Re: Leslie's "The End of the World" (was Re: Does anybody know t

Nick Bostrom (
Thu, 13 Aug 1998 22:02:22 +0000

Mark wrote:

> Nick Bostrom [] wrote:
> >But there is a version of the doomsday argument that bites on this
> >possibility too. Rather than counting the number of observers, you
> >should (arguably) count the number of "observer-moments" - i.e. time-
> >segments of observers. In that case you get little respite from
> >assuming that birth rates will drop drastically.
> But why should age matter?

I'm not clear that it does (there are different views on that). But consider the following thought experiment. Suppose that for one year, every morning you will suffer for amnesia for 10 minutes. Now suppose you find yourself in that situation: it's morning and you suffer from amnesia. You are asked to guess what day it is. You say you have no clue. You are then required to bet on whether it's Monday, and also on whether it's Tuesday, and wether it's Wednesday etc. Seven different bets. If I were you I would assign a probability of 1/7 to each of these hypotheses.

Then you are asked to bet on whether it's *not* Sunday. It follows you should assign a 6/7 chance to that.

But why? Well, it seems that what you are doing is saying to yourself: "There are these two intervals, Sunday and not-Sunday. But not-Sunday is six times as long as Sunday, so that should be six times as probable."

Once you have started to assign probabilities to time segments like this, it is easy to see how to generalize it to the context of the DA.

(Thanks for asking the question. I'm actually a little clearer about this after having to think out this answer.)

> If many humans today are going to live for
> billions of years, then most of the older humans must at some point have
> been a human living today.

That does not follow. Most old people that will ever have existed might be people who will not be born for many millenia yet. This does not preclude that today's humans could live for billions of years.

>That means that any
> number of SIs can be created from a limited number of organic humans, and
> all will remember being an organic human at some point. In order to get
> to be an SI they must have started out as a human, but any number of SIs
> can be created from one human.

I see what you mean but I don't think it helps.

> If the Singularitarians are correct that only
> one human will transcend to an upload and take over the universe, and
> if the time argument is correct, then I should be far more likely to be inhabiting a long-lived body than a short-lived one and hence the odds are
> high that the human in question will be me, since I'm currently inhabiting
> this body.

And under these assumptions you should also be more likely to inhabit a later stage of the body of the person who will transcend. Given that you inhabit an early body-segment, you are no more likely than anybody else to become the Superbeing. (And this might even be evidence against the hypothesis that there will be such a Superbeing.)

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