The ’1400′ number was indeed picked somewhat arbitrarily and merely to illustrate the point that our life expectancy goes up even for quite levels of AI risk. I thought this point was worth making because I’ve seen the opposite claim made or insinuated by some anti-AI advocates.
Later models in the paper take into account more factors—in particular, temporal discounting and diminishing marginal utility in QALYs. In these more full-fledged models, it is not the case that we get “launch immediately” for all non-extreme settings of the parameters even if we postulate that post-superintelligence lifespans could be a billion years.
(In general, as I said in the preamble, I recommend not focussing overly on the exact numbers that pop out of these simple models. The formal models are mostly intended to serve as illustrations of various possible reasoning assumptions and the general patterns they imply.)
Coming back to this, I’m confused about why I wrote that a billion years would simplify the model to “launch immediately” for all non-extreme parameters (before you even talk about discounting or decreasing marginal utility). E.g. if the initial risk is 50% and there’s a 10% reduction per year, you obviously don’t want to launch immediately if you’re trying to maximize the average lifespan of currently living people.
The ’1400′ number was indeed picked somewhat arbitrarily and merely to illustrate the point that our life expectancy goes up even for quite levels of AI risk. I thought this point was worth making because I’ve seen the opposite claim made or insinuated by some anti-AI advocates.
Later models in the paper take into account more factors—in particular, temporal discounting and diminishing marginal utility in QALYs. In these more full-fledged models, it is not the case that we get “launch immediately” for all non-extreme settings of the parameters even if we postulate that post-superintelligence lifespans could be a billion years.
(In general, as I said in the preamble, I recommend not focussing overly on the exact numbers that pop out of these simple models. The formal models are mostly intended to serve as illustrations of various possible reasoning assumptions and the general patterns they imply.)
Coming back to this, I’m confused about why I wrote that a billion years would simplify the model to “launch immediately” for all non-extreme parameters (before you even talk about discounting or decreasing marginal utility). E.g. if the initial risk is 50% and there’s a 10% reduction per year, you obviously don’t want to launch immediately if you’re trying to maximize the average lifespan of currently living people.