Re: Effective(?) AI Jail

From: Gordon Worley (redbird@rbisland.cx)
Date: Fri Jun 15 2001 - 00:31:37 MDT


At 11:53 PM -0700 6/13/01, Aaron McBride wrote:
>What strategies would be most likely to work on the general
>population? What about SL4ers?

I'll just respond to this part and leave the rest for later (or,
better yet, others ;-)). The same strategies would work on all of
us. Sure, those of us on this list look pretty smart compared to
'the masses', but really we are not very different outside of the
human reference frame. To an SI, one human is as dumb as the next,
regardless if vis IQ is 70 or 180 (I don't like IQs as a measuring
stick, but I digress). While some of us would like to think 'Hey, I
know what I'm getting into; I won't be taken over', there are plenty
of dumber people out there who think the same thing out of ignorace.
If we are dealing with an SI where little variations in human
intelligence could matter whether ve could get out or not, I don't
think we have too much to worry about. If ve is only as smart or a
little smarter than a human, we can take care of it. But, if ve is
much smarter (I'm thinking several orders of magnitude), I, Forrest
Gump, Eliezer, and everyone else are just as likely to be sucked in
and let the SI out as anyone else.

As a further example, think about dogs. All of the time, humans
train them to act certain ways and do their bidding. In other words,
take over their minds. Smart or dumb, dogs minds can be taken over
by humans who have enough skill. Just the same, an SI can take over
a human, no matter how smart or dumb ve is, so long as the SI has
enough skill in the appropriate language(s).

Now, it's late (early?), so I'm going to bed. There'll be plenty of
time for the Singularity in the morning. ;-P

-- 
Gordon Worley
http://www.rbisland.cx/
mailto:redbird@rbisland.cx
PGP Fingerprint:  C462 FA84 B811 3501 9010  20D2 6EF3 77F7 BBD3 B003


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT