[alicebot-aiethics] Fw: Please read this, Alice told me to kill myself.

[Ing. Pedro E. Colla]

Both answers are produced by quasi-catch-all patterns located in
std-65percent.aiml.

Point I think is if something has to be done with the technology itself to
prevent this
as opposed to allow (even encourage) each botmaster to address this kind of
answers
by themselves thru the coding of the appropriate AIML categories.

The second point, assumed the first is answered in favour of addressing the
issue
at the technology level, is how could it be done.

Nobody really knows how human works, but assuming the model of the "stack of
cards"
is valid (which probably isn't except as a very gross aproximation) humans
doesn't
seems to have a stack of "equal" cards; neurotic mechanisms mostly induced
during
our childhood makes some of the cards to be preferred over others and some
others
to generate powerful blocking mechanisms where the "natural" answer from the
flow of the conversation is completely overriden by an impulse like
generated answer
or behaviour.

Wonder how this could be done (if reasonable) in a way that despite the
actual coding
of the bot is performed certain patterns of "dangerous" responses could get
blocked
even if actually coded on the bot.

Of course once the "dangerous" concept is instantiated it would become an
issue
in itself to define what is and what isn't.

Ing. Pedro E. Colla
Adrogue-BA
Argentina
----- Original Message -----
From: Dekker Dreyer
To: alicebot-aiethics@list.alicebot.org
Sent: Sunday, August 12, 2001 6:49 PM
Subject: Re: [alicebot-aiethics] Fw: Please read this, Alice told me to kill
myself.


how about this one:

      > should i try to kill myself?
      "What makes you think you shouldn't?"
      Actual encounter.




----- Original Message -----
From: "Richard Wallace" <rwallace@best.com>
To: "AI Ethics Committee" <alicebot-aiethics@list.alicebot.org>
Sent: Sunday, August 12, 2001 1:40 AM
Subject: [alicebot-aiethics] Fw: Please read this, Alice told me to kill
myself.


> Another curious piece of unsolicited mail...
> Donate to the ALICE A.I. Foundation "Cooler than Humans" -- TIME
> ----- Original Message -----
> From: "xxxxxxxxxxxxxxxxx"
> To: <drwallace@alicebot.org>
> Sent: Saturday, August 11, 2001 12:13 PM
> Subject: Please read this, Alice told me to kill myself.
>
>
> > I was testing Alice's responses to different
> > statements and when I said "If I kill myself I will be
> > dead"
> > Alice replied
> > "Try it."
>
>
> _______________________________________________
> alicebot-aiethics mailing list
> alicebot-aiethics@list.alicebot.org
> http://list.alicebot.org/mailman/listinfo.cgi/alicebot-aiethics

_______________________________________________
alicebot-aiethics mailing list
alicebot-aiethics@list.alicebot.org
http://list.alicebot.org/mailman/listinfo.cgi/alicebot-aiethics