On Thu, Jan 29, 2004 at 11:33:15AM -0800, Hal Finney wrote:
> What about arguments that attempt to estimate the fraction of observers
> who are in simulations versus in base realities, such as Nick Bostrom's
> Simulation Argument, www.simulation-argument.com?
> Are you saying that such arguments are pointless, and that no matter
> how convincing they became, both choices would be equally rational?
These arguments are not pointless, but they only make sense if you assume
a particular measure or class of measures to be used. An argument that
counts the number of copies of observers only make sense if you assume
that you should care about each copy equally regardless of whether it
exists in a simulation or in base reality.
> Or would you say that it is rational to reject observations? After all,
> among the infinity of universal distributions there are enough to justify
> rejecting any specific observation as a "flying rabbit", a special case
> exception which is built into the UTM that defines the distribution.
> There exist universal distributions which can accommodate any such
I'm not sure what you mean by 'justify rejecting any specific observation
as a "flying rabbit"'.
> Doesn't this philosophy ultimately reject all evidence, and further,
> make it impossible to make predictions?
No, it just means that you have assume a measure when making
predictions, and that arguments about implications of evidence can only
usefully occur between people who use similar measures.
> There is a universal measure
> which is consistent with my past observations and yet lets me conclude
> that the sun won't rise tomorrow. Is it just a matter of taste and not
> rationality that determines my beliefs on this matter?
Yes, theoretically there exists a universe where the sun won't rise
tomorrow, and you could assign it a larger measure than the universe where
the sun does rise. You could also prefer the taste of dirt to the taste
of ice cream. If we don't call the latter irrational, why should we call
the former that?
> If my understanding of these questions is correct, we have to find
> a stronger set of rules and constraints on rationality, for the term
> to have a useful meaning. Maybe we don't have them yet, but it isn't
> acceptable to call such a wide range of behaviors rational.
I'm not sure how to respond to that. Can you explain why it isn't
acceptable to call these behaviors rational?