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

[spike]

>...On Behalf Of Emlyn
Subject: Re: [ExI] turing test, was: RE: ai class at Stanford

>... And we talk to the man who coded Cleverbot, a software program that
learns from every new line of conversation it receives...and that's chatting
with more than 3 million humans each month...

http://www.radiolab.org/2011/may/31/  --  ...Emlyn

JA!  This is exactly what I have been wondering about for some time.  I
think it is a simple enough coding task, even I might be able to do it
singlehandedly, in you-know-what.  In a teen chat room, the dialog is simple
minded indeed, scarcely able to pass for artificial idiocy.  I think I could
write code that reads the archives, creates from that a lookup table and
after reading a few megabytes, can generate passable teen drivel on its own.
Then with fresh responses from unaware teens, it could continually expand
and update that table to keep it stocked up with the latest hipster terms
and usages.

NOW THINK ABOUT THIS, for it would meet at least the loose definition of a
learning program, it would pass the Turing test for at least the unaware, it
would eventually be good enough that it would make a passable companion to
those who are scarcely passable companions themselves, it would be able to
discuss the latest teen idol, aaaaaand... it wouldn't even be all that
difficult to code.  I think this would be a fun little project, one on which
I could use the help of one of you hipsters.  What is an example of a teen
chat site?  Is there a site that can be downloaded as one big revolting glob
of teen text?  Has anyone here ever tried to download such a thing?  If you
can, and will, do send it, and I will try to write up some code.

More difficult question now: if I can code a teen chatbot, is it ethical?
Morally OK to do that?  What bad consequences can be imagined?  What good
consequences could be imagined?

Since teens are emotionally delicate, or in most cases already have enough
problems without finding they have just been having an hour long
conversation with a spreadsheet, are there counter suggestions?  Would it be
morally OK to go to a religion chat site and pull the same gag?  How about a
motorcycle group?

Regarding this last question about a motorcycle group, I belong to a group
in which new people come along regularly and ask questions that have been
asked a jillion times before.  I think I could write code that would
recognize questions identical to a previous one from the archives and answer
it.  This would be perfectly OK, and would be very open about being
software.  For instance, we regularly get stuff like "Does Fram make an oil
filter that would fit the cavalcade?  Joe Newbie"  The software would get
the guy's name and answer immediately "Ja, Joe Newbie, use a PH4023, and if
you can't find that, you can try a Delco 1660 which is a buck cheaper and
works fine.  Answered by SpikeBot."

That isn't trying to fool anyone, but helps the silly prole who asked the
question.

spike