From: Krekoski Ross (rosskrekoski@gmail.com)
Date: Wed Jul 23 2008 - 10:51:23 MDT
On Wed, Jul 23, 2008 at 6:59 PM, B Ziomek <bziomek@gmail.com> wrote:
>
> Physic humans with minimal upgrades or direct uploads: I'd say that if a
> singularity of any kind does occur there will be some intelligences left
> which are regonizable in cognizance. Whether existing in a sandbox
> simulating the current world or fleeing the inner system, who knows, but I
> find it extremely unlikely that any AI wouldn't want to keep at least a few
> of its creators around, at least in a stored form for study.
fair enough.
> I may be making
> a dangerous assumption here, but it seems unlikely that a logical being
> would destroy something that may come in useful in the future in exchange
> for a comparatively miniscule amount of storage space (heck at worst our
> genome is what, 5 MB? at least the plans to make a human would be seemingly
> guaranteed to survive and singularity).
True, but again, 5MB of complexity (actually less) is miniscule
compared to the amount being generated daily by any individual.
Granted if we were to abstract away the generated complexity of all of
humanity we may get some redundancy (I dont know, it depends on what
philosophical framework one draws upon) but certainly it should be
more than 5MB.
>
> But beyond this, how much will our true successors resemble today's
> humanity? I think it's impossible to make a concrete guess, but that the
> lower boundary of similarity is pretty far away.Uploaded humans will just be
> emulations of running on vastly superior hardware. As anyone who's ever run
> a hardware emulator on a modern computer knows, the amount of processing
> power it takes is usually far larger than the processing power the hardware
> being emulated had, and hence the process is very inefficient.
I argued a similar point a few months ago-- If humans were to be
emulated perfectly, the hardware running the emulation would be
necessarily more complex (Kolmogorov) than the emulated reality, down
to the quantum level if any quantum effects have any effect on random
input available to individual humans (chaos theory applies, I'm not
talking about quantum effects in the brain, but the non-predictability
of, for example, the microdynamics of thermal systems). It would be
thus more efficient to observe, rather than emulate the system. A
vastly superior intelligence would certainly be aware of this fact,
and if it is at all interested in self-improvement (defined presumably
as an increase in complexity and cognitive capability since this type
of goal would most probably fall near to the target area the initial
designers aimed for) then it may very well wish to preserve humanity
(as broadly construed by this list) in some recognizable form simply
because it allows for a diversity of direction. I'm talking here about
direction in terms of the likely evolutionary paths of human systems
of thought that a preserved humankind would take as distinct from ones
an isolated AI would take-- it is an advantage to have a diversity of
intelligent systems if self-improvement is an end goal.
Its somewhat analogous to the argument that biodiversity is a
scientific (and thus ethical, as well as economic) end to itself. The
only reason we harvest and exploit the biosphere to the extent we do
now is because we're tremendously short-sighted.
At the very
> least the uploaded humans would modify themselves to execute more natively
> on whatever hardware exists, and once the self-tampering starts, when will
> it stop?
>
Yes self modification is fine. But we don't really know how that'll
turn out yet. Kinda like chimps trying to predict the stock market.
Ross
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT