Re: Why would AI want to be friendly?

From: Samantha Atkins (samantha@objectent.com)
Date: Wed Sep 27 2000 - 01:46:13 MDT


"J. R. Molloy" wrote:
>
> Eugene Leitl writes,
>
> > The only realistic way how to get at that initial set of data is
> > either growing it with an evolutionary algorithm (here the exact
> > nature of the computational engine is less crucial) or scanning a
> > critter. In latter case you'll have to write a very very precise yet
> > high-performance neuronal emulation engine.
>

Or you could start with humans and continuously augment (voluntarily)
with first external but more and more integrated and then internal
hardware and software. This seems to me the best way to keep humans in
the loop and to end up with something human compatible and reasonably
likely to be friendly and caring about humanity.

And this is inline with the main track of human augmenting computer
work.

> Exactly right, and as a consequence, no one is going to write that neuronal
> emulation engine when the alternative (growing an initial set of data with
> evolutionary algorithms and genetic programming) obviously surpasses manually
> coding an AI.
>

Starting with that net and concept formation skills and building up that
and surpassing it should also work.
 
> > I'm not sure why many apparently smart people still want to do it the
> > Cyc way. Codify everything explicitly, using an army of human
> > programmers, until the thing can limp off on its own. I just don't see
> > it work, because even groups of people are just not smart enough for
> > that.
>
> Precisely so... What the Cyc way ends up with is a vast (but manageable) data
> base that does what its namesake implies: It renders encyclopedic knowledge.
> Like the human caricature of a "walking encyclopedia" it is full of facts and
> information, but it can't do any real or original thinking. If Cyc passes the
> Turing test, it does so as a result of the incompetence of human interviewers,
> not because of Cyc's artificial thought processes. IOW, just because Cyc has all
> the answers, that doesn't mean it can formulate an intelligent question.
>

But what a great resources to hook into the WebMind or to have at the
disposal of more capable AI! Don't throw that work out. It is a useful
piece. If nothing else it is a huge glob of training material.

- samantha



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:31:14 MST