Re: PHIL/AI: Humongous Lookup Table

Gregory Houston (vertigo@triberian.com)
Tue, 11 Feb 1997 22:23:47 -0600


CurtAdams@aol.com wrote:
> However, it fails the Turing test quickly as once you say a
> sentence nobody has spoken before, it will fail. It's very easy to devise a
> sentence with a vanishing probability of having been spoken before and
> trivial to devise such a conversation. Almost any innovative analogy will
> do.

Would the table fail the test if it responded saying, "I do not
understand. Could you please rephrase your statement?" I personally do
not comprehend every obscure analogy that I hear either, and thus have
to respond in the the aforementioned fashion. Does this mean that I
would fail the Turing test?

Imagine the computer is confronted with a statement which it has no
reply for. What if it were then to say, "I do not understand the
question. If I were asking you the same question, how would you reply?"
Once it was given a response it could then store the question and the
reply in its table. Thus the next time it was confronted with this
question it would have a response. There are certainly faults here, the
computer would of course be particularly gullible, recording any kind of
nonsense some one might offer it, but it still seems to me like a step
towards learning ... no matter how simplistic. To reduce its gullibility
or to increase its apparent intelligence, the computer might run some
simple checks on the reply similar to those we use in forms
verification, e.g., you ask for someone's email address and you have cgi
or javascript check to make sure there is an ampersand in the reply.
With a little thought, I'm sure this could get a bit more useful, but if
the computer came to the conclusion that it might be being "toyed" with,
it could respond, "Are you sure that is a valid response to your
question?" If the person replied "no", the computer could scrap the
whole thing, but if the person said "yes, it is", the computer could be
programmed to always accept it as if everyone was honest, or it might
determine by some other means, fuzzy logic perhaps, if it wants to risk
the chance of keeping the statement and reply within its table for
future use.

I'm not too thoroughly familiar with this ... so bash away.

-- 
Gregory Houston
vertigo@triberian.com
816.561.1524