From: Lee Corbin (lcorbin@tsoft.com)
Date: Mon Jun 10 2002 - 23:30:49 MDT
Hal writes
> And on further thought about the speed issue, I am convinced that,
> unbelievable as it may seem, it is better by this logic to run slower
> than faster! Running your program N times slower is essentially
> equivalent to running it repeated N times.
>
> Suppose your life goes through states ABCDE. Then you could run it,
> say, 3 times over: ABCDEABCDEABCDE. Now you could look at these runs
> as being interleaved:
>
> A B C D E
> A B C D E
> A B C D E
>
> and they collapse to:
>
> AAABBBCCCDDDEEE
To make sure that I'm following you, let me express your idea
in my own language. I'll also add some particulars of the
instantiation which may enable some criticism of your proposal.
You suppose that someone's life went through five states. (I
wonder if changing "states" to "stages" would work better,
because I think that just 5 states results in almost zero
benefit.) Anyway, the machine loads state A, then calculates
the next state, which turns out to be B. It then stores B,
and reloads A, recalculates B, and again stores B, and so on.
Since each state is calculated three times in your example,
and so might be calculated equally well in galactic size
disjoint regions of spacetime, it follows that this is
equivalent to three separate runs. I completely agree.
I would assign three times the benefit, and agree that
the way you're doing it,
AAABBBCCCDDDEEE
is equivalent to your first interleaved case.
> In the 3 times slower version you can think of the first B as being
> caused by the first A, the second B as being caused by the second A, and
> so on. So you can still see it as 3 repetitions, all being interleaved.
> It's equivalent.
Yes.
> Going back to the spatial issue, you could have 3 computers running your
> program:
>
> ABCDE
> ABCDE
> ABCDE
>
> Or you can think of this as a single 3-times bigger computer, going
> ABCDE. It's just a question of looking at it by rows or columns.
But now are there three independent causal chains, or just one?
Isn't this the same issue that you raised yesterday? I declare
that it's all in the number of separate calculations that took
place; just, analogously, NASA would be happier if you calculated
a moon trajectory three separate times rather than use a single
computer with big parts.
> The bottom line is, redundancy is redundancy. You can be N times bigger,
> N times slower, or N times repeated. Either way you have each state
> occupying N times more of spacetime than in the singular case.
Hmm. I still don't get this. That is, don't you believe the
hypothesis that a conscious experience (of say 5 minutes' duration)
is the equivalent of a TM going through a billion states or so?
If so, how can it matter whether the TM is a clanky machine on
wheels that takes a minute per state, or a slick Alpha that zooms
through the states? The conscious experience would be the same
(I claim). The report of the subject would certainly be the
same. The subject would express the same value for the experience.
(Now it is true that I regard those as only necessary to consciousness,
but not sufficient. There is the extremely puzzling case of the giant
lookup table to which we must return presently---so narrative and
testimony isn't necessarily indicative. But it does still remain
for you to explain why one causal succession of states is to be
regarded differently than another causal succession of identical
states.)
> I know I am not exactly providing air-tight arguments for these points, but
> my intuition is very strong that they are all equivalent. I can try
> to come up with better arguments later.
Sure.
> An amazing implication of this perspective is that uploading would be a
> bad thing! Putting yourself onto a computer that ran 1000 times faster
> would be like killing yourself to a 99.9% degree; likewise if the circuit
> elements were smaller so that your brain ran on a sugar cube computer.
> I wonder if someone who believes in the virtue of redundancy and who
> comes to accept the equivalence of these various ways of being redundant
> would refuse to accept uploading. It would be like having 1000 identical
> copies of yourself running and getting all but 1 of them instantly killed.
Well, I think that you've just supplied a reductio ad absurdem
that casts doubt on your fundamental conjecture. We know for
a fact that uploading is a good thing ;-)
I believe in redundancy, but don't believe in those equivalences.
And indeed for me, but perhaps not for you, it would be a very bad
thing if my runtime got cut 99.9% by having 999 duplicates halted.
Lee
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:14:44 MST