What if your thoughts and everything you perceive are nothing but bits in a computer simulation designed to satisfy the curiosity of scientists with capabilities far beyond anything known to human beings?
Maybe you’re thinking, “I already saw that movie.” Or, “What’s the point in speculating on some abstract philosophical theory that we can never test anyway?” Or maybe you just think it all sounds pretty far-fetched.
But some philosophers are taking this idea, called the “simulation argument,” very seriously. Physicists have gone even further, suggesting that we might even be able to detect evidence that confirms it, if we know where to look.
In 2003, philosopher Nick Bostrom of the University of Oxford made the first rigorous exploration of the simulation argument. The simulations he considered are different from those in movies like “The Matrix,” in which the world is simulated but the conscious minds are not—that is, where biological human beings with human brains interface with the simulated world. In Bostrom’s simulations, human consciousness is just another figment of the simulation.
Bostrom assumes that the human mind is substrate-independent: that human consciousness isn’t strictly dependent on the biological brain itself, and that if we could physically replicate that brain in sufficient detail in another form (such as within a computer) it would also have the subjective experience of consciousness. The replication doesn’t have to be perfect. It just has to be good enough that the replicated being has a human-like subjective experience (a “mind”). An advanced civilization with sufficient computing power to pull this off would be classified as “posthuman.”
What is the probability, then, that we ourselves are simulated minds?
To calculate the probability that a randomly-selected human-like mind—let’s call it “you”—is a simulation, you would divide the number of simulated minds by the total number of all human-like minds (both simulated and non-simulated, or ”real”).
It might seem that there’s no way to make sense of these quantities, but keep in mind that the “simulated minds” in this case come from a posthuman civilization running detailed simulations of its own past. The total number of “simulated minds” will be a multiple of the “real minds” of the humans that existed before they reached posthuman status. This multiple will be the average number of simulations run by the society (although this argument doesn’t rule out multiple human-like societies existing, if you’re a fan of the somewhat-similar Drake equation). So if you divide both the numerator and denominator by the number of “real minds” (even though we have no idea what that number is), you reach the following:
Now we get to play with the numbers to see what happens. If the total number of simulations is very small, the ratio is very small. But if the total number of simulations is very large, the ratio will be close to one. Arguing from some reasonable assumptions about what drives the number of simulations, Bostrom explains that we can expect at least one of three scenarios:
1. The fraction of civilizations that survive to the posthuman level is very small.
2. The fraction of posthuman civilizations interested in running simulations is very small.
3. The probability that you are a simulated mind is a simulation is very high.
If options 1 and 2 are correct, we can relax: we’re probably real. But if you think that many civilizations do survive to become “posthuman,” and that many of those posthuman civilizations are indeed interested in running simulations, then option 3—that you are just a computer simulation—becomes a serious probability. It’s hard to tell, of course, as we have no direct experience with posthuman civilizations and their preferences, and there are a number of philosophical objections worthy of debate: Why would any posthuman civilization want to do this? Are Bostrom’s assumptions actually reasonable, or do they miss something vital? Is substrate-independence true, or is it impossible to replicate a human mind?
Can physics offer any insight here? English cosmologist John D. Barrow addressed this question in a 2007 essay published in the book “Universe or Multiverse?,” in which he argued that the simulations might have limits. Even if posthuman simulators “have a very advanced knowledge of the laws of Nature, it’s likely they would still have an incomplete knowledge of them,” wrote Barrow. Any flaws or gaps in this knowledge “would of course be subtle and far from obvious, otherwise our ‘advanced’ civilization wouldn’t be too advanced.”
If these gaps exist, as Barrow reasons, the result would be either glitches in the working of reality, or update “patches” to fix a glitch before it causes a problem. (Recall that in “The Matrix,” local changes to the Matrix caused déjà vu.) These patches could result in changes, over time, to the laws of nature. Barrow concludes:
[…] if we live in a simulated reality we should expect occasional sudden glitches, small drifts in the supposed constants and laws of Nature over time, and a dawning realization that the flaws of Nature are as important as the laws of Nature for our understanding of true reality.
An un-peer-reviewed 2012 attempt at a more rigorous physical analysis of the situation, “Constraints on the Universe as a Numerical Simulation” by physicists Silas R. Beane, Zohreh Davoudi, and Martin J. Savage, reached the conclusion that “in principle there always remains the possibility for the simulated to discover the simulators.” Their specific prediction was that there might be limitations on cosmic ray energy levels if we live in a simulation. However, the above statement might be overstating their own case: rather than discovering the simulators, if we find that cosmic rays violate these limitations, we will have instead disproved that we’re in a simulation!
This paper also predicts why our posthuman descendants might want to simulate our universe: to test out string theory. Currently, string theory is overpowered by a vast landscape of possible versions of string theory, and scientists haven’t figured out which one might describe our universe. Detailed simulations would allow posthuman races to test hypotheses about these universes, ruling out possible versions of string theory to zero in on the one that describes their real universe.
If this is the case, Barrow must be right that even the simulators are working from incomplete knowledge. A civilization that is simulating the universe to explore the string theory landscape must not know everything about the laws of physics, and therefore we can reasonably expect gaps and flaws in the simulation.
Indeed, these ideas suggest that our entire known universe is itself only a small part of a grand experiment to understand the most fundamental mysteries of the universe. And for the scientifically-minded among us, many may find that a worthy purpose to our simulated creation.
Editor’s picks for further reading
Are You Living In a Computer Simulation? The Simulation Argument
Find an extensive list of articles, videos, and other media discussing the “simulation debate” at this site curated by Nick Bostrom.
Backreaction: The simulation hypothesis and other things I don’t believe
Sabine Hossenfelder explains why “the simulation hypothesis is not among the things that keep me up at night.”