Re: reasoning under computational limitations

Nick Bostrom (
Thu, 1 Apr 1999 22:08:57 +0000

Wei Dai wrote:

> If computational uncertainty is covered under probability theory, maybe it
> can shed some light on the Self-Indication Axiom (SIA). Suppose you are in
> a universe with either 1 or 10 people, depending on whether the 100!-th
> digit of PI is odd or even. (There is a machine that computes the digit
> and then creates either 1 or 10 people.) Should you believe that the
> 100!-th digit of PI is more likely to be even than odd? This is similar to
> the thought experiment in your DA paper where a machine creates 1 or 10
> people depending on the outcome of a coin flip, but without the problem of
> dealing with what it means for the coin flip to be "fair".

Yes you can do this, but I don't see how it helps decide whether the SIA is true. (For other people following this thread: The Self-Indication Axiom states roughly that upon finding that you exist, you should increase your probability for hypotheses according to which many people exist. (There would be "more slots for you to be born into" if there were more persons.) It can be shown that adopting the SIA leads to the exact cancellation of the probability shift required by the DA. This would seem the main reason to adopt the SIA. However, there are big problems with the SIA which seem to make it ultimately unacceptable.)

Nick Bostrom Department of Philosophy, Logic and Scientific Method London School of Economics