On Sat, Feb 8, 2020, 5:28 PM Matt Mahoney <mattmahone...@gmail.com> wrote:

> 1. I flip a fair coin. I peek at the result without showing you. It is
> heads. What is the probability that the coin came up heads?
>

Probability does not describe the state of a system. It describes the
belief by an agent that a future event will happen, on a scale from 0 to 1.
Different agents can have different beliefs based on different knowledge or
different models and both be correct.


> 2. While you sleep I plan to make 5 atom for atom identical copies of you
> in different rooms. Then I will roll a die and pick one of the six still
> sleeping copies to kill instantly. Knowing this, what is your subjective
> probability that you will wake up in the same room?
>

Probability can be accurately estimated by a large number of independent
trials. If an event happens m times out of n, 0 < m < n, then the
probability is m/n.

If I copy you every night for n nights (making 5^n copies of you), then on
average each copy would remember waking up in a different room 5/6 of the
time.

Making a copy of you and killing either the original or the copy both have
the same effect as doing nothing. If this seems counter intuitive, it's
because humans evolved to fear death, and as social animals, to grieve the
death of others. A computer does neither because it wasn't programmed to.


> 3. (Bostrom's simulation argument). There is a 1% chance that sometime in
> the next billion years that we will create a computer simulation of the
> present day world and run a million copies of it. What is the probability
> that we are living in a simulation?
>

Probability is a measure of belief in future events. But since there is no
test for living in a simulation, there are no future events to predict and
the question is meaningless.


------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Tefd74cfe5df991e0-M931e15a56e04daed64c53061
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to