The Larger World ArchiveClaims · Sources · Evidence
Space, physics and the cosmosPhilosophical argumentPhilosophy · 2003 – present

The Simulation Hypothesis

Bostrom's argument is not “we are probably simulated”. It is a trilemma, and one of the three horns is boring. The internet took the exciting horn and dropped the rest.

A computer-generated Julia set fractal. Bostrom's 2003 argument is a trilemma with three h
A computer-generated Julia set fractal. Bostrom's 2003 argument is a trilemma with three horns — and the internet only ever quotes the third. Image: Adam majewski · CC BY-SA 4.0 — Wikimedia Commons.

What the paper says

In 2003 the philosopher Nick Bostrom published “Are You Living in a Computer Simulation?” in The Philosophical Quarterly. The argument is conditional, and it has three parts. At least one of these must be true:

  1. Cognitive civilisations go extinct before reaching a “posthuman” stage capable of running detailed ancestor simulations.
  2. Posthuman civilisations are not interested in running such simulations, or do not run many of them.
  3. We are almost certainly living in a simulation — because simulated observers would vastly outnumber real ones.

The reasoning in brief: if it is possible and cheap to run enormous numbers of detailed ancestor simulations, then simulated minds would outnumber genuine ones by a huge factor. A randomly selected observer should then expect to be simulated. So either nobody gets that far, or nobody bothers, or we are simulated.

Bostrom is explicit that he does not know which horn is true, and that the argument is not evidence of anything in particular. The interesting content is the structure, not the conclusion.

Why “glitches” prove nothing

The internet version points at déjà vu, a repeated texture, a coincidence, or a frame-rate artefact in a video and says: code. Every one of those has an ordinary explanation in neuroscience, perception, and compression — and none of them would be expected to occur if a simulation were well built.

Worse, “glitch” arguments are self-defeating: a simulation capable of rendering a civilisation for 13 billion simulated years, at the resolution of quantum field theory, is not going to leave a visible seam in a doorway.

There is one real physics-flavoured line of work worth naming: in 2012, Beane, Davoudi and Savage published “Constraints on the Universe as a Numerical Simulation”, asking what a simulated universe would look like on a lattice and whether current cosmic-ray data could constrain it. It is a legitimate, published, speculative paper — and it constrains possible simulation schemes, it does not detect one. That is the correct level of claim.

The problem with the trilemma

It assumes substrate independence. The counting argument treats minds as patterns that can be run on anything. If consciousness depends on the specific physical stuff of brains, “running” a mind is not like running software, and the numbers never add up.

It assumes simulations are cheap and numerous. Simulating one mind is fine; simulating a planet's worth of physics at quantum resolution, for billions of years, may be physically impossible even for a posthuman civilisation — which lands you on horn one or two.

It is not testable. As stated, the hypothesis predicts exactly the same observations as the non-simulated alternative. Compare many-worlds, which at least has a defined mathematical structure that could in principle be distinguished.

It has the Last Thursdayism shape. Any disconfirming evidence can be explained as part of the simulation. See Last Thursdayism — the argument is the same, with a server instead of a deity.

What is genuinely worth taking from it

Two things. First, it is a clean demonstration of how a serious philosopher can make an unsettling argument while explicitly refusing to claim it is true — which is exactly what most of the fringe material in this archive fails to do. Second, it raises a real question about reference classes: if you are going to reason about your own existence probabilistically, you need to know what class of observer you are in, and nobody knows.

It also, incidentally, produced a useful cultural test. Ask someone what they think of the simulation hypothesis and you learn whether they distinguish a structural argument from a vibe.

Sources for this page

  • Bostrom, “Are You Living in a Computer Simulation?” (2003), full PDFsimulation-argument.com
  • Wikipedia, “Simulation hypothesis” — the trilemma and its criticsen.wikipedia.org
  • Beane, Davoudi & Savage (2012), “Constraints on the Universe as a Numerical Simulation”arxiv.org
  • Britannica, “Simulation hypothesis”britannica.com