On Sun, Oct 26, 2014 at 4:42 PM, justcamel via AGI <[email protected]> wrote:
> On 26.10.2014 17:24, Steve Richfield wrote:
>
>> It would take a capability beyond present imagination - probably more CPUs 
>> than there are atoms in the Milky Way galaxy, beyond even anyone's 
>> conception of God, to do such a thing. Why would this even be interesting?
>
> Not really. As I wrote before ... quantum mechanics could be a processing 
> time saver. There might be no atoms in the Milky Way at all unless you take a 
> measurement. Just as regions of the map nobody is currently using are not 
> rendered/processed by the "World of Warcraft" servers.

Yes that's true. It would take 10^120 bit operations to simulate the
physics of the universe according to Lloyd (
http://arxiv.org/abs/quant-ph/0110141 ). But to you, it would be
indistinguishable from the 10^18 simulated neural spikes coming from
your simulated sensory organs over your lifetime. And even that could
be reduced further by the simulated filtering of your perceptual
neural circuitry down to 10^9 bits of long term memory.

But I have two different problems with Bostrom's simulation argument
(the original can be found at http://www.simulation-argument.com/ ).
First he is assuming that whoever is simulating us is just like our
future selves. Actually we cannot know anything about the universe
that simulates us when even our own thoughts are the result of its
computation. It can have us believe whatever it wants us to believe.
It might be that in this other universe that space, time, and matter
are abstract concepts it invented for our world. We can say nothing
about the upper limit of its computing capacity, obviously.

My second problem with Bostrom is that he misuses probability theory.
He argues that if there is one real universe and 9 simulations of it,
then there is a 90% chance that we are in one of those simulations.
That is not what "probability" means. Probability is a mathematical
model of belief, not of reality. It means that if we perform 100
trials of an experiment and we believe that those trials are
independent, and that a certain outcome is observed in 90 of those
trials, then we say that there is a 90% chance of that outcome
occurring on the next trial. We will believe that the trials are
independent if we do not know of any program shorter than 100 x H(0.9)
bits long which outputs the observed results, where H(p) is the
entropy of p = p log 1/p + (1-p) log 1/(1-p).

This is not what Bostrom is claiming. We cannot do an experiment where
we test 10 universes and find that 9 of them are simulations. If you
are a simulation, then you will believe what the simulation has
programmed you to believe. In any case, probability theory doesn't
work in cases where you have done 0 experiments. Different people will
assign different probabilities without any mathematical justification,
because no such justification is possible.


-- 
-- Matt Mahoney, [email protected]


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to