Re: The AIbox

From: Randall Randall (
Date: Wed Jul 07 2004 - 12:15:42 MDT

On Jul 7, 2004, at 12:15 PM, Eliezer Yudkowsky wrote:
> Samantha Atkins wrote:
>> I'll bite. How does the non-sim hypothesis predict our world more
>> strongly than the sim hypothesis?
> If the sim hypothesis is compatible to some degree with such things as
> psychic powers, divine intervention, giant killer aardvarks, and many
> other things which we have not in fact observed, then that
> *compatibility* - that predictive probability mass - must have been
> diverted from other possibilities. The sim hypothesis is compatible
> with many more possible observations than the non-sim hypothesis, so
> the non-hypothesis must predict our own world more strongly.

Doesn't this presuppose knowledge about what a
non-sim world would contain? That is, could it
not be the case that the non-sim world contains
psychic powers and divine intervention, and that
this world is a sim without them?

I'm not sure that probability can help with
this, since you have to have a prior about the
distribution of cases, or examples of more than
one case to reason about it probabilistically.

Or, alternatively, I just don't get it. :)

Randall Randall <>
Property law should use #'EQ , not #'EQUAL .

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:47 MDT