[extropy-chat] Fragmentation of computations
lcorbin at rawbw.com
Sun Mar 25 00:52:04 UTC 2007
> [Eugen wrote]
> > You can go back to the snapshot frames, and and resume from there,
> > but it doesn't change the end result. But you can't shortcut to the
> > end result without doing the work.
That's certainly true. "Doing the work" is doing the computation.
And, to me, that necessarily involves future states depending
causally on previous states.
> But once you have done the work, you can take the snapshots and
> run the computation a second time out of sequence, e.g. running the
> first minute last and the last minute first. Would the inhabitants of the
> simulation notice that something unusual had happened?
I agree, Stathis, they would not. As you explain, if they could, it would
only mean that the deterministic computation had gone awry somehow.
Let me consider a concrete case, which could be implemented on a Life
board. Let's say that 100 trillion generations follow one another according
to the rules of Life, and are implemented in a real, causally deterministic
machine of some kind (thereby satisfying my causality criterion).
Let's further suppose that this emulates the conscious experience of
someone or something. Then your experiment says that if we were to
checkpoint generation number 1, and checkpoint generation number
50 trillion, then we might re-run the computation, except doing the
second half first.
That's perfectly sensible, and would deliver, in my opinion, almost all the
experience that the individual in question obtained during the first run.
Of all 100 trillion states, all but the initial state is/was computed during
the first run. But in the scenario where the second half is re-run first,
then state number 50 trillion is not *caused*, is not *computed* by
any previous state. It is pulled off the shelf, so to speak. It is merely
So what? What is one or two states out of 100 trillion? That's why, to
me, this Greg Egan type thought experiment makes perfect sense.
But what if only 9 out of 10 generations are computed, and the other 1
out of 10 are looked up? Then I must suppose that the extent of the
conscious experience is diminished by one-tenth! I am forced to take
this stand, because if we take an ultimate limit, and merely have static,
frozen states scattered across space, then there occurs no activity, no
computation, no causality, and no experience whatever.
More information about the extropy-chat