QuestionsIs the world real?The simulation argument: three possibilities, one of which is true
Is the world real?Probably a simulation

The simulation argument: three possibilities, one of which is true

The simulation argument (Bostrom); the simulation trilemma; are we living in the Matrix?

Nick Bostrom born 1973One of three things is true: we die out before becoming "posthuman", or posthumans never simulate their ancestors, or we almost certainly live in a simulation.
See profile
What it claims

At least one of these three things is true: (1) almost no civilization reaches the posthuman stage; (2) almost no posthuman civilization runs such simulations; (3) we are almost certainly in a simulation. Bostrom does not choose among the three, and suggests spreading one's confidence roughly evenly among them.

Why

If many advanced civilizations run simulations of human histories, almost all experiences like ours would be simulated.

The same idea through history
René Descartes1641In the first Meditation, Descartes imagines an all-powerful "malignant demon" who deceives him about the sky, the earth and his own body. It is the ancestor of doubt about the external world. Bostrom nevertheless says his argument is of a different kind: it does not start from doubt, it starts from a technical fact (see the note on this below).Hilary Putnam1981Putnam tells the brain in a vat, wired to a computer that sends it sensations. Bostrom says on his website that his argument is "fundamentally different" from this story: it is not meant to raise a sceptical problem, but to draw a conclusion about the world from what technology may one day do.
Nick Bostrom2003The version presented here: the three possibilities and the reasoning about the number of simulated minds. Bostrom does not defend the third one alone: he defends the disjunction. He himself says he gives the simulation hypothesis a "substantial probability".
Elon Musk2016In June 2016, at the Code conference, Elon Musk gave a popular summary: if video games become indistinguishable from reality and are played everywhere, the odds that we are in base reality are one in billions. This "we probably live in a simulation" version is the one the general public remembers, and it is why this page sits in the camp "Probably a simulation": Bostrom himself does not go that far.
In short

"The simulation argument: three possibilities, one of which is true" is an argument by Nick Bostrom ("Are You Living in a Computer Simulation?", Philosophical Quarterly, 2003, 2003) on the question "Is the world real?" (answer: probably a simulation). Its claim: At least one of these three things is true: (1) almost no civilization reaches the posthuman stage; (2) almost no posthuman civilization runs such simulations; (3) we are almost certainly in a simulation. Bostrom does not choose among the three, and suggests spreading one's confidence roughly evenly among them. According to AskPhilosophy's logic check (Z3 solver), the conclusion does follow from its 4 premises. AskPhilosophy presents 3 objections, 3 of them with a documented reply, and the original text behind each premise, checked word for word.

"Are You Living in a Computer Simulation?", Philosophical Quarterly, 2003
Valid

Imagine that our descendants become so powerful in computing that they can simulate, in their computers, millions of complete human histories, with simulated people who think and feel. Then simulated minds would be far more numerous than "original" ones. In 2003, Nick Bostrom draws from this not one but three possibilities, at least one of which is true: either almost no civilization reaches this stage (it dies out first), or almost none of those that reach it run such simulations, or we are almost certainly in a simulation. Careful: Bostrom does not say the third is true. He defends the disjunction, and says one should spread one's confidence among the three. What the general public remembers ("we are probably in the Matrix") is a simplified version, which is why the argument sits in this camp.

P1A computer running the right program would have real conscious experiences: consciousness does not depend on the stuff one is made of (neurons or chips) but on the organisation realised. Bostrom admits this point is not entirely uncontroversial, but he takes it as given.
P2A "posthuman" civilization (technologically very advanced) would have enough computing power to simulate a great many human histories, using only a tiny part of its resources.
P3So if a non-negligible share of civilizations reach this stage, and a non-negligible share of those run such simulations, then almost all the people who have experiences like ours are simulated people.
P4"Weak" indifference principle: if we know that a share x of the people with experiences like ours are simulated, and nothing in our experience tells us whether we are among them, then our degree of confidence that we are simulated should be x.
So at least one of these three things is true: (1) almost no civilization reaches the posthuman stage; (2) almost no posthuman civilization runs such simulations; (3) we are almost certainly in a simulation. Bostrom does not choose among the three, and suggests spreading one's confidence roughly evenly among them.

click a premise to reject it

If you accept every premise, logic forces you to accept the conclusion. So the debate is about the premises.
Checked by a logic solver: the premises cannot all be true while the conclusion is false.
P2 and P3 say: if the first two possibilities are false (civilizations reach the posthuman stage, and some run simulations), then almost everyone like us is simulated. P4 takes us from this proportion to our own case: since nothing distinguishes us, we are very probably among the simulated. P1 is needed for simulated people to count as people "like us". So: either (1), or (2), or (3). The reasoning is valid; what is debated are the premises, above all P1 (simulated consciousness) and P4 (indifference).
Analogy. In a town where 999 out of 1000 people wear a superhero costume, and if you cannot see your own outfit, the best guess is that you are probably in costume too. Bostrom's reasoning is the same, with "simulated people" in place of "people in costume". But first you have to know that there really are that many in costume: that is where the three possibilities come in.

Objections

3

Each objection is a conversation: click to see who replies and where the debate stands.

B
Brian Weatherson 2003 · attacks premise 4

Weatherson accepts the reasoning up to the last step but rejects the indifference principle. He notes that our evidence may not be limited to what we feel: seeing a real dagger is not the same evidence as "seeing" a simulated dagger. Then it cannot be said that nothing distinguishes a human from a simulated person, and the proportion x no longer gives our degree of confidence.

Bostrom replies
Nick Bostrom 2005 · Bostrom's reply (2005)
Summary: Bostrom replies that the best answers to the sceptic avoid making perceived appearance an absolute proof. There are cases (like a hall of distorting mirrors) where we would have good reason to distrust our senses. He also devotes sections to the exact interpretation of the principle and to an objection based on "grue"-like predicates.
Where the debate stands (our summary)Weatherson and Bostrom agree on the logic of the early steps. The disagreement is about the final step: can one go from a proportion of simulated people to a personal degree of confidence, when our evidence may be more than our sensations alone? Open point: neither convinced the other (our summary).
John Searle 1980 · attacks premise 1

Searle argues that a simulation is not a duplicate: no one believes a simulated fire burns the neighbourhood, or a simulated rainstorm drenches us. Why should a simulation of understanding understand anything? If Searle is right, Bostrom's "simulated" people would have no consciousness, and would not count among people like us.

a defender replies
see the argument
No reply possible: objection raised in 1980, after the author's death (null).
David Chalmers defender of the argument · 1995
Summary: Chalmers argues that what matters for consciousness is organisation, not matter. If neurons were replaced one by one with chips doing the same job, experience should neither fade nor start dancing without the person noticing: so a system with the same organisation has the same inner life.
The debate goes on (our summary) Searle wrote in 1980, before Bostrom's argument (2003): he could not reply to it. Bostrom does not answer Searle directly either, but he writes that his thesis "is not entirely uncontroversial" and that it does not assume a strong form of functionalism (our summary of his text).
Where the debate stands (our summary)The point of disagreement: does consciousness depend only on a system's organisation (Chalmers, functionalists) or also on what it is made of (Searle)? Open point: no position is widely accepted. If Searle is right, the first link of the chain fails, and with it the third possibility.
J
Jonathan Birch 2013 · attacks premise 2

Birch points to a tension: Bostrom's argument assumes we have good evidence about the physical limits of computation (to say posthumans could simulate), while allowing that we might lack good evidence about our own constitution (since we might be simulated). If we were simulated, the evidence on which P2 rests would be suspect. Birch calls this "selective scepticism".

Bostrom replies
Nick Bostrom Bostrom's reply (website FAQ)
Summary: Bostrom replies that it is false that, in a simulation, nothing can be known about the underlying reality. Two conditional claims remain known: if we are simulated, the underlying reality allows simulations; if we are not, our evidence about technology holds at face value. In both cases, the disjunction of the three possibilities remains true.
J
Jonathan Birch 2013
Summary: Birch examines the variants of the argument that try to avoid this problem, one of which is close to the one in Bostrom's FAQ. He argues they do not preserve the original conclusion: the argument loses either its force or its empirical character.
Where the debate stands (our summary)Both sides agree one must ask what our evidence is worth if we are simulated. The basic disagreement: does Bostrom's conditional version suffice to keep the conclusion, or does it make the argument lose what distinguishes it from a mere sceptical scenario? Disputed point, no consensus (our summary).
What today's voices say
Defend it
Neil deGrasse TysonHe found the idea hard to refute in 2018, then said an objection made him reconsider in 2020.
Lex FridmanHe devoted episode 83 (2020) to the simulation argument with its author, Nick Bostrom. We found no source stating his own verdict, so this only records that he discussed it.
Nick BostromHis own argument: "the simulation argument itself shows only that one of three possibilities is true". He does not say which one, takes the simulation hypothesis seriously, and says the argument has held up very well.
David ChalmersCalls Bostrom's simulation argument clearly worth taking very seriously, and gives a gut credence of maybe 10 to 20 percent.
Criticize it
Sabine HossenfelderShe calls the simulation hypothesis pseudoscience: it makes a bold claim about the laws of nature without paying attention to what physics already knows. In her words, popularizing it is not innocent fun: it mixes science with religion, 'which is generally a bad idea'.
Sean CarrollHe is skeptical: our universe is so large and complex that he asks why anyone would go to all this trouble. He agrees most simulated worlds would be simple.
Donald HoffmanHe likes the point that we do not see reality as it is, but disagrees on two things: the bottom level is not physical space-time, and computation cannot give rise to conscious experience.