Bostrom’s trilemma

Picture a civilisation that gets advanced enough to run detailed simulations of its own past, with people in them who think and feel. Bostrom’s paper argues that at least one of these is then true. Almost no civilisations reach that stage, because something stops them first. Or almost none that reach it choose to run such simulations. Or we’re almost certainly living in one. If ancestor simulations get run at all, simulated minds vastly outnumber original ones, and a mind picked at random is probably one of the many.

The argument is short, and it’s careful about what “almost” means. It doesn’t say which of the three holds. That sets it apart from earlier versions of the question. Plato, Zhuangzi, Descartes and Putnam ask whether you can know. Bostrom asks about probabilities, and shows you can’t make all three possibilities small at once.

People often try to find a fourth option. The qualifications are real: how many minds a simulation holds, whether simulated civilisations run simulations of their own, what counts as maturity. They change how much weight each of the three carries. The argument itself rests on two premises, substrate independence and an indifference principle. Denying either one is the main route to a fourth possibility.

The paper also explains why this site looks the way it does. Bostrom’s argument contains a number, the simulated minds per original mind. That number depends on how much computing a simulation takes. The sandbox makes that explicit. The Trilemma window treats the scenario you chose as the thing being simulated. It works out how many copies one universe-scale computer could run, and lets you set the other two probabilities. The third bar is just arithmetic.

Bostrom has said he spreads his credence across all three. Informed people have settled on each one. The trilemma doesn’t tell you where to stand. It only says that at least one of three uncomfortable propositions is true.