Jonathan Colvin writes:

[quoting Stathis Papaioannou]:
>The situation is different when I am considering my copies in
>the future. If I know that tomorrow I will split into two
>copies, one of whom will be tortured, I am worried, because
>that means there is 1/2 chance that I will "become" the
>torture victim. When tomorrow comes and I am not the torture
>victim, I am relieved, because now I can feel sorry for my
>suffering copy as I might feel sorry for a stranger. You could
>argue that there is an inconsistency here: today I identify
>with the tortured copy, tomorrow I don't. But whether it is
>inconsistent or irrational is beside the point:
>this is how our minds actually work. Every amputee who
>experiences phantom limb pain is aware that they are being
>"irrational" because there is no limb there in reality, but
>knowing this does not make the pain go away.

This is incorrect, I think. At time A, pre-split, there is a 100% chance
that you will *become* the torture victim. The torture victim must have once
been you, and thus you must become the torture victim with probability 1.
There's no inconsistency here; you are quite right to be worried at time A,
because you (at time A) *will* be tortured (at time B). The inconsistency
comes with identifying (you at time A, pre-split) with (one of the you's at
time B, post-split). There can be no one-to-one correspondence.

You are right to say that I will definitely become the torture victim, insofaras there will definitely be a torture victim and he will have been me pre-split. However, what I am interested in is the *apparent* probability that I will be tortured, however that may come about. Here is another example. You are offered two choices:

(a) A coin will be flipped tomorrow. If the result is heads, you will be tortured; if tails, you will not be tortured.

(b) You will be copied 10 times tomorrow. One of the copies will be tortured, and the other 9 will not be tortured.

By your reasoning, there is a 50% chance you will be tortured in (a) and a 100% chance you will be tortured in (b), so (a) is better. But I would say the probabilities are (a) 50% and (b) 10%, so (b) is clearly the better choice.

Subjectively, there is *always* a one to one correspondence between an earlier and a later version, even though from a third person perspective the relationship may appear to be one to many, many to many, or many to one. This is in part why reasoning as if observer moments can be sampled randomly from the set of all observer moments gives the wrong answer.

--Stathis Papaioannou

REALESTATE: biggest buy/rent/share listings

Reply via email to