- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]

From: Wei Dai <weidai.domain.name.hidden>

Date: Sat, 21 Feb 1998 18:15:19 -0800

Ok, here's another probability paradox. Suppose in the quantum suicide

experiment the assistant offers the experimenter a bet. The experimenter

gives the assistant $2 before he pulls the gun, and in return the

assistant will give $3 to the experimenter after pulling the gun, if the

result is a click instead of a bang. The assistant clearly has a positive

expected return from this bet. But if the experimenter believes she will

hear a click with 100% certainty, she also has a positive expected return.

These paradoxes show that the method of computing sensory probabilities

suggested by Tegmark is flawed. I think the right method involves removing

the restriction that

sum_over_X P(X|Y) = 1

where Y is current perception and X is future perception.

This sum should be greater than 1 if there is a posibility of copying

(cloning) and less than 1 if there is a possibility of death. In the

single-round quantum suicide experiment this sum would be 1/2. In an

experiment where an experimenter is copied with probability 1/2 the sum

should be 3/2. In general the sum should be the ratio between the measure

of one's continuation and the measure of oneself.

Let me give a specific example. Consider an experiment where at time 1 a

coin a flipped and the result observed by the experimenter, and at time 2

the experimenter the duplicated if the the coin landed heads. I suggest

the experimenter should have the following beliefs at time 0:

1. I will observe heads at time 1 with probability 1/2, and tails with

probability 1/2.

2. I will observe heads at time 2 with probability 1, and tails with

probability 1/2.

At time 1, if she observed heads:

3a. I will observe heads at time 2 with probability 2, and tails with

probability 0.

At time 1, if she observed tails:

3b. I will observe heads at time 2 with probability 0, and tails with

probability 1.

Received on Sat Feb 21 1998 - 18:16:04 PST

Date: Sat, 21 Feb 1998 18:15:19 -0800

Ok, here's another probability paradox. Suppose in the quantum suicide

experiment the assistant offers the experimenter a bet. The experimenter

gives the assistant $2 before he pulls the gun, and in return the

assistant will give $3 to the experimenter after pulling the gun, if the

result is a click instead of a bang. The assistant clearly has a positive

expected return from this bet. But if the experimenter believes she will

hear a click with 100% certainty, she also has a positive expected return.

These paradoxes show that the method of computing sensory probabilities

suggested by Tegmark is flawed. I think the right method involves removing

the restriction that

sum_over_X P(X|Y) = 1

where Y is current perception and X is future perception.

This sum should be greater than 1 if there is a posibility of copying

(cloning) and less than 1 if there is a possibility of death. In the

single-round quantum suicide experiment this sum would be 1/2. In an

experiment where an experimenter is copied with probability 1/2 the sum

should be 3/2. In general the sum should be the ratio between the measure

of one's continuation and the measure of oneself.

Let me give a specific example. Consider an experiment where at time 1 a

coin a flipped and the result observed by the experimenter, and at time 2

the experimenter the duplicated if the the coin landed heads. I suggest

the experimenter should have the following beliefs at time 0:

1. I will observe heads at time 1 with probability 1/2, and tails with

probability 1/2.

2. I will observe heads at time 2 with probability 1, and tails with

probability 1/2.

At time 1, if she observed heads:

3a. I will observe heads at time 2 with probability 2, and tails with

probability 0.

At time 1, if she observed tails:

3b. I will observe heads at time 2 with probability 0, and tails with

probability 1.

Received on Sat Feb 21 1998 - 18:16:04 PST

*
This archive was generated by hypermail 2.3.0
: Fri Feb 16 2018 - 13:20:06 PST
*