TL;DR: P(simulation|important) > P(simulation|¬important)
If we are living in a particularly important century, the probability that we're in a simulation increases.
Disclaimer: This post is kind of weird. Sorry.
Contents
Is this century important?
This month, Effective Ideas are offering mini prizes for blogs that discuss Holden Karnofsky's series on The Most Important Century. That's the prompt this post takes up, from a slightly odd angle.
Karnofsky's argument is that the current century might be the most important one in history, and his follow-up argues that digital people would be an even bigger deal than the simulations we usually imagine. My question here is different: if the century matters this much, does that change how likely it is that we're inside a simulation?
P(simulation)
Before anything else, what is the probability that we're in a simulation at all? There's a well-known survey of this, and arguments in the Reality+ space about how strange the simulation hypothesis is to take seriously.
The usual treatment runs the other way: ask what fraction of observers in a universe like ours would find themselves in a simulation, and combine that with how many simulations get made relative to how many substrate-born observers exist. The constraints come from things like the TOP500 list of supercomputers and estimates of global computing capacity — all of which are crude, and get cruder the further out you push them.
The simulation hypothesis
The hypothesis itself: our world might be a running simulation, and we are the digital inhabitants of it. Sceptics have replied in print that the whole framing is confused, which is fair enough — the interesting move is to ask what the hypothesis predicts rather than whether it feels odd.
Here's the thought. If the century we're living in is unusually important, then the fact that we find ourselves in it becomes evidence about the kind of world we're in.
Why would we be simulated?
The usual proposed motivations for running ancestor simulations are research, entertainment, and insurance — the argument that a sufficiently advanced civilisation might run copies of its own past for instrumental or recreational reasons. The duplicator argument is one place this gets developed in detail.
Note what all of these motivations have in common: they are about the simulators and what they want. None of them is about us, and none of them selects for the century we happen to be in. So the importance of the century can't, on its own, explain why a simulation is running.
P(important | simulation)?
But we can still ask the conditional the other way round: given that we're in a simulation, how likely is it that the century is important?
And here there is a plausible asymmetry. If the century is very important — if it contains the transition to something genuinely different — then a simulator has a strong incentive to run it, because it's the interesting one to watch. If the century is unremarkable, there is little reason to simulate it at all; you'd run it only by accident, or as a background process.
So we have something like:
Note the direction. This is a claim about the conditional probability of simulation given importance, not about importance given simulation. That's exactly the shape of argument I want to make, and it's why the framing is not just a restatement of "important things get simulated."
Anthropic reasoning all the way up
There's a familiar move here: we should expect to find ourselves in a place that is typical of where observers like us end up. Applied recursively, this is the standard anthropic-style reasoning used to argue for our position in the universe — a selection effect over observers rather than a causal claim.
Taking that seriously, the position of this century is not random either. Observers who happen to exist in a simulation of a pivotal century are, in a sense, over-represented relative to observers in simulations of dull centuries — because pivotal centuries are the ones that get simulated more, and each simulation generates its own observers. That is a self-referential loop, and self-referential loops are where this kind of argument usually goes wrong, so it's worth keeping the loop explicit rather than hiding it.
Filling out the formula for P(simulation | important)
Bayes' rule makes the pieces explicit. Writing i for "the century is important" and s for "we are in a simulation":
So the argument needs three quantities, each of them shaky in a different way:
| Quantity | What it depends on |
|---|---|
| P(s) | Base rate of simulations, from computing capacity and the costs of running them — plus how many observers sit in each. |
| P(i) | How likely this century is to be important in the first place, which is where Holden Karnofsky's priors enter. |
| P(i | s) | The key term: whether important centuries are preferentially simulated, relative to how often each century is simulated at all. |
The third term is doing all the work. If simulations are drawn uniformly across centuries, then P(i | s) ≈ P(i) and nothing interesting happens. The inequality only survives if there is a real preference for simulating important centuries, and that preference has to come from the simulators' motives rather than from any property of the century itself.
Which is why the motives in the previous section matter so much: research purposes, entertainment, and insurance differ enormously in how much they favour pivotal periods. A simulator who wants to study a transition would run the interesting century; a simulator who wants entertainment might not care either way. The argument's strength inherits all of that uncertainty.
Ramifications
If the century is important and its importance raises our probability of being simulated, then some of the usual debates about simulation are being conducted on the wrong footing. It isn't only "is the simulation hypothesis true?" — it's "does the significance of our era feed back into the prior?"
Two consequences seem worth flagging.
- Importance and artificiality are entangled. If being important makes simulation more likely, then arguing that this century is pivotal partially undermines arguments that we are in "ordinary" base reality. The two claims pull in the same direction, which is uncomfortable if you're trying to reason from either alone.
- The preference term is the whole argument. Every step above that isn't about simulators' motives is bookkeeping. Which raises a question I can't fully settle here: are there reasons to think the things running simulations would pick pivotal centuries over boring ones, or only reasons to think they might? The references gathered here — including this list of intellectual, philosophical and technological revolutions — sketch how many candidate transitions there might be, and hence how many simulations a thorough simulator would need to run.
I'm not claiming the simulation hypothesis. I'm claiming something narrower: that the importance of the current century is not background colour for the simulation debate, it is one of the terms inside it. That's uncomfortable, and I don't have a clean way to make it comfortable.
More posts
-
Validation as a bottleneck for agency
On how our capacity to act is a result of our social environments, and why supportive networks matter more than self-encouragement.
-
A minefield of attractor states
A version of the argument that we could be living in the most important century, with an emphasis on why this century's level of influence is distinct.
-
What if we're already digital?
The simulation hypothesis and its implications for the importance of the current century.