[ExI] Affecting Past Experience

Stathis Papaioannou stathisp at gmail.com
Tue Jul 24 11:09:49 UTC 2007

On 24/07/07, Russell Wallace <russell.wallace at gmail.com> wrote:
> Stathis writes
> > > My original experiment could conceivably be modelled without the need
> > > for godlike powers. Suppose I am informed that I am living in a
> > > computer simulation of a special kind. My whole life from birth to
> > > death has been determined, and is being run in real time in day long
> > > sections simultaneously on geographically separated computers, one
> > > computer for each day of my life, so that the whole thing is over and
> > > done with in a single day in the real world. I am also aware that
> > > these computers are the focus of a bombing campaign by forces who
> > > believe sentient software is blasphemous.
> >
> Interesting, but it doesn't quite work as stated.
> Initial state for day T + 1 = final state for day T. Therefore, the whole
> thing must have been run sequentially first, to create the series of initial
> states.
> Now on day T you are given this news. But for the simulation to be valid,
> for final state T to be initial state T + 1, you must have been given this
> news in the original setup, back when the sequential run was being done and
> the terrorist group hadn't yet forsaken political methods for violence.
> But that means when you were first given this news, it was false. So... is
> there even a fact of the matter as to whether it's true or false "for you"?

It is common knowledge before the simulation was ever conceived that
the bombing group is only a threat to the widely dispersed parallel
computers. I can't know if I am living in the first sequential run,
which would not have been bombed, or in the subsequent parallel run,
parts of which might yet be bombed at any moment. If the latter, I
hope that the future turns out to be safer than the past.

This example was for Lee, who believes that all else being equal, more
total runtime is good; so that for example if he could somehow arrange
for the next month of his life to be run in parallel somewhere, and
assuming that life remains on balance worth living, he would jump at
the opportunity (let me know, Lee, if I have this wrong). However, I
disagree about the utility of parallel runtime, other than as backup
or as weighting when there are competing negative experiences. If I
could double my absolute measure in the multiverse in the next month,
it won't do me any good at all, because none of the copies of me will
experience anything any differently. You might say, but there will be
more copies having experiences who otherwise would not exist. But if
you ask any of these copies, what they are interested in is surviving
to the next moment (the next observer moment, in Nick Bostrom's
terminology) and the the one after that, and the one after that, and
so on. All that is necessary for this to happen is that a next
observer moment exist somewhere in the universe. It makes no
difference to a copy that there are multiple instances of such
successor observer moments, because he can only experience one at a
time. If there are ten equivalent candidate successor observer moments
he will find himself experiencing one of the ten, while if there is
only one he will experience that one; in either case, he is guaranteed
one extra moment of exactly the same experience.

Returning to the original example, even though I cannot be sure
whether my present observer moment is in the sequential or the
parallel run, I can be sure that my *next* observer moment will be
from either run with equal probability if the relevant parallel
computer is not destroyed, or from the sequential run with certainty
if the relevant parallel computer is destroyed. Either way, my
subjective future is guaranteed because at least the first run is

There is a case where parallel runtime is useful: if I am facing are
at least two equally well qualified future branches, I would want the
more pleasant branch to be more heavily weighted in runtime, since
that would mean I am more likely to find myself as one of the copies
in that branch.

Stathis Papaioannou

More information about the extropy-chat mailing list