[ExI] ai class at stanford

[Adrian Tymes]

On Tue, Aug 30, 2011 at 11:11 PM, spike <spike66 at att.net> wrote:
> The more I think about this however, the more obvious the idea becomes.  A
> robo-FAQ is an obvious thing to need or want, and it's a cool coding
> project.  Therefore it must have been done a thousand times before now.  Has
> anyone here seen a robo-FAQ?  Is the code public domain?

I have seen attempts at it.  Most often, the reason it gets pulled is because it
keeps triggering on stuff that is similar to the newbie question but not quite -
even when the questioner takes pains to distinguish the question from what's
in the FAQ.

To take one example:

<person> "How do I install X?"
<bot> "Hello!  I see you are asking how to install X.  To install X, take disc Y
and do Z."

So far so good, right?

<person> "My computer crashed, and I'm trying to install X on my new box.
When I put the disc in and run the command I get an error message about an
invalid key."
<bot> "Hello!  I see you are asking how to install X.  To install X, take disc Y
and do Z."
<person> "I just said I took the disc and did that thing.  Arrgh!"

Actually, the X installation tool thought - mistakenly - that it was
being pirated.
But that's beside the point.

Also consider those phonemail menu trees.  The point is (usually) not actually
to make people frustrated so they go away; the trees are an effort to help
people find the information they want efficiently, with labor saving as a bonus.
The frustration factor comes from how often people call in with requests that
are not actually in the set of information available to the tree.