Are we living in a simulation?
If advanced civilisations can run vast numbers of simulated worlds, simulated minds would vastly outnumber real ones. So which are you more likely to be?
- Reading time
- 4 min
- Difficulty
- Intermediate
- Branch
- Reality
- First posed
- Nick Bostrom, 'Are You Living in a Computer Simulation?', Philosophical Quarterly, 2003
The setup
Computers have gone, in a single human lifetime, from rooms of vacuum tubes to devices that render realistic worlds in real time. Extrapolate a few thousand years, or a million. A sufficiently advanced civilisation might have the computing power to simulate entire worlds, including the brains of every person in them, in enough detail that the simulated people would be conscious.
If such civilisations exist and choose to run many such simulations, perhaps of their own history ('ancestor simulations'), then the number of simulated minds could vastly exceed the number of biological ones. Each simulation might contain billions of people; a civilisation might run millions of simulations.
In 2003 Nick Bostrom turned this into a careful argument. He did not claim we are in a simulation. He claimed that at least one of three propositions must be true: (1) almost all civilisations at our stage go extinct before becoming able to run such simulations; (2) almost all civilisations that could run them choose not to; or (3) we are almost certainly living in a simulation.
Why it matters
The argument uses only modest assumptions: that consciousness can be produced by computation, that enough computing power is physically possible, and a principle of indifference about where you are. Yet it concludes that one of three dramatic possibilities is true. Many people find it impossible to dismiss, and impossible to accept.
It also reframes old questions. The simulation hypothesis is a scientific-sounding version of skepticism and of religious ideas about a creator outside the world. A simulator would, in effect, be a god to us.
Key terms
- Simulation hypothesisThe hypothesis that our world, including ourselves, is a computer simulation run by some more advanced civilisation.
- Substrate independenceThe assumption that a mind depends on the pattern of information processing, not on the material it runs on, so a mind could exist in silicon as well as in neurons.
- Principle of indifferenceThe rule that, without any reason to favour one possibility over others, you should assign them equal probability.
- PosthumanA hypothetical future being, descended from humans, whose capacities have been so extended by technology that it is no longer human in the ordinary sense.
- Anthropic reasoningReasoning that takes into account the fact that we exist to make an observation. For example, we should not be surprised to find the universe compatible with life, since we could not observe one that was not.
The positions
We are probably simulated
The third disjunct is the likeliest. Given what we know, we should assign a significant probability to being in a simulation.
The trilemma holds, but not option three
The argument is valid, but options one or two are more likely: civilisations die out, or do not run such simulations.
The argument does not work
The argument rests on false or untestable assumptions, especially that simulated beings would be conscious and that we can reason about our 'typicality' this way.
It would not matter much
Even if we are simulated, our world is real: the objects around us exist as computational structures, and our lives are no less meaningful.
Arguments and objections
We are probably simulated
- Best argument
- There is no known physical law forbidding the computing power required, and our own civilisation is visibly heading toward ever more detailed simulations. If even a small fraction of civilisations run many simulations, simulated minds would dominate. If you have no evidence about whether you are simulated, the principle of indifference says you should assume you are typical, and the typical mind would be simulated.
- Strongest objection
- Each step involves enormous uncertainty: whether consciousness can be simulated, whether posthumans would want to, whether the computing power is possible. Multiplying uncertainties does not produce a confident conclusion.
- The reply
- Bostrom agrees, which is why he presents the argument as a three-way disjunction rather than a proof.
The trilemma holds, but not option three
- Best argument
- There are many reasons civilisations might not run ancestor simulations: ethical concerns about creating billions of suffering people, lack of interest, or regulations. And perhaps no civilisation survives long enough. The argument does not tell us which option is true; it is a reason to worry about option one as much as to wonder about option three.
- Strongest objection
- Options one and two require near-universal agreement across all civilisations. If even a small number of eccentric individuals with enough power run simulations, the numbers swing toward option three.
The argument does not work
- Best argument
- If consciousness depends on biology, simulated people would not be conscious, so there would be no simulated observers to outnumber us. The indifference principle is also questionable: we know we are conscious and embodied, and the reference class of 'all observers' may be ill defined. And a hypothesis that predicts nothing different from the real-world hypothesis may not be a hypothesis at all.
- Strongest objection
- Doubts about machine consciousness are themselves uncertain, and even a modest probability that simulated beings are conscious keeps the argument alive.
It would not matter much
- Best argument
- David Chalmers argues that the simulation hypothesis is not a skeptical hypothesis but a metaphysical one. If we are in a simulation, then tables, trees and people exist; they are just made of bits rather than (or as well as) particles. Virtual objects, on his view, are real objects.
- Strongest objection
- Many people feel that discovering their world was a simulation would be devastating, not neutral. It would mean that its creators could end it, edit it, or watch it.
Who argued what
- We are probably simulated Elon Musk
- The trilemma holds, but not option three Nick Bostrom
- The argument does not work John Searle, Sabine Hossenfelder
- It would not matter much David Chalmers
Variations
Nested simulations
If we are simulated, our simulators might be simulated too, and so on. If our own descendants run simulations, the stack deepens. Some have noted that each level would need fewer resources than the one above, so the nesting cannot go on forever; there must be a base reality somewhere.
Glitches
Could we detect a simulation by looking for errors, such as the universe showing signs of finite resolution or computational shortcuts? Some physicists have proposed experiments. Others note that a sufficiently advanced simulator could edit our memories of any glitch.
Go deeper
An old idea in new clothing
The worry that the world is not as it seems is ancient. Plato's cave imagined prisoners who mistake shadows for reality. Zhuangzi's butterfly dream asked whether waking life is itself a dream. Descartes' demon could feed us a false world. The simulation argument is distinctive because it does not merely raise the possibility; it argues that, given certain assumptions, the possibility is probable.
Related ideas
Further reading and sources
- EncyclopediaStanford Encyclopedia of Philosophy: search “simulation hypothesis”
- WikipediaSimulation hypothesis
- PaperAre You Living in a Computer Simulation?, Nick Bostrom
- BookReality+, David Chalmers
Open questions
- If you found out you were simulated, what would change about how you live?
- Would a simulated world be less real, or just differently made?
- Would it be ethical for us to simulate conscious beings?