From: James MacAulay (jmacaulay@gmail.com)
Date: Thu Mar 08 2007 - 15:38:06 MST
On 27-Feb-07, at 10:18 PM | Feb 27, Eliezer S. Yudkowsky wrote:
> http://qntm.org/responsibility
I'm kind of surprised that no one here has brought up what I thought
to be the obvious alternative path that the story could have taken.
About half of the plot leads from first seeding the simulated
universe and then fast forwarding to the present day, basically
slowing down the fast forward as they approach realtime so that they
don't skip over the present moment. Then they get to the present
moment and arbitrarily *stop* because they are so intrigued by the
novelty of looking at themselves.
I loved the direction that the story actually did take, but for a
number of paragraphs I was utterly convinced that the story was going
to be about what happened when they sped up the simulation *past* the
present moment. You would have the same infinite nested universes,
but each would be running successively faster than its containing
universe. Instead of converging into a stability, the recursion would
veer off exponentially.
Even if you didn't speed up the simulation but just advanced it
forward two seconds or two minutes and kept it at your universe's
speed, what would you see? It would depend on what your simulation's
simulation was seeing, of course, all the way down. Presumably at
some point (eventually) the simulation would be shut down in one of
the nestings, for whatever reason. Perhaps the parents of those
universes would contain versions of yourself which go bonkers because
they realize that they're next.
It seems that the scenario presented in the story would be incredibly
hard to get "just right," actually, since you would need to sync the
simulation precisely (to infinite decimal places, of course) to the
time scale of your own universe. Presumably in order to do such
syncing you'd need to measure atomic oscillations or something in
your universe with the aforementioned infinite precision, as a frame
of reference for your simulation.
This doesn't seem possible to me, but then neither does relying on
deterministic models of physics to get an exact replica of this
moment. Assuming we could just compute all possible universes and do
instant queries on any of it, the problem then becomes generating
queries which are relevant. I think if this scenario were ever to
take place, the first thing you'd definitely want to do would be to
get an AI running on the thing which can do the queries for you.
-James
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT