Damien R. Sullivan:
> I'm still rather bothered by the fact it's wrong at most moments in time.
> In a species that lasts 100 million years members can have been concluding
> that they were about to go extinct. For 100 million years.
You can't derive any time predictions from the DA until you have specified your estimates of future population figures. If the hypothetical species you refer to knew that they had a constant population, for example, and you ask what they would believe with (say) 97.5% confidence, then you'll find that only for the first 2.5 million years would they conclude that doom will strike before 100 million years.
> Does the logic of the DA not apply to your own lifespan?
You would have to use the observer-moment version. But I think it does not apply, since what I call the no-outsider requirement isn't satisfied. (There are other observer-moments beside those belonging to your life.) But it might be a problem for John Leslie since he doesn't accept the no-outsider requirement. A French mathematician, J-P Delahaye, noticed this and called it the Baby paradox -- newborns predicting that they will die soon.
> > is correct. For most of them, it is true that they are about in the middle
> > of the set of members of their species. That's just the law of averages.
> > So of all those species members who apply the DA to conclude that "doom
> > is imminent" (specifically, that they are about halfway along in the
> > membership of their species), most of them are correct.
> Is this the DA conclusion? I'd thought it was that room really was imminent,
> not that you're in the middle.
The conclusion is that you are probably somewhere in the middle. However, due to recent population explosion (~10% of all humans ever born are alive today) this translates into a fairly short time span - especially if we think uploading and space colonization would add to population figures.
> Consider 3 species. One has a constant number of members for almost all of
> its existence. Another has exponential growth and then sudden collapse. A
> third has a Gaussian distribution over time. It seems to be only members of
> the last can conclude they're in the middle. Members of the second should
> conclude they're near the end. Members of the first can't conclude much of
> anything; they have a 1/3 chance, say, of being in the middle third of their
> history. Not very specific.
Ah, but by "near the end" we don't mean "near the end of the time inverval during which the species will exist", but rather "near the end of the sequence of individuals in this species".
> Another thing I've never figured out is what we're supposed to do with the
> conclusion of the DA, even if correct.
I can tell you that: we should change our view about what the future will be like. More importantly, we should redouble our efforts to avoid all possible threats to the existence of our species. By doing that we change the priors and improve our odds.