The Potatium Shockwave
SHAREphiliptrammell.com/blog/41  ·  1 Aug 2019  ·  #41
If you hang around utilitarians for very long, you will learn the following jargon.
·"Hedonium" is the hypothetical substance that is optimized for feeling good. Whatever your brain is on a beautiful day, when you’ve just won a Nobel Prize, and you’re on drugs, the substance of your brain is then a bit “closer to hedonium” than usual.
·“Dolorium” is likewise the hypothetical substance optimized for feeling bad.
·When you are in dreamless sleep, or having a boring day you’re not particularly enjoying or disliking, your brain is at “hedonic zero".
·The best possible future, from a total hedonic utilitarian perspective, is presumably a universe filled with hedonium. This might be achieved by launching robot-rockets programmed to convert other planets and stars into hedonium—and, presumably, into more robot-rockets capable of doing the same. The grand vision is ultimately a sphere of hedonium, centered at Earth, expanding ceaselessly until the end of time. This process is sometimes called a “hedonium shockwave”. The sooner the better.
·One implication of total hedonic utilitarianism is that a universe with very many experiences just slightly better than hedonic zero is better than a universe with fewer experiences each of which is in bliss. Parfit (1984) called this the repugnant conclusion. In illustrating an experience just slightly better than hedonic zero, his memorable (1986) example is the experience of listening to muzak and eating potatoes.

Suppose creating a good experience comes with some fixed costs (e.g. the energy it takes to mine the substance to be turned into the experience), plus some variable costs which are convex in the hedonic intensity of the experience. This assumption seems very weak and reasonable, to me—at least as reasonable as the assumption that we will one day be able to launch self-replicating hedonium-producing robot-rockets. It is simply the production function we face with respect to experiences today. It costs some roughly fixed amount to create an additional human or animal, and it costs some amount to make that creature happy; and happiness is concave in consumption, meaning that the second dollar spent on making a creature happy doesn’t provide them with as much extra happiness as the first dollar does.

Under this arrangement, spreading half as quickly—i.e. spending twice as much time and energy on optimizing the experiences into which the robots are converting a given planet—will eventually fail to produce twice as much welfare from that planet as would have been produced by the rough-and-ready experience-creating job. And this limit will not generally be reached before the planet has been fully optimized, just as having a second child often produces more welfare than spending twice as much on one’s first child, even though the first child is not in a state of perfect bliss. The robots will then maximize total good feeling by only somewhat optimizing the stuff of a given planet, or whatever, before moving on to the next. More precisely: if the fixed cost of creating an experience is F, and the variable cost in hedonic intensity is some smooth, convex function V(h), the robots will maximize welfare by creating experiences not of maximal hedonic intensity but of hedonic intensity h : F + V(h) = hV′(h). This will be positive, but as far as I can tell there’s no strong reason to think it will be high.

To paraphrase the great Jack Handey: I hope God likes potatium, because that’s what he’s getting.

41
This scenario seems conditional on there being a fixed number of robots working on things. Presumably, a late-stage civilization which has figured out hedonium probes will have technology advanced enough to stop at each planet exactly long enough to create a replica of itself which whizzes off again and one more robot which can spend a longer amount of time optimizing the available matter for hedonic experience; the only bottleneck there is a matter tradeoff, where it might turn out that it's possible to more densely pack a potato-y experience than a state of ultimate bliss into the planet by a larger factor than the factr by which bliss is better than potatoes. This latter tradeoff seems plausible but not highly probable to me, and the tradeoff is such that I would expect things to not be very potato-y because 1/x punishes things badly for being near 0. Although even in the scenario where Von Neumann probes are impossible for some reason, amount of new space accessible per time grows (at best) as the O(t^2) surface of a sphere, and will shrink as space expands (making matter less dense). So over time the future value of sending out a probe diminishes, and the value of spending time optimizing new planets increases - I think most models with finite time horizons end up converting a large fraction of their light-cone into matter which is pretty well optimized (up to the density constraints mentioned above).
RavenclawPrefect  ·  17 Sep 2019 9:11 PM
POST
1