From: Josh Cryer (joshcryer@gmail.com)
Date: Thu Jul 01 2004 - 02:43:10 MDT
On Wed, 30 Jun 2004 20:43:25 -0700, justin corwin <outlawpoet@gmail.com> wrote:
>
> On the simulation of opaque undetectable sandbox worlds:
>
> On Tue, 29 Jun 2004 19:20:12 -0700, Josh Cryer <joshcryer@gmail.com> wrote:
> > Long time reader (off again on again), just wanted to post about this.
> > I do not think a transhuman superintelligent mind could necessarily
> > *know* it was inside a box. You invent a universe for it, you plop it
> > in there, volia.
>
> I'm going to go ahead and say the creation and running of an
> undetectable virtual world with a coherent apparent casual history is
> probably beyond us. On first blush it sounds like a much harder task
> than an AI, actually.
I don't think there's some magical rule in the universe that says that
humans can't create such a thing, just for the record. I do not
believe that there exist useful things that an AI could understand
which a sufficiently learned human could not. Sure, an AI could know
five quadrillion digits of PI, that does not necessarily mean that
that is remotely useful to humans. If anything, an AI could make
things simplier for humans to understand, technological simplification
could be the actual results rather than the opposite.
But we don't need to simulate the whole universe, just what areas
we're allowing the AI to explore or experience. Remember, we want
useful stuff from the AI, in our universe it (the AI) could be in a
"box" somewhere in a lab in some company which is attempting to create
AI. (Whoa, try to twist your brain around that, a AIBox as an AI in a
box in a virtual universe!) It'd only know the books we've given it to
read, newspapers, and so on. At this point the simulator would need
only be in the box that it really is on, we don't really put it into a
virtual universe until it designs technology for us that allows us to
create the box in the first place. We can essentially use the AI to be
one step ahead of itself. Kind of sneaky, but workable.
Of course, you might argue that stopping the simulation to work on
anything the AI invents for us would clue it in to time passage (say
we turn the machine off to build nanotechnology it designed; which
results in the stars changing their alignment, etc, it would see an
instant change in the alignment of many many events it can observe),
but this is sufficiently quelled by recording all data inputs that the
AI would recieve and plugging them into the resulting virtual universe
that we have created. Basically, the start of the virtual universe
will be the end of the real universe. It could be totally unnoticable
to an AI. An AI can't know what doesn't exist. We're essentially
talking about an AI trying to get out of a box that has no evidence
for its existance.
Josh Cryer
This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:22:42 MST