re: Torture vs Specks
Specks, every time, unless we revise the original writeup to fix some of the very relevant details. So much so that if I see someone pick torture, I immediately suspect them unfit to touch ethics, especially utilitarian ethics. It should be self-evident to anyone with minimal compassion competence that suffering does not scale like that, and can't just be homogenized like that.
Measurement of suffering must take into account its impact relative to the sufferer. A negligible momentary negative is profoundly, qualitatively different than a life-destroying, arguably worse-than-never-living or worse-than-death fate. This does not free us of hard choices between more suffering for fewer minds and less suffering for more minds, but as written I doubt if the author even really appreciates this enough.
An infinity of nothings is still nothing, and an indefinitely large amount times an indefinitely small amount is still probably not particularly large. Of course it depends on the ratio of magnitudes between the large and the small - but:
If you're a human trying to make real-world decisions, and you come to me claiming that you have that level of reliable precision with utils/qualys, my first reaction is going to be "fuck off, no you don't. also, how the hell did you manage to get more hubris than me?" Both due to practical limitations of us mere humans, but also in principle, because:
You have to contend with per-mind ethics and something like the uncertainty principle - in ethics just as in physics, you can only get so precise before you hit a fundamental noise floor, where the effects of the measurements and of uncertain unknowns wash out further precision. Past that point, even if "shut up and multiply" is technically "correct", it's out-of-touch masturbation to think that the result of that multiplication is usable. More concretely:
The error bars of uncertainty on numbers get multiplied too. If you try to do something stupid like multiply a tiny, fluctuating, and uncertain ethical value (like the impact of dust specks in eyes) by incomprehensibly large numbers of people, those error bars get incomprehensibly large too. For most questions, they probably get large enough to swallow so much of the result range that you can't actually use the result to pick a decision, because the possible results include all or most possible results which would warrant completely different decisions.
Even if the correct result to the "torture vs specks" thought experiment is torture, this result has limited practical applicability. Both due to the above, and because once you integrate the future you realize that there are many non-negligible downsides over time to choosing greater suffering for some just to avoid a negligible inconvenience for most.