From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Fri Sep 28 2001 - 16:06:29 MDT
aominux wrote:
>
> No matter how disgustingly sick this sounds, after spending a certain amount
> of time in a hedonistic environment you WILL begin to desire pain. It
> doesn't matter whether or not you can conceive this now.
>
> I doubt any of us could go million subjective years before desiring a
> pre-singularity simulation, let alone an infinity. The sick truth is that
> we don't desire happiness; we desire sensory stimulation.
>
> Besides, since you could consume an entire subjective human life in a
> fraction of a second in post-singularity times, it wouldn't seem like much
> of a sacrifice. We will/do desire variety and we will get it.
It seems to me that the dichotomy between pleasure and pain is not a very
complex one. An alternating sequence of pleasure and pain does not
display much more complexity than a pure tone of either type. What you
want - assuming that you want to keep the basic human reinforcement
architecture in the first place - is continuing emotional involvement in a
scenario of ever-growing complexity, where the subjective intensity levels
vary quantitatively (at least) in a way linked to the complex decisions
and complex sensory information returned. Pain is not necessary for this,
although pleasure does have to be complex and involved rather than simple
and dissociated.
Of course, in saying this I am violating the post-Singularity
unknowability constraint. The point I want to establish is that I can
think of at least one way to, starting from a basically human frame of
reference, construct a scenario for non-bored transhumans. I'm not saying
that this is THE scenario, but I'm saying that it's possible to imagine a
scenario where there are better things to do than carve table legs for a
thousand years a la _Permutation City_. To my mind this has always been
the single largest flaw in Greg Egan's works; in both _Permutation City_
and _Diaspora_, he doesn't seem to have been able to think of anything
interesting to do with a few thousand subjective years. I don't think
you'd get bored even with the strictly human possibilities in a few
thousand years.
> On a related note, the odds that we are not in a simulation are impossible
> to compute since we would have to assume the that we were in a simulation.
> However, if we trust that our world is based on a backup of the future's
> past, it doesn't take much time to figure the probability of a single human
> life being in the "real world." Humans have been around for far under one
> million years and we still have billions of years until our sun expires.
> Many many many more humans will exist beyond this point in history than
> before it. You chances of experiencing pre-singularity firsthand are slim
> to none.
Heh. Fall off a cliff and... "You thought you were one of the Eldest?
Hah, not only that, you thought you worked for the Singularity Institute?
Yeah, right! Just who do you think you are, anyway?"
Note two things, however:
1) It doesn't solve the Fermi Paradox, since the ancestral simulation is
a simulation of an old galaxy that has not been colonized by aliens,
rather than being a simulation of a very young galaxy that was the first
to give rise to intelligent life - leaving us with the same Fermi Paradox
for the ancestral world being simulated;
2) My own subjective experiences appear to violate the Friendly AI rules
that I would expect to be the outcome of the current scenario, if it is
historically accurate. I mean, I can see complete failure of Friendliness
as a possibility, but find it hard to believe a partial failure that
allows nonconsensual ancestral simulations;
3) There will be many many many more citizens in the future than there
are now, but not necessarily many many many more humans; the fraction that
choose to hang around the human intelligence level may very well approach
zero. If there are simply Better Things To Do, only the fact that there
is an existing subpopulation of Luddites heading into the Singularity
makes it at all possible that there may be an unbroken line of Pedestrians
descended from Pedestrians ten million years into the future. And would
those long-selected-for Pedestrians be running immersive delusive
ancestral simulations of SL4 list members? How long would a Pedestrian
stay Pedestrian who engaged in that kind of hobby?
However, the ancestral simulation argument is probably the only rational
reason to even speculate whether the life we see and inconveniences we
undergo have some kind of value or coherence which is not immediately
evident.
> Sincerely apologizing for grammatical errors and sl4 violations,
The main one is quoting the whole post you replied to. Would everyone
please just take the time to delete it?
-- -- -- -- --
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 - 04:00:37 MDT