Re: ESSAY: How to deter a rogue AI by using your first-mover advantage

From: Tom McCabe (
Date: Wed Aug 22 2007 - 17:45:58 MDT

In a reality simulation, the simulator can put
*whatever the heck they want* into the Internet
archives. If the RAI is being simulated by an FAI,
then it has no way of knowing what actions will cause
the FAI to shut it down. It's a simulated world, so
there's no moral obligation on the FAI to stop the RAI
from taking it over; and the FAI is free to make up
whatever messages it wants and stick them in the
Internet to see what will happen, even if it winds up
doing the exact opposite of what those messages

 - Tom

--- Kaj Sotala <> wrote:

> On 8/22/07, Robin Lee Powell
> <> wrote:
> > The problem is that you're ascribing demonic
> attributes to RAI when
> > golemic failure is *far* more likely: RAI isn't
> going to care about
> > your threats to destroy it, no matter how phrased,
> any more than it
> > cares about the fact that whoever asked it to
> calculate C won't be
> > around to recieve the answer. RAI has clearly
> undergone subgoal
> > stomp (that is, pursuing a subgoal is causing it
> to not realize that
> > it won't be able to complete its master goal,
> which is to give
> > whoever asked the answer to the calculation C).
> Nothing you say
> > will make any difference, but RAI is clearly so
> poorly designed that
> > it's not paying any attention to anything that's
> not directly in the
> > subgoal path.
> If it's destroyed, that too prevents it from
> achieving its subgoal.
> If its reasoning is so crippled that it doesn't
> realize it should take
> precautions to *protect itself * in order to achieve
> a goal (be it a
> subgoal or supergoal), I have difficulty seeing how
> it could ever be a
> threat in the first place. If you miss out on
> something that
> elementary, you're not going to figure out the
> harder bits, either.
> --
> |
> Organizations worth your time:
> | |

Looking for a deal? Find great prices on flights and hotels with Yahoo! FareChase.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:58 MDT