Note that Bostrom doesn’t advocate preemptive nuclear strikes in this essay. Rather he says the level of force should be no greater than necessary to “reduce the threat to an acceptable level.”
When the stakes are “astronomical” and many longtermists are maximizing EV (or using Maxipok) and are consequentialists (or sufficiently consequentialist), what’s an acceptable level of threat? For them, isn’t the only acceptable level of threat the lowest possible level of threat?
Unless the probability difference is extremely small, won’t it come down to whether it increases or decreases risk in expectation, and those who would be killed can effectively be ignored since they won’t make a large enough difference to change the decision?
EDIT: Ah, preemptive strikes still might not be the best uses of limited resources if they could be used another way.
EDIT2: The US already has a bunch of nukes that aren’t being used for anything else, though.
There are going to be prudential questions of governance, collateral damages, harms to norms, and similar issues which swamp very small direct differences in risk probability even if one is fixated on the very long run. Hence, an acceptable level of risk is one which is low enough that it seems equal or smaller than these other issues.
Note that Bostrom doesn’t advocate preemptive nuclear strikes in this essay. Rather he says the level of force should be no greater than necessary to “reduce the threat to an acceptable level.”
When the stakes are “astronomical” and many longtermists are maximizing EV (or using Maxipok) and are consequentialists (or sufficiently consequentialist), what’s an acceptable level of threat? For them, isn’t the only acceptable level of threat the lowest possible level of threat?
Unless the probability difference is extremely small, won’t it come down to whether it increases or decreases risk in expectation, and those who would be killed can effectively be ignored since they won’t make a large enough difference to change the decision?
EDIT: Ah, preemptive strikes still might not be the best uses of limited resources if they could be used another way.
EDIT2: The US already has a bunch of nukes that aren’t being used for anything else, though.
There are going to be prudential questions of governance, collateral damages, harms to norms, and similar issues which swamp very small direct differences in risk probability even if one is fixated on the very long run. Hence, an acceptable level of risk is one which is low enough that it seems equal or smaller than these other issues.