or thing."
...
> 
> Anyway, would be interesting to see this technology integrated with the
> search engine of a domain specific web site. For example, using it to find
> that book you forgot the name of.  Erik


The important issue is how effectively the 20 questions
software learns.  If enough people play the game, then
eventually it should get something specific like brand
and model of a car or bike.  That would be cool.

Recall the eliza experiment a couple years ago where
someone rigged eliza to the internet teen chat group.
He had a whole bunch of teens converse with eliza,
never realizing it was software.  Many of those who did
figure it out did so because the software would respond
instantly, faster than a person can type.  That can
be fixed.  That looks to me like machines have passed
the Turing test.

Next step is to take the 20 questions software and do
a Turing test with informed contestants.  This software
is good enough that it is difficult to tell if a human
or a machine is working the other end.  Friends, we
have survived to see computers pass Turing's criterion.

spike