No, we’re not living in a simulation. Here’s why.
The idea that we might be living inside the simulation of some more advanced future civilisation sounds like something a stoner might dream up on a Saturday night after watching one too many sci-fi movies. Yet in recent years it has been floating around the internet as an ostensibly serious and respectable intellectual position, especially among tech utopians like Elon Musk. In my view, not only is the logic supporting this outlandish conjecture full of holes, implausible assumptions and contradictions, but the whole idea is pernicious and should be resisted.
The argument originates with Oxford University philosopher Nick Bostrom. In a 2003 paper titled “Are You Living in a Computer Simulation?” Bostrom argued that a sufficiently advanced civilization might be able to simulate consciousness. Given the rapid advancement of technology, especially in video games and virtual reality, it’s conceivable that future technologies could run highly detailed simulations indistinguishable from reality. Since these simulated civilizations would also eventually advance enough to simulate worlds themselves, then the number of simulated universes would end up vastly outnumbering the original single reality. Thus, he argued, it is overwhlemingly probable that we are living in a simulation rather than “base reality”.
This argument is often bolstered by claims based on elements of quantum physics. For example in this Medium post by Paul Pallaghy, the author argues that quantum entanglement effects “virtually prove we are living in a simulation”. The fact that quantum correlations occur instantaneously over arbitrary distances is, according to him, “highly consistent with the universe being a simulation. And not much else.” He continues: “The game source code — quantum mechanics — determines the outcome of correlated outcomes light years apart. If it’s in code, it’s then irrelevant how far apart these correlated objects are.”
There is a universal human tendency to cast the universe in the image of our own core metaphors and obsessions. Feudal Europe, with its rigid hierarchies, projected this same social structure upwards into the spiritual world, imagining a “chain of being” extending down from God the Father — the ultimate feudal lord — through angels, kings, popes, greater and lesser lords, all the way down to lowly serfs and finally poor old animals, labouring at the bottom of the stack like Yertle’s turtles.
In the Industrial Revolution, scientists reimagined the world as a giant machine following Newton’s laws, a clockwork universe. If God had a place at all, He was relegated to the role of industrialist. He was the manufacturer of the machine, the Prime Mover who set the whole thing in motion. It was a view of the world that directly reflected the increasing domination of the natural world by industrial machinery.
It’s no surprise, then, that in our contemporary world, with its ubiquitous computers and digital simulations, we’ve started to recast the world as a kind of VR experience or video game. It’s as if we’re trying to squeeze ourselves inside the devices that we stare into day in and day out, completing, in a sense, our fantasy of control over reality. We might not quite be able to magic the world into any form we want it, but our descendants can. Life is a game in their computer. Hooray!
That this is a fantasy should be made clear by examining the logic of the argument. Let’s start with Bostrom’s. To describe his reasoning as “speculative” is an understatement. I will ignore his initial supposition that consciousness can be simulated (actually, created) in a computer, even though we do not understand consciousness at all. Bostrom imagines a Russian-doll type situation, where simulated civilisations end up creating their own simulations, which then advance far enough in their development to repeat the feat and so on, without end.
The problem with this should be obvious to any computer programmer. It’s referred to as a “stack overflow”, and it’s a common enough programming mistake that the biggest website for peer coding assistance is named after it. If a piece of computer code runs itself — a situation known as “recursion” — as in the imagined Russian doll simulations, it requires more memory to be allocated each time. If a limit is not imposed on this process, the computer will rapidly exceed its memory limits and crash. This can’t be avoided by invoking some vague notion of superior future tech. Computers are physical devices and cannot transcend the inherent limitations of their memory.
The technical difficulties are not limited to the problem of infinite recursion though. Currently physicists are incapable of accurately calculating the states of even a single complex atom. (I’m not sure exactly how far they’ve gotten in the periodic table now, but it’s not far.) The idea that a whole universe can be simulated to such a level of verisimilitude that its inhabitants can make arbitrarily detailed measurements and never notice a flaw is a stretch, to put it mildly.
Furthermore the argument ignores the physical limits which nature places on computing devices. Moore’s Law — the observed doubling of computational power every eighteen months which held throughout the early years of computer development — is a historical phenomenon. It is not a law of nature. Computers cannot exceed the inherent information processing limitations of matter itself, and a small classical or quantum computer can never simulate a larger system than itself with perfect fidelity. In fact, the only “computer” with the amount of information required to simulate a universe is… the universe itself.
When this point is made, typically the proponents of the simulation hypothesis argue that the simulations don’t need to actually simulate things down to the atomic level. Like in a video game, only the observed parts of the world need to be “rendered”, and in doing so, many computational shortcuts can be employed. You don’t need to calculate the whole state of every atom that makes up the moon. You can just calculate the observed parts to whatever resolution is required. The galaxies can be more or less “painted on”.
Sigh. So these people are saying that the vast size of the universe is just an illusion set up for the game? That only in “base reality” do all those galaxies in the Hubble photographs with their billions of stars and planets actually exist? Really? OK, if you say so. But again there’s a problem. When computational shortcuts are employed, inconsistencies inevitably occur. In games, the simulations inherent limits are protected from being exposed because players don’t have unlimited freedom to interrogate their surroundings. You can’t get a microscope and examine the table in front of you, or get on a plane and fly over the oh-so-convenient mountain range that bounds the world to see what’s out there. We, however, are free to conduct any experiment or observation we choose.
The simulation programmers can’t know in advance what observations the simulation inhabitants are going to make. This is a law of computer science. You can’t predict the result of an arbitrary computation any other way than by running it. Therefore to ensure that all the possible observations of all the “sims” in the game will always be consistent with one another, the simulation has to compute things to a very deep level — one that only gets deeper as the sims’ scientific capabilities improve.
Despite this the world appears incredibly consistent in every detail. Of course, it’s possible that the program erases our awareness of the inconsistencies. But as far as arguments go, this really is a disgraceful last resort. It might be true, but as Descartes realised in his famous reflections on what it is possible to doubt, we also can’t know that we aren’t manipulated into believing the world is real by invisible demons. We can admit that it is impossible to prove we aren’t inside a manipulated reality, but Occam’s Razor tells us that that’s a poor reason for believing it. It’s simpler and therefore more rational to believe that that waddling feathered thing is a duck than it is to believe it is an ingenious robotic duck simulator.
It’s when quantum mechanics gets trotted out in support of the simulation hypothesis that things really get silly, however. The argument runs that a number of features of quantum physics tend to lend to support to the idea that reality might be simulated. For example, the quantised nature of matter and energy might be a consequence of the limits of the underlying computer, similar to the pixels on a screen. The observer effect, where quantum systems only assume definite values when observed, might hint at computational shortcutting. And entanglement effects might be the result of code simultaneously setting values at separate locations. Or something.
Again, fatigue — and the growing length of this article — prevent me from iterating all the problems and misconceptions inherent in this vague, impressionistic reasoning. But I don’t need to get into the nitty gritty of quantum mechanics to discount these arguments. Let us just consider this. If quantum physics are a side effect of simulation, then the physics of base reality cannot be quantum. Otherwise the argument would be self contradictory, since base reality is by definition not simulated and quantum effects could therefore not be an argument for simulation. However, once the concession is made that base reality does not operate according to quantum physics, the argument has fallen into self-contradiction anyway.
If base reality does not operate according to quantum physics, what physics do apply there? Newtonian physics perhaps? That is ridiculous. Newtonian physics emerge out of quantum physics and don’t explain the way the world works at all. Electrons collapse into the nuclei of atoms in a Newtonian world. Quantum physics is the basis for everything in chemistry, and therefore in biology too. It is simply the basis of the entirety of the reality we perceive. This world cannot be a simulation of a world with totally different, unimaginable, non quantum physics!
A simulation is by definition a likeness, but this universe cannot be a likeness of any universe that does not operate according to quantum physics, because such a world would look nothing like ours. Certainly it can’t be a simulation of the past of our descendants (or of the descendants of the people we putatively simulate), which was Bostrom’s original idea.
Quantum physics is the only physics we know of, and any other laws according to which base reality might operate are not physical laws in the sense we understand that word. So for all we know base reality might be the world of the Olympian gods and goddesses. Or maybe we’re the dreams of a purple troll in Xanadu. Whatever the case, it becomes hard to call our reality a simulation, since it’s unlikely that our concept of a computer has any meaningful correlate in a world without something as basic as atoms. We are back in the realm of theology, just theology with a misleading scientific gloss.
Yes the whole argument is a preposterous flimflam of tech-bro hubris, wishful thinking, projection, and confused quantum woo-woo. It’s silly; it’s unbelievable. It is also, as far as alternative ontologies go, harmful. It devalues people, nature and reality itself by relegating everything to the status of digital assets in a giant game. Never mind climate change; it’s all just a simulation! It pushes us further into the digital hypnosis that grips our society, convincing us that reality is image and image reality, an ideology of surfaces that is deeply problematic, not only for its fundamental superficiality, but because it offers a false refuge from the very real crises our civilization faces. It tells us that when we finally look up from our screens, we’re still only looking at one. And that is not only delusional, it’s a cop-out from the responsibility of deep care that we owe one another and the world.
— — — — — —
POSTSCRIPT:
I’ve enjoyed the commentary this piece has engendered. I decided to add a somewhat more detailed discussion in this postscript for those who have read Bostrom’s paper and may be interested.
A point that many commentators make is that my critique regarding plausible future technology is based on my current limited understandings, and that I can’t know what capabilities we may have in the future, including the discovery of new physics etc. This is true, but what I am attempting to do is question assumptions which Bostrom seems to see as self-evident but which absolutely deserve to be challenged. This is the key weakness in Bostrom’s argument.
In order to evaluate our real best guess at the probability that we are in a computer simulation we can’t just look at the probability that we are in a simulation given that advanced civilizations can and will create such simulations. Rather we must evaluate our level of certainty about a wide range of assumptions on which the scenario of massive scale ancestor simulation depends. Thus we need to assign probabilities to the such questions as:
- What is the probability that civilization progresses into a future that has greatly advanced technology, but that this technology does not rise to the level required to simulate the experience of billions of individuals living their lives to a degree indistinguishable from real life in every detail?
- What is the probability that, given such tech does end up being possible, the civilization decides to devote the massive resources required for running these computers to something other than billions of ancestor simulations?
- What is the probability that future civilization possesses the technology and the desire to simulate ancestors, but only has the resources or need for a single simulation, maybe of only a thousand or a million individuals? In other words, some simulation occurs, but not at the extreme scale required to outnumber “base reality” life experiences.
- What is the probability that future civilizations, however advanced, continue to have resource constraints that force them to allocate their computing resources to more pressing concerns than historical simulations?
- What is the probability that future civilizations with extremely advanced computing power would see no more value in simulating ancestors than we would in running tic-tac-toe simulations ad infinitum?
- What is the probability that the deep future looks completely different from anything we can project based on current technology, understanding and experience and simulating of ancestors is simply not a relevant possibility?
Now, having assigned probabilities, justify your numbers based on scientific evidence or first principles reasoning.
The fact is that these are all imponderables. Yet Bostrom makes it seems like he more or less knows the answers. We could invert the reasoning here and ask, what is the probability that human beings five hundred years in the past could predict how we would be using technology today? We know the answer to this question at least. It is zero. They could not and did not predict it. They could not begin to dream of it. How then are we in a different position with regards to a future orders of magnitude further away, when change is accelerating?
Bostrom has picked out one tiny possible future and, by clever but deceptive reasoning, magnified that world so that it appears far more likely than it really is. He has conjured into existence an imaginary world of planet-sized computers with unlimited resources for historical simulation games which he can’t compellingly explain why anyone would want to run, and by placing it in the reader’s mind, supported by an assumption of infinite technological progress, has made seem almost plausible. In doing so, however, enormous gaps in evidence and reasoning have been swept under the rug. He has summoned out of thin air unlimited powers to create worlds, and then used them to create a very specific future world that he then claims proves something about the nature of our reality now. It does not.
Finally, in response to those who argue that I can’t know the true nature of reality, of course I don’t. But I am not trying to argue that there’s nothing more to the world than meets the eye. I believe we are embedded in an infinite field of relations that is in principle unknowable in its absolute nature. This is a topic I am currently writing a book about, and intend to write more about on Medium. However, this is not an argument for Bostrom’s Simulation Hypothesis. If anything, it is simply an argument for humility in the face of the great unknown.