After watching a talk by Stuart Armstrong it struck me that I had drastically underestimated the force of the atronomical waste argument. Previously I had considered the utility of colonizing the Milky Way. However, in the light of his talk colonization of billions of galaxies did not seem very unlikely (say below a 0.01 probability). This adds a whole bunch of orders of magnitude.
Now it seems to me that even this estimate is an underestimate and rather than being a drastic underestimation it is an astronomical underestimation. The reason is that a standard total utilitarian should consider not only a few of the most likely scenarios and their values but also add together "a tail" of all other scenarios with a probability above 0 and their utilities. In my epistemic state prior to hearing about Armstrong's talk I should reasonbly at least have factored in, say a 0.1 probability of colonizing tens of galaxies, perhaps a 0.001 probability of colonizing billions of galaxies etc, since I had no reason to think that there are no possible ways to effectivize colonization. These estimates should then be updated when new evidence on ideas or methods for colonization are available.
Even spending a little time on this it seems like the expected utility provided by this "tail" would be enormous. For example, consider the possibility of travel across arbitrary distances (one interpretation of the "wormhole" idea in science fiction). Assuming space is infinite, If such a technology and the technology Armstrong describes in his talk is possible (self-replication von-neumann probes that can be powered by a star and use planets around the star to create massive solar power arrays) an exponential expansion that would last as long as the stars would be possible. After replicating a and harnessing the power of one star the probe could use the power to create, say 10, new probes and send through separate wormholes to distant parts of the universe. Assuming that stars will continue to shine for 10 billion years and that this cycle could be completed in 10 years (Armstrong calculated 40 years to strip mine mercury to construct an enormous solar powered mass driver to hurl billions of von neumann probes towards other galaxies at speeds close to c) then it follows that the number of probes (and thus, colonized galaxies) would increase tenfold every ten years and Bostrom's original Astronomical waste calculation should be increased by about *a billion orders of magnitude*.
I think it is fair to say that this number is not astronomical, but *astronomically astronomical*.
Is this scenario likely? Probably not. But I have a hard time justifying a subjective probability of less than, say, 1/10000 that such super-luminar travel and the other conditions are possible. So in a naive expected utility calculation the expected utility of reducing x-risk would still be approximately a billion orders of magnitude.
Also note that wormhole-travel is just one possible technology. Other, perhaps more far fetched, technologies that could lead to far greater utilities than this include time travel and finding some renewable source of energy and the creation of lab-universes.
Is this just another example of Pascal's mugging/wager? Perhaps. But I think that an important point is that in such a case we cannot draw any clear boundary between pascalian and non-pascalian cases. There is a continuum of more and more "pascalian" cases, starting with low probability/high impact actions such as eating vegan or donating money to effective altruism, the chance of affecting a moderate space expansion, a smaller affecting colonization of the milky way, an even smaller colonization of colonizing billions of galaxies, of colonizing astronomical numbers, of getting unbounded amounts of utility all the way up to infinite and higher-order-infinite outcomes.
(If you are a negative utilitarian an analogous argument could be stated about the astronomically astronomical importance of reducing future suffering).
Now it seems to me that even this estimate is an underestimate and rather than being a drastic underestimation it is an astronomical underestimation. The reason is that a standard total utilitarian should consider not only a few of the most likely scenarios and their values but also add together "a tail" of all other scenarios with a probability above 0 and their utilities. In my epistemic state prior to hearing about Armstrong's talk I should reasonbly at least have factored in, say a 0.1 probability of colonizing tens of galaxies, perhaps a 0.001 probability of colonizing billions of galaxies etc, since I had no reason to think that there are no possible ways to effectivize colonization. These estimates should then be updated when new evidence on ideas or methods for colonization are available.
Even spending a little time on this it seems like the expected utility provided by this "tail" would be enormous. For example, consider the possibility of travel across arbitrary distances (one interpretation of the "wormhole" idea in science fiction). Assuming space is infinite, If such a technology and the technology Armstrong describes in his talk is possible (self-replication von-neumann probes that can be powered by a star and use planets around the star to create massive solar power arrays) an exponential expansion that would last as long as the stars would be possible. After replicating a and harnessing the power of one star the probe could use the power to create, say 10, new probes and send through separate wormholes to distant parts of the universe. Assuming that stars will continue to shine for 10 billion years and that this cycle could be completed in 10 years (Armstrong calculated 40 years to strip mine mercury to construct an enormous solar powered mass driver to hurl billions of von neumann probes towards other galaxies at speeds close to c) then it follows that the number of probes (and thus, colonized galaxies) would increase tenfold every ten years and Bostrom's original Astronomical waste calculation should be increased by about *a billion orders of magnitude*.
I think it is fair to say that this number is not astronomical, but *astronomically astronomical*.
Is this scenario likely? Probably not. But I have a hard time justifying a subjective probability of less than, say, 1/10000 that such super-luminar travel and the other conditions are possible. So in a naive expected utility calculation the expected utility of reducing x-risk would still be approximately a billion orders of magnitude.
Also note that wormhole-travel is just one possible technology. Other, perhaps more far fetched, technologies that could lead to far greater utilities than this include time travel and finding some renewable source of energy and the creation of lab-universes.
Is this just another example of Pascal's mugging/wager? Perhaps. But I think that an important point is that in such a case we cannot draw any clear boundary between pascalian and non-pascalian cases. There is a continuum of more and more "pascalian" cases, starting with low probability/high impact actions such as eating vegan or donating money to effective altruism, the chance of affecting a moderate space expansion, a smaller affecting colonization of the milky way, an even smaller colonization of colonizing billions of galaxies, of colonizing astronomical numbers, of getting unbounded amounts of utility all the way up to infinite and higher-order-infinite outcomes.
(If you are a negative utilitarian an analogous argument could be stated about the astronomically astronomical importance of reducing future suffering).