The Hypothesis
The concept of a computer simulation is familiar enough to the modern reader. It is a model world built by a computer scientist to test his or her theories of meteorology, the spread of diseases, economics and so forth. The proponent of the Simulation Hypothesis begins by supposing that there are no limits to the development of this technology: It may be that our scientifically advanced descendants will be able to build and run simulations that replicate life on Earth with exhaustive accuracy—digitally reconstructing not only the atomic composition of every object on Earth but also the neurological structure of every human brain. And this, they suggest, has the unsettling entailment that the postulated simulation might include a simulated but conscious version of you and me.1
Present day simulations single out a particular natural phenomenon for analysis. What possible purpose could such unconstrained simulations serve? Westerhoff offers a suggestion.2 We often wonder how history might have turned out if some small but crucial detail of the past had been different. What if, say, Mao Zedong had died of a heart attack during his famous swim across the Yangze River? To us such questions are unanswerable. But perhaps not to our descendants. They could (so the theory goes) run a simulation of Earth between 1875 and 2018, a simulation that matched to history at every point with one exception: Mao Zedong dies on July 16, 1966.
And here, claims Westerhoff, arises a still more unsettling possibility: The possibility that we are living in one of these simulations; the possibility that, say, Mao Zedong did die in 1966 and the architects of the simulation are interested to see how human history would have turned out had he lived. Nor should our ignorance of our unreality come as a surprise: Since the historical persons on which we are modelled did not believe they lived in a simulation, nor do we.
It is a wild hypothesis. But if we are willing to indulge for a moment its key presuppositions, it also has a certain probabilistic force. And this is because there is in principle no obvious limit to the number of simulations our descendants might choose to run. It is not unreasonable to suppose that they would run tens of thousands or even millions of simulations. And in that case the 100 billion actual humans who have ever existed on Earth might comprise a tiny fraction of the sum total of conscious beings—simulated and actual—who have ever existed. And in that case the probability that you are a simulated human being is on balance greater than the probability that you are an actual human being.3
On the face of it the Simulation Hypothesis (like Last Thursdayism, like Berkeleyan Idealism) would appear to be undisprovable: Faced with any datum advanced against the hypothesis, the proponent could claim that that datum is also part of the simulation. If that were how things stood the hypothesis would still be rationally unaffirmable.4 However, I am now going to argue that the Simulation Hypothesis is demonstrably false.
A First Pass: Westerhoff
Westerhoff himself considers an argument against the scenario.
Since the computer supervening over the simulation could not be infinite in its computational resources, there is, he says, a regress problem for any simulated world that can run its own simulations, which simulations, ex hypothesi, could run their own simulations in turn, and so on, infinitely. And so any world in which simulations are possible or even an accessible concept is probably not itself a simulation: The architects of the simulation, if they exist, would need to calibrate the program to avoid this scenario.
A Second Pass: Natural Theology
Westerhoff’s counterargument is of limited force. A proponent of the Simulation Hypothesis could just postulate that there is some undiscovered constraint in the simulated physics of our universe which prevents an infinite regress of simulations.5 But this need not trouble an opponent of the hypothesis which faces far more serious difficulties.
The first arises from Natural Theology. Since in its usual formulation, and also by definition, the Simulation Hypothesis imagines that our observable universe is a simulation of an actual universe, even allowing that we are in a simulation fails to discharge all the traditional arguments of Natural Theology. We can therefore argue of the actual universe (which we observe in the mirror image of the simulation) what philosophical theologians have always argued; namely, that Theism is an inference to the best explanation for the ex nihilio origination of material reality, the fine tuning of the laws and constants of physics, the origin of life and human mental and moral experience.
What possible relevance does Natural Theology have to the Simulation Hypothesis? It might be argued that the Simulation Hypothesist could simply set the question of the existence of God to one side. Its relevance is this: If these arguments obtain and God exists there are good a priori grounds for believing that a perfect moral agent would not allow fallible moral patients to themselves become moral agents over other moral patients in an unlimited and unconditional way—in the manner of Roko’s Basilisk. So the proponent of the simulation theory has an insupportable burden of proof to shoulder. To make his argument plausible he must prove that God does not exit.
Here, as a last resort, a Simulation Hypothesist might deny that our simulated universe bears any meaningful resemblance to the actual universe. The philosophical cost of this reply is high (since it would greatly attenuate the grounds for postulating the hypothesis in the first place) and profits him not at all. For the most forceful and indefeasible argument for the existence of God is the Cosmological Argument which obtains so long as a single finite and contingent particular is observed.6 And since the simulated universe, if it exists, is itself a contingent and finite particular, the Cosmological Argument obtains even if we allow that we can know nothing at all of the actual universe. As before, to make his argument plausible, the Simulation Hypothesist must first discharge the most forceful argument of Natural Theology in order to prove that God does not exit.
The Death Blow: The Hard Problem
The foregoing difficulties are considerable. But they are trivial compared to the central problem with the Simulation Hypothesis.
In postulating conscious minds that exist in a computer, the hypothesis presupposes that consciousness can be instantiated in a physical substrate and thereby presupposes in turn and without argument a solution to the so-called Hard Problem of consciousness. In fact, it can be shown that mental states are irreducible to the physical in principle. A key feature of the hypothesis is therefore falsifiable. To warrant serious attention, the proponent of the Simulation Hypothesis must first complete an insurmountable task. He must solve the Hard Problem by demonstrating how mental states (intentionality, qualia, first person ontology) are susceptible of reduction to the physical.7
The Simulation Hypothesis has a certain grip on the popular imagination— perhaps in particular among a generation who have grown up playing computer games. It does not, however, stand up to careful scrutiny.
————————————————————
[1] It is worth noting a problem that arises with the Simulation Hypothesis right out of the gate: Recovering enough information about persons long-dead to simulate them is fundamentally impossible since most of the information would have been dissipated as heat and radiated away from Earth at light speed. No finite computing power, however powerful, could complete the task.
[2] Jan Westerhoff, Reality: A Very Short Introduction.
[3] Others, while entertaining the outlandish hypothesis, are more conservative in their probabilities. David Chalmers has estimated the probability that he is living in a simulation at 20 percent.
[4] Hoffman and Rosenkrantz take the view that if something belongs to a universal and commonsense ontology, “then there is a prima facie presumption in favour of its reality. Those who deny its existence assume the burden of proof.” Swinburne has formalised this idea into a basic principle of epistemology which he calls the Principle of Credulity: We should, in the absence of evidence to the contrary, believe that things are the way they seem to be. An unprovable and undisprovable hypothesis that conflicts with our universal and commonsense ontology is therefore to be dismissed on pain of irrationality.
[5] On this view the discovery of such a constraint would provide inductive evidence that we are in a simulation.
[6] See The Cosmological Argument.
[7] See the Argument from Consciousness summarised here in five parts: Qualia, Intentionality, Privileged Access, Nonphysicality and the Conclusion.