15 · The Computational Universe

This is Part II's closing hearing: computation, having examined every other branch of philosophy, finally asks whether reality itself will take the stand — and here the two courts run out of ground to stand on separately.

Is reality itself on the witness stand?

On 2 June 2016, at Recode's Code Conference, Elon Musk told the audience there was "a one in billions chance we're in base reality," reasoning that video games had gone from Pong to photorealism in forty years, so a sufficiently advanced civilization would run simulations indistinguishable from the world we take for granted. The claim traces to Nick Bostrom's 2003 paper "Are You Living in a Computer Simulation?" — but Bostrom's actual argument is not that claim. It is a trilemma: at least one of the following is true — almost all civilizations go extinct before they can run high-fidelity ancestor simulations; those that survive are almost universally uninterested in running them; or we are almost certainly in one (Bostrom, Philosophical Quarterly 53, no. 211, 2003: 243–255). Bostrom assigns no precise probability to the third branch and warns explicitly against overconfidence in any of the three. A careful philosophical argument, built to be cross-examined by philosophy, was cross-examined instead by a technology conference, and lost precision in the exchange. That collapse — trilemma into slogan — is itself worth entering as evidence before the argument proper begins.

Stated as Bostrom wrote it, the argument is a disjunction reached by elimination, hedged at every step: it does not say what we are, only that one of three things must be. Held to that standard, it is a piece of careful anthropic reasoning about future computing capacity — not a metaphysics, and not proof of anything about the world we are actually in.

The proposal that physical law is itself computation runs from Konrad Zuse's Rechnender Raum (1969; translated Calculating Space, MIT, 1970), which modeled space as a cellular automaton computing its own next state, through Edward Fredkin's Finite Nature hypothesis — that space, time, and every physical quantity are ultimately discrete and finite — to Seth Lloyd's Programming the Universe (2006), which treats every particle collision as a bit-flip, and to the Wolfram Physics Project, launched 14 April 2020, which models space as an evolving hypergraph from which physical law emerges. The skepticism deserves equal floor, not a footnote. Sabine Hossenfelder has argued that no one has shown how a discrete algorithm reproduces the continuous symmetries of general relativity, and that searches for a preferred discretization scale have found nothing (Hossenfelder, "The Simulation Hypothesis Is Pseudoscience," Backreaction, February 2021). Scott Aaronson makes the sharper point: the hypothesis's problem is not that it has been shown false, but that essentially no observation could count against it — a defect prior to any question of truth ("Because you asked: the Simulation Hypothesis has not been falsified; remains unfalsifiable," Shtetl-Optimized, 3 October 2017).

John Conway's Game of Life, publicized by Martin Gardner in Scientific American in October 1970, runs on three rules and produces gliders, oscillators, and — decades later — a complete computer built from its own pieces. Rule 110, one of 256 possible elementary one-dimensional automata, was conjectured universal by Stephen Wolfram in 1985 and proved so by Matthew Cook, who worked out the proof while bound by a nondisclosure agreement with Wolfram Research that delayed its publication roughly a decade ("Universality in Elementary Cellular Automata," Complex Systems 15, no. 1, 2004: 1–40). Both cases make the same point without metaphor: full computational power, and the unpredictability that comes with it, can hide inside a rule simple enough to state in one sentence.

Every "random" number a computer produces is the deterministic output of an algorithm seeded by some prior state; it looks unpredictable only because the algorithm is opaque to whoever is asking. This restages compatibilism's oldest move as engineering. Hobbes and Hume, the classical compatibilists, held that freedom never meant uncaused — it meant acting from one's own will, unconstrained by external force — so a fully determined pseudo-random generator and a fully determined human can both exhibit exactly the unpredictability-to-an-observer that ordinary judgments of freedom actually track. If physical law itself turns out to be the deterministic output of a cellular automaton, that does not resolve the question; it removes the last place to hide from having to answer it.

Every principle above runs the same shape, and the shape is the turn this book keeps making: a computational claim meets a question philosophy has held open for centuries, and neither side gets the last word. Bostrom's trilemma stays open; Hossenfelder's objection stays unanswered; compatibilism stays contested. That is not this method failing to reach a verdict. It is what an honest cross-examination looks like when the witness is reality itself, and the trial is still open when the file closes.

The people, briefly. Bostrom, a Swedish-born philosopher at Oxford, wrote the 2003 paper after first circulating it in 2001. Zuse, better known for building one of the first working programmable computers in the 1930s and 1940s, turned to digital physics afterward. Fredkin worked the idea for decades until his death in 2023. Lloyd is a mechanical-engineering professor at MIT and an architect of quantum computing. Wolfram, having built Mathematica, launched his physics project as an open, crowd-participation effort in 2020. Hossenfelder is a German theoretical physicist known for public skepticism toward untestable claims dressed as physics. Conway devised Life in 1970; Cook proved Rule 110 universal and then spent years fighting to publish it.

Connections

Chapter 02's question — are objects in a simulated world real — is this chapter's necessary predecessor. Chapter 09's Church–Turing thesis sets the limit on what "computing the universe" could even mean. Chapter 11's audit of Shannon's bracket, and Wheeler's it-from-bit, is where the speculation in this chapter meets its own foundations. Chapter 10's question of what is reachable, not merely true, asks whether any simulator could compute reality at the resolution reality displays.

Open questions

If no experiment could ever distinguish a simulated universe from a "base" one, has the argument stopped being the kind of claim philosophy or physics can adjudicate at all? Does proving Rule 110 universal make Wolfram's hypergraph more likely to be physics, or only more likely to be capable of it, pending some independent reason to believe the update rule? And is compatibilism actually answering the free-will question, or quietly redefining "free" until the harder version of the question can no longer be asked?

Go deeper

• Seth Lloyd, Programming the Universe: A Quantum Computer Scientist Takes On the Cosmos, Knopf, 2006.
• Nick Bostrom, "Are You Living in a Computer Simulation?", Philosophical Quarterly 53, no. 211 (2003): 243–255 — simulation-argument.com/simulation.pdf; Matthew Cook, "Universality in Elementary Cellular Automata," Complex Systems 15, no. 1 (2004): 1–40.
• Computer Simulations in Science, Stanford Encyclopedia of Philosophy.