From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sun Jun 23 2002 - 16:08:06 MDT
Ben Goertzel wrote:
>
> Ok, suppose that an expedition sent to a nearby star gives the AI info
> about rare elements, and suppose this info is the key needed to create
> faster-than-light travel, which then enables the AI to travel to the
> galactic core, which then allows it to travel in time...
Exactly, Ben. What *looks* like a statement "about as plausible as any
other", time travel in 2123, turns out to require an incredibly improbable
concatenation of events to come into being. So why did you say 2123 rather
than 2023+15sec or 200023 (to allow for a round-trip to a distant stellar
object?) Because you're used to science fiction stories that postulate time
travel coming into being around a hundred years in the future.
WRONG PRIORS.
> Yeah, it seems an unlikely series of events. But how good can we
> possibly be at estimating probabilities about post-singularity events?
> Obstacles may arise which we simply cannot forese now.
Look, when you say 2123, I am perfectly justified in assuming that the
foundations of this statement lie in science-fictional convention about
"when time machines get invented", rather than the incredibly improbable,
overcomplex description needed to actually place time travel on the order of
100 years in the future (much too long to be a theoretical problem that can
be solved within the solar system, much too short to cover any interesting
spatial differences). If you'd been thinkin' about it you woulda said
either 2023 or 200023.
-- Eliezer S. Yudkowsky http://intelligence.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT