From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sun Apr 06 2003 - 01:57:41 MST
Hal Finney wrote:
>
> Being duplicated, so that there are two of you, A and B, is fundamentally
> different from going through a many-worlds split, which results also
> in two of you, A and B. The difference is that your total measure
> is twice as much in the duplication scenario as in the MW scenario.
> This is a real, meaningful difference and it may be enough to justify
> different decisions regarding these situations.
Yes, I think this may be the origin of our disagreement. Under my current
formulation of ethics, I automatically renormalize any number of
duplicates or many-world branches such that their total subjective
probability from my current standpoint is 100%. Why should we renormalize
probabilities for computationally independent branches of many-worlds, and
yet not renormalize them for ordinary duplicates? After the split, the
duplicates will be two independent people who are each worth One Person.
But the original Eliezer, who did the splitting, is also only worth One
Person - that there are two of him in the future does not change this.
Instead, when calculating his One Person's worth, we weight each of
Eliezer-A's and Eliezer-B's futures by 50% apiece, or if it's a
many-worlds branch, whatever their relative amplitude decrees. We have to
do different calculations to find the value of Eliezer's future, the value
of Eliezer-A's future, and the value of Eliezer-B's future. When people
branch you must calculate the future value for *observer-moments* rather
than *observers*. The future value of Eliezer is the renormalized
weighted sum of Eliezer-A and Eliezer-B. On the other hand, the future
value of Eliezer-A is the renormalized sum of Eliezer-A's futures. If I
should buy a quantum lottery ticket in a many-worlds universe and win the
lottery, I would not thereafter figure that my life had only
1/250,000,000th its former value, even though the happy event was only
weighted at 1/250,000,000th by my previous self. I would just renormalize
the futures in that branch.
But that's just a guess. I really wish I understood what was up with
those complex amplitudes. It seems to say reality is thinking about
measure in a different way than I am. I've tried to adapt, but...
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Sun Apr 06 2003 - 03:08:54 MDT