RE: When Programs Benefit

From: Lee Corbin (lcorbin@tsoft.com)
Date: Wed Jun 12 2002 - 13:52:27 MDT


Hal writes

> A minor variant on Lee's experiment is to have the *first* simulation
> halt if you push the button, but not the second simulation (he had it
> vice versa)....

and concludes after some excellent analysis

> So the question is meaningless. In general, you can't speak of a
> resumption as being associated with a specific earlier run of a program.

Exactly so.

> To me, this casts doubt on the intuitive sense that there is "one person"
> associated with the first run, and "another person" associated with the
> second run, with the first person getting snuffed out when he pushes
> the button while the second person continues to run. That model just
> doesn't work.

Quite right. We have *one* person. That person cannot be
distinguished into two at all, even if the subject attempts
use of the tricky pointer "*this", by the subject, to attempt
to refer to only the currently executing process by appeal
to some sort of system or universal clock.

What we *do* have two of is *instantiations*. One executes at
spacetime coordinates X and the other at coordinates Y. (This
is from the universal or objective point of view.) Such
identical executions can easily be caused to happen outside
each other's light cone so that no direct causal relationship
exists between them. And I say that just as no causal relationship
exists between them, your approval and disapproval shouldn't vary
either.

> Instead, there is at best ambiguity about which person it is who
> continues to run. [There's not even that.] (In fact, John Clark
> goes so far as to deny that there are two people involved, and I
> agree that this is questionable.) [More than questionable.]

Indeed I hope that we all do come to agree that only
one person executes here, (which is an implication of
the broader principle that duplicates are self).

> We can even look at the cycle-to-cycle progress of the programs even when
> no macroscopic suspension is involved, and question whether it makes
> sense to associate one person with one run and a second person with
> the other run. Between each cycle there is a microscopic suspension,
> which based on the argument above means that we could with equal validity
> assume that the two runs swap identities every cycle. That's a pretty
> absurd view of the behavior of identities (by which I mean people from
> the first-person view), again calling into question the validity and
> usefulness of the one-person-per-run model.

Yes. The whole question, however, remains whether this
one person benefits from redundant beneficial runs (or
is harmed by a redundant cruel and horrible runs).

Of course, I have been claiming that this single person
actually suffers whenever he has a bad run. I claim that
he actually experiences experiences each time that he is
run. (To me, my claims appear practically tautologous.)

How people can value differently a local execution of a
program depending on events arbitrarily far away in space
and time is quite a mystery.

Your emotions do not deceive you in this case when your
knowledge of actually ongoing terrible events elicits
your horror and your attempts to prevent them. Any of
the correspondents in this thread would simply *not*
allow abstract considerations to prevent him from acting
against local horror. If you could truly understand
the torment that a program was experiencing every time
that you ran it, you simply would never run it. Facile
arguments such as "this run doesn't matter because it's
been proved that everything happens over and over" don't
fly morally, ethically, or logically.

This, alas, is the 8th posting today, so that's all from
me for now.

Lee



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:14:45 MST