Contribution
A short paper that turns the size of the reachable universe into a practical conclusion. It counts the potential lives lost to delay. It then shows that the same numbers make the probability of eventual settlement matter far more than its timing. It is the source of the expected-value argument for prioritizing existential risk, using the concept from Bostrom 2002 (§III, n. 8).
Key insight
Delay is measured in years and the future in billions of years. So anything that changes whether settlement happens outweighs anything that changes when. “If our actions have even the slightest effect on the probability of eventual colonization, this will outweigh their effect on when colonization takes place” (§III ¶3).
Assumptions
- Lives that could be created would be worth living (§II ¶1), called a weak assumption.
- Aggregation without time-discounting. The conclusion holds for any evaluation that “does not count one person’s welfare for less just because there are many other persons” and “is not relativized to a particular point in time (no time-discounting)” (§II ¶3). Partial aggregation and “some degree of time-discounting” are said to leave it standing (§II ¶4).
- Minds can be run on computers, for the larger estimate (§I ¶4, n. 5). A biological estimate is also given.
- Infinite consequences are set aside (n. 7).
Key results
- The rate of loss (§I).
- The Virgo Supercluster has about 10^13 stars. A star with nanotechnological computing gives 10^42 operations per second; a human brain about 10
- So about 10^38 potential lives are lost per century of delay, or about 10^29 per second (§I ¶4).
- With 10^10 biological humans per star, 10^23 in all, the loss is over 10^13 per second: “over ten trillion potential human beings” (§I ¶5–6).
- Total utilitarian conclusion (§§II–III).
- Advancing settlement by one second brings about more than 10^29 (or 10^13) lives (§II ¶2).
- But a single percentage point of existential risk reduction “would be worth … a delay of over 10 million years” (§III ¶2).
- “For standard utilitarians, priority number one, two, three and four should consequently be to reduce existential risk” (§III ¶3).
- Person-affecting conclusion (§IV).
- On that view, obligations run “primarily towards currently existing persons and to those persons that will come to exist”. Extinction is bad because it makes lives worse, not because it loses potential lives (§IV ¶1).
- If no one now alive will reach the cosmic resources, extinction costs about 40 years for each of six billion people: a very important concern, “in the same big ballpark” as poverty, hunger and disease, and not dominant (§IV ¶2).
- But given a chance (Bostrom suggests not less than 1%) of a singularity or life extension within current lifetimes, the expected benefit to people now alive of “living for perhaps billions of subjective years” may be astronomical (§IV ¶3–5).
- So the person-affecting utilitarian must balance safety against speed: “it is not enough that humankind survives to colonize; it is crucial that extant people be saved” (§IV ¶6). The total utilitarian always picks safety.
- Non-identity (n. 14). On the person-affecting view, actions with large effects on the future change which people come to exist, so “there would typically be no future persons whose level of well-being is affected by our current actions”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Each second of delayed settlement of the local supercluster loses the potential for at least 10^13 (biological) to 10^29 (computational) worthwhile lives | weak to moderate as numbers, by Bostrom’s own account (“what matters … is not the exact numbers”); depends on speculative technology | §I |
| C2 | On any aggregative, undiscounted axiology, this delay is an enormous loss of value | moderate: follows from C1 and the axiology | §II |
| C3 | For total utilitarians, reducing existential risk outweighs accelerating development | strong conditional on C1–C2 and expected-value reasoning | §III |
| C4 | Person-affecting utilitarians should also give existential risk great weight, but balanced against speed | weak: rests on a guessed 1% chance of radical life extension and on non-declining marginal utility of lifespan | §IV |
Concepts
- astronomical waste: the potential value lost as resources that could sustain worthwhile lives go unused.
- existential risk: an outcome that “would either annihilate Earth-originating intelligent life or permanently and drastically curtail its potential” (§III ¶2).
- person-affecting utilitarianism: obligations primarily to existing persons and those who will come to exist (§IV ¶1). Bostrom calls the formulation simple, “not necessarily the best possible one” (n. 9).
Connections
- Bostrom 2013 (LIT-1773) restates the argument with other figures (10^16 lives on Earth alone) and defines existential risk formally.
- Greaves and MacAskill (LIT-1684) cite this paper as precedent and generalize it beyond total utilitarianism.
- Narveson (LIT-1769) and the non-identity problem (LIT-1768). Section IV and note 14 are a person-affecting treatment: they owe nothing to merely possible people, and non-identity removes most future people from view.
- Thorstad (LIT-1688) questions the astronomical value of risk reduction this paper first asserted.
Bearing on the record
- The owner’s thesis (CLAIM-491). These are inferences; Bostrom
does not discuss motivation or the self.
- Two grounds, side by side. On the total view the obligation is to an impersonal sum of value, owed to no one. On the person-affecting view it is grounded in the good of people now alive. Bostrom shows that the two give different advice (safety alone, or safety balanced against speed).
- The person-affecting case is ordinary self-interest. The reason it gives a living person to care about the far future is that they, or people now alive, might live billions of subjective years in it. That is concern for the future as one’s own future, with no larger self needed. It is a third reading beside the thesis’s extended self and the impersonal good, closest to Pitcher’s on QUESTION-301.
- Nothing on motivation. The paper says what “standard utilitarians ought to” do (abstract) and gives no thought to why anyone is moved.
- Discounting. The conclusion needs no time-discounting but survives “some degree” of it (§II ¶4), a weaker dependence than Greaves and MacAskill concede.
- Proposed dated note on CLAIM-491: “Bostrom’s Astronomical Waste (NOTE of 2026-10-29) gives two grounds for concern with the far future and shows they diverge. On the total view the far future is owed to an impersonal sum, and its advice is safety alone. On a person-affecting view it matters because people now alive might live billions of years in it, which is ordinary self-interest, and its advice is to balance safety against speed. A thesis-friendly reading of longtermist concern needs no extended self where that second ground operates.”
- No THEORY filed.
- No ML instruction.
Limitations
- Author’s HTML, not the journal text. Not compared.
- The figures are speculative, as Bostrom says (§I ¶6). The computational estimate assumes uploads and star-scale computers.
- The person-affecting section rests on guesses: a probability of radical life extension “of less than, say, 1%” called unwarranted (§IV ¶4), and non-declining marginal utility of lifespan (§IV ¶5).
- Mixed views are left open (§IV ¶6: they “might or might not” yield either bottom line).
- Typos: “One exception are fetuses” (n. 14).
Open questions
- Do people who care about existential risk do so for the person-affecting reason (their own or their contemporaries’ long future) or for the total one? Schubert et al. (LIT-1748), when read, may bear on it.