I find the idea that net utility among people in our world today is negitive to be an extremly implausible one. Even people who have brief periods of time where their utility is so far negitive and attempt suicide (but survive) generally regret it later, as that kind of deep depression is usually something that passes pretty quickly and before long they find themselves back in a positive-utility state regretting that they almost ended their life. Most people overall enjoy life, I think, and generally want to extend it for as long as possible.
Life can be hard, and it's certanly nowhere close to optimal (yet?) but overall it's definitely a positive utility state. So much so, that the possibility of losing life (fear, danger, illness, ect) is itself one of the most utility-lowering things that can happen to a person.
As for the other point; when comparing two possibile futures you also have to consider the possibility of either one happening. The only way you could get a future with an AI torturing vast numbers of people forever is if someone specifically designed it to do that. That's not impossible, but it's far, far more likely that a person would want to create an AI that would make large numbers of people happy, either altruistically, or for economic or political reasons, or as a cooperative effort between people, ect. (And of course it's also possible that we fail and the AI just kills everyone, or that super-AI ends up not happening or not being as important as we think ect.) But just looking at those two probabilities, "AI-tortures-everyone" vs "AI-makes-people-happy", if the second one is far more likely then the first one, then you need to give that greater utility weight.
Perhaps I was a bit misleading, but when I said the net utility of the Earth may be negative, I had in mind mostly fish and other animals that can feel pain. That was what Singer was talking about in the beginning essays. I am fairly certain net utility of humans is positive.
I've started listening to the audiobook of Peter Singer's Ethics in the Real World, which is both highly recommended and very unsettling. The essays on non-human animals, for example, made me realize for the first time that it may well be possible that the net utility on Earth over all conscious creatures is massively negative.
Naturally, this led me to wonder whether, after all, efforts to eradicate all consciousness on Earth - human and non-human - may be ethically endorsable.This, in turn, reminded me of a recent post on LW asking whether the possibility of parallelized torture of future uploads justifies killing as many people as possible today.
I had responded to that post by mentioning that parallelizing euphoria was also possible, so this should cancel things out. This seemed at the time like a refutation, but I realized later I had made the error of equating the two, utility and disutility, as part of the same smooth continuum, like [-100, 100] ∈ R. There is no reason to believe the maximum disutility I can experience is equal in magnitude to the maximum utility I can experience. It may be that max disutility is far greater. I really don't know, and I don't think introspection is as useful in answering this question as it seems intuitively to be, but it seems quite plausible for this to be the case.
As these thoughts were emerging, Singer, as if hearing my concerns, quoted someone or other who claimed that the human condition is one of perpetual suffering, constantly seeking desires which, once fulfilled, are ephemeral and dissatisfying, and therefore it is a morally tragic outcome for any of us to have emerged into existence.
Of course these are shoddy arguments in support of Mass Planetary Biocide, even supposing the hypothesis that the Earth (universe?) has net negative utility is true. For one, we can engineer minds somewhere in a better neighborhood of mindspace, where utility is everywhere positive. Or maybe it's impossible even in theory to treat utility and disutility like real-valued functions of physical systems over time (though I'm betting it is). Or maybe the universe is canonically infinite, so even if 99% of conscious experiences in the universe have disutility, there are infinite quantities of both utility and disutility and so nothing we do matters, as Bostrom wrote about. (Although this is actually not an argument against MPB, just not one for it). And anyway, the state of net utility today is not nearly as important as the state of net utility could potentially be in the future. And perhaps utilitarianism is a naive and incorrect ethical framework.
Still, I had somehow always assumed implicitly that net utility of life on Earth was positive, so the realization that this need not be so is causing me significant disutility.