| |
|
I thought I'd write a few more thoughts on this.
OK. Lets grant substrate independence and assumptions about computing power and the possibility of AI and ancestor simulations. But having granted them and in some sense forgotten them, lets remember that we have agreed to forget them.
So the first part of Bostrom's argument says (more or less) that either lots of ancestor simulations get run, because if you can run one you can run lots, or hardly any get run. Fair enough.
He then goes on to state a "bland indifference principle" that says, assuming lots of these simulations get run, we should believe that we are simulations.
(Actually, Bostrom is only trying to justify a disjunction in the first instance, which I think is more or less ok. But clearly, the fuss is about arguing that we should believe we are sims, given that we are not going to die out.)
Now here is where I think there is a trivial sounding problem that actually puts the argument in the same class as brain in a vat type scenarios. The thing is, the indifference principle is only true if you assume that we have no reliable information about whether we are sims or not, as Bostrom notes.
But in order to argue that we have no knowledge about whether we are sims or not, our experiences must be, to a certain extent, deceptive. However, if our experiences are deceptive, there is no good reason to assume that they are only deceptive, or potentially deceptive, about this one thing. As a for instance, we could return to the assumptions on computing power we had at the start and posit some alternative scenarios. Perhaps AI is possible but extremely difficult both in terms of time and resources, so that very few ancestor simulations could ever be run. What credence should we attach to this possibility? Well, if we accept Bostrom's arguments at the start, not much. But if we accept that we are likely to be simulations, and our experiences are manufactured, then how can we suppose that the simulation is a faithful representation of the physics in the "real world"? I think we would be forced to admit that our doubt at our being "genuine" would have to be coupled with doubt about physical constraints on computer power. (You might try to argue about the probable nature of an ancestor simulation, but any basis of plausibility simply begs the question.)
Ironically, I think the argument is only convincing if you believe it isn't valid.
The weak link for me is the claim that our experiences give us no evidence which lets us decide if we are sims or not. The problem is that we can never have any idea about this sort of question and that the only sensible response to this kind of skepticism (and I think this is where Bostrom is envatting brains) is to act as if it isn't true, even though the possibility exists that it is true and we are being routinely deceived.
Actually, I think Bostrom is a step away from realising this. He argues that as the proportion of sims approaches 1, so should our credence that we are sims approach 1, because at the number 1, we would be sure to be a sim. Except that, at the number 1, we have the situation where nothing came before posthumans except simulations. But this is clearly nonsense. As we approach that number 1, we should be increasingly convinced that Bostrom's argument is invalid.
(This is more or less the point Weatherson develops, I think, though he uses a lot more symbolic manipulation.) |
|
|
