Re: AI Jailer.

From: James Higgins (jameshiggins@earthlink.net)
Date: Tue Jul 16 2002 - 11:05:21 MDT


Moshe Looks wrote:
> James Higgins wrote:
>
> The trick is not to have the AI justify a new design, but the original
> design
> that was implemented by human programmers. If the current (possible
> redesigned)
> AI is Friendly, it is only by virtue of the original design. The
> "fool-the-humans-strategy" that you suggested would not work here, since
> AI cannot alter the the original design, only comment on it.

But if it is truly that much more intelligent it could come up with ways
to prove that a flawed friendly design was in fact friendly. This
should not be a hard problem for a transhuman. It might be helpful by
giving us some insight, but we couldn't actually believe anything
written by the AI. I'd like to point out that Humans can write papers
attempting to prove points that are not in fact true, yet sometimes they
are believed for many years because there is no substantial opposing
evidence. Science is full of this, it is routine to have numerous
theories about a subject that oppose each other, yet all of them can
make a good case. In the case of a transhuman it would be much harder
for us to spot any holes so it could essentially tell us a story and,
unless we had decades to verify it, we'd have to take it on faith.

James Higgins



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:40 MDT