‘Simulation Theory’ is popular lately so I am building a new summary piece on it. The following article repeats material elsewhere on my site but in scattered places, and with some new and connecting material, to provide a thorough and current treatment of the question.

-:-

Are we and the universe just a giant computer simulation run by aliens in some external meta-universe? It’s the new craze to think so. It’s worth thinking about; but recent enthusiasm for it far exceeds its probability. Simulation theory isn’t new. It’s just a techno-redux of various forms of Idealism or even the Hindu-Buddhist concept of Maya. The problem all have faced is that they fail to explain the specific complexity and behavior of even human consciousness much less the world outside it (apparent or not). They all “handwave” away this problem, but that handwaving conceals a fatal fault in the theory when expressed in terms of epistemic or indeed even existential probability. The pre-tech versions suffer all the defects of the specified complexity problem of supernaturalism generally (see The Argument from Specified Complexity against Supernaturalism and The God Impossible). The modern version replaces that defect with an analogous one: it has to posit two entire universes (and in a specific causal relationship) just to explain one. But those aren’t the only problems.

The Cartesian Demon Problem

The relationship between modern and ancient “sim” theories is instructive. Because it reduces everything to the same question. Put simply, “How do we know that what we see around us is the real deal, and not some grand illusion perpetuated by an unseen force?” And not in the sense that consciousness itself is a sim, and so it represents the outside world with a functionally corresponding illusion—we all agree that’s the case (see What Does It Mean to Call Consciousness an Illusion? and the Barkasi-Sant’Anna Defense of Naive Memory Realism). Rather, we mean in the sense that even the external world is an illusion—that it literally does not exist as represented, but is merely communicated to our mind as if it existed.

This would include our minds themselves, which in this case are not likely produced by the classic 1950s sci-fi imagining of “brains in a vat” (amusingly reimagined in The Matrix), but there wouldn’t likely even be brains, per se—just the sim. Everything, ourselves included, would just be computer code processed in a grand computer circuit. But apart from the convoluted metaphysics, this is the same thing as Maya and any other mind-makes-world Idealism. Which reduces all versions of this thinking to one common principle: they are all Cartesian Demons.

In other words, some analog to René Descartes’ evil demon would be the constructor of us and all we survey. In sim theory, that means an inscrutable alien race or AI running some equivalent to The Matrix (sans bodies in tubes). Some such Cartesian Demon argument has been around for a long time. It suffers from being far less probable than a more mundane scientific realism. Because it requires vastly more complex explanations…for which there is exactly zero evidence. In that sense, this argument is basically just a secular version of theism. “There is no evidence for the God that you are proposing explains all reality.” “Well, of course not, he hides all the evidence. Because he’s super-amazing.”

If you didn’t catch the fallacy there, let me walk you through it. Sim theory is just another “Satan placed all the fossils” idea (just replace “Satan” with “aliens”). The theory by itself contradicts vast amounts of observations, which ordinarily constitutes a decisive falsification of a theory. This is especially the case for the most basic version of this argument, which is solipsism, wherein only you exist and your mind is making everything up, and that’s everything, up to and including the notion of you even having a brain (see my takedown of that hypothesis in my Critique of Rea; it performs poorly as a predictive model). But the same problem obtains for external simulators. You have to “explain away” all the evidence against that, with convoluted hypotheses as to why what we expect to see doesn’t show up—including why the sim managers hide from us, and why the sim is so objectively pointless and poorly designed (and literally every other argument against God you can think of).

For instance, if reality is engineered by a conscious Mind, we should expect reality to conform to its whims. Reality would then look like some externally controlled simulation. Including values-directed laws of physics, like “survival of the kindest,” or the appearance of helpful angels, or respawns after accidental deaths. Even if the Mind is evil or has goals uninterested in our welfare, we should see comparable manipulation and engineering of the universe in aid of those goals. Which is why sim theory enthusiasts always have to gerrymander a bizarre goal to fit our actual observations, which is as defective an epistemic procedure as Ptolemy inventing epicycles to keep the solar system geocentric, or—again—any apologetic for God you can think of (starting to notice a theme here?).

Whereas a universe that isn’t being directed, should only exhibit the same mindless goal-directed behavior all mindless universes would. For example, all natural universes of any sufficient random complexity will inevitably manifest evolutionary algorithms, and equivalents to the nucleosynthesis cascade, crystallization, autocatalysis, and so on—which require no intelligent programming because they automatically fall out of any complex random system: hence the the New Wong-Hazen Proposal Refutes Theism as well as sim theory. What we observe are phenomena that are inevitable given the starting conditions, and require no engineer or external controller to cause or maintain. To evade this falsification, defenders of God, er, I mean, The Simulator, have to invent a vast array of “excuses” to explain all that evidence away. They basically invent a Deity that conveniently, for some really weird reason, only designed the universe to look exactly like an undesigned universe. The world looks like a world with no engineer—sim or super.

The unavailability of this excuse for solipsism is what proves it false (our mind cannot simultaneously be the Cartesian Demon and its victim). But the availability of this excuse for God, er, I mean, external Simulators, still does not rescue them from being massively improbable. The immediate problem is that Simulationists don’t realize how vastly complex their theory is. They think a “Mind” is a really simple thing. When in fact it’s actually the most complicated thing in the universe. And if you are positing aliens or AI, you aren’t just positing a Mind. You are positing an entire additional universe, and a complex convenient history for that extra universe explaining the appearance of those aliens or AI and their development and running of this particular universe in it as a sim. Which collectively is actually more complex than this whole universe.

That’s right. Due to the laws of computation, a simulated universe must necessarily always be less complex than the mechanism simulating it. Ergo, Simulation theory entails positing an extra universe more complex than the universe you are positing it to explain. In truth, the simplest possible way to simulate a universe is to just create the universe. Simulation hardware is a waste of time and resources. No extra hardware or computational architecture is then required to store and process all the data—much less “store and process” the aliens and whatever else must exist in their universe to explain how they exist at all, and have the resources (and historically gerrymandered motives) to simulate our universe within theirs. Plus there will be no risk of bugs or errors, which is a bonus. But it is in fact a universal law of computation that the simplest simulation of a thing is the thing itself (see chapter nine of Labyrinths of Reason). You only put it in a computer if you need to strip away almost all of its contents to study a single phenomenon; or to have the power to mold and change it at will. Yet we clearly don’t live in either scenario.

So proposing an even more complex thing (on no basis of any evidence) to explain our already-absurdly-complex universe is generally going in the wrong direction, scientifically speaking. But even logically speaking it entails positing something less probable than the random formation of our universe without the Simulator. Because you have to presume the random selection or formation of that other, more complicated universe, even to get to the possibility of it simulating ours. So you have the same improbability of a natural origin of our universe times the added improbability of that other universe leading down a path to inspiring anyone to create ours. The resulting combined improbability is always going to be less than the probability of just being in that first universe to start with. Of course, to get that second improbability that you have to multiply by the first one, you have to presume an even more unusual collection of contingent events within that universe, to explain why anyone in that universe would come to exist who would want to simulate our universe—by which I mean not just any universe, but our universe, in the form and function we observe it has (a crucial point I’ll get back to); and then you have to add in the improbability that of all universes we could be in, we just happen to be in that one. So the prior probability of any Simulation hypothesis is astronomically small. Indeed, beyond. Origin theories like Chaotic Inflation are vastly simpler, and rely on posits that already have evidence supporting their existence or likelihood. So they’re much more likely—which means, in terms of frequency, most universes will arise that way, not the other, convoluted way (don’t worry, I’ll be getting to Nick Bostrom’s attempt to get around this problem).

Another way to look at the Cartesian Demon problem is to force yourself to think through what it actually requires. And realize that each peculiar thing you have to imagine adding on only reduces the prior probability of those features having ever come together. Cartesian Demons simpliciter entail all sorts of observable evidence for their existence, because they would just obviously be controlling everything like a cat with string, which we’d all notice, all the time (think Q or that kid in The Twilight Zone); you have to make the thesis more and more complicated to get more and more of that missing evidence explained away, building a conspiracy theory around its exhausting effort to hide its very existence and manipulations from us. As I said in Epistemology without Insurmountable Regress or Fallacious Circularity, the weakest Cartesian Demon is your friends pranking you (a somewhat more convoluted and thus more complex explanation of what you are experiencing than that you are just experiencing the actual state of the world right now); a much stronger but still weak Cartesian Demon is The Truman Show (which is far more complicated in terms of the system required to realize it); a far better Cartesian Demon is The Matrix (which is far more complicated still, in terms of the system required to realize it); so to get even better than that (even all the way to a perfect Cartesian Demon) requires a vastly more complex hypothesis than even that (necessarily—as I discussed before in The God Impossible). And that’s all just to construct and describe its powers and motives and how and why it has them and never fails at them, thus successfully hiding the fact that we are in its clutches.

Worse, you then also have to still propose a whole extra universe on top of the Cartesian Demon anyway, in which the Cartesian Demon can exist and which makes its powers realizable. Just skipping the vast added complexity of the demon and sticking with the only universe you need in the explanation anyway is by definition far simpler. This is why Cartesian Demons are simply too improbable to credit on present undeniable evidence (particularly the undeniable present experience of all these logical facts). By which I mean epistemically improbable, of course. We may well be brains in vats or whatever technological equivalent. But we have no reason at all to believe that’s in any way likely. Mediated perception of an external reality more or less accurately modeled by our brain is just a far simpler and thus epistemically more likely explanation of how these perception-events are occurring as they are. Adding a whole other universe, and a whole other weird scheme on top of that, is just too many epicycles to credit. Multiplying entities beyond necessity.

So we can be pretty sure we aren’t in a Simulation. Not absolutely sure. But pretty sure. Our epistemic status warrants no belief that we are, to any appreciable or pertinent degree. Hence it’s actually a fairly solid scientific conclusion that we are not. “We are in a Simulation” is pseudoscience, just like “Alien lizard people control all the world’s governments.” So, far from being “unsolvable,” this question is pretty well solved. Not solved in the sense of “we can be absolutely certain we aren’t in a Simulation,” but neither is anything else in science. “We can be absolutely certain water is made of hydrogen and oxygen” is also false. But no one concludes from that trivial truth that we haven’t solved the problem of what water is made of.

Such is the actual state of this question. But not for lack of trying…

The Nick Bostrom Problem

In an effort to get around all those problems, Nick Bostrom developed the Simulation Argument. Which relies on something new we’ve realized: there very likely are countless other universes; so positing them is not ad hoc anymore. But his argument from that to the conclusion that we might be in a simulation suffers from one fatal flaw: a hidden false premise.

The Bostrom argument starts with the premise that if every “posthuman” civilization in every universe repurposed all the matter in its universe to run sims, then the number of sims would exceed real universes, and so the probability we’d be in a sim is greater than that we’d be in one of the fewer real universes. The argument depends on this premise, take note. Because it trades on an equivocation fallacy over the word “sim.” Otherwise the argument can be summarized fairly elegantly as:

A technologically mature “posthuman” civilization would have enormous computing power. Based on this empirical fact, the simulation argument shows that at least one of the following propositions is true: (1) The fraction of human-level civilizations that reach a posthuman stage is very close to zero; (2) The fraction of posthuman civilizations that are interested in running ancestor-simulations is very close to zero; (3) The fraction of all people with our kind of experiences that are living in a simulation is very close to one.

This statement is entirely correct and true. In fact it’s a really good example of analytical philosophy nailing a point. But note that, as stated, this does not declare (3) true or probable. It just says, the only way (3) can be false, is for either (1) or (2) to be true. Formally, Bostrom claims all three must be equally likely, so far as we know; informally, he always shills for (3). But he’s wrong. We do know (1) is likely to be false; civilizations are far harder to kill than the Bostroms of the world think (see Is Society Going to Collapse in 20 Years? and Are We Doomed?). Existential “threat worriers” far too quickly conflate “collapsing civilization into a reset” with “wiping out the human race.” Civilization progresses no matter how often it gets reset. Rome fell. The West collapsed in ruin (Yes, the Dark Ages Really Were a Thing). But just over a thousand years later we were advancing beyond its every achievement. We just don’t have any way to kill civilization. We can collapse one. But that’s not the same thing. And given hundreds of thousands of years, collapses are mere blips. The net progress of civilization is assured. Meanwhile, I just surveyed all the reasons why (3) is likely to be false. What we have left is (2). And alas, we do not have any reason to believe (2) is unlikely at all; in fact, it is almost certainly true—at least to a frequency that will swamp and thus scuttle Bostrom’s entire reasoning.

This is an example of one of the most common mistakes philosophers make: not thinking things through. Bostrom argues that “if (2) is true, then there must be a strong convergence among the courses of advanced civilizations so that virtually none contains any relatively wealthy individuals who desire to run ancestor-simulations and are free to do so.” What’s unlikely here is “individuals who desire to run ancestor-simulations.” Also notice this first equivocation fallacy: his argument requires all the resources of the entire universe to be policed-up into running ancestor sims, but now for some unexplained reason he’s talking about one dude running a few sims. The question is: will an entire civilization repurpose all the resources of existence to this; not “will some whacko run a weirdo sim on occasion.” Because his argument needs there to be more “ancestor sims” than real universes, remember? Also remember that even a single simulated universe must be smaller than the universe it is being simmed in; it is therefore not likely a single whacko can run a sim of this scale. They would need to be Emperor of the Universe and command the entire resources of it—without anyone shutting that evil fucker down. These are tall improbabilities.

The desire to use a universe’s entire processing capacity to run games and paradises would always overwhelm this strange niche interest—which exposes the key equivocation fallacy in Bostrom’s argument: he snuck in the word “ancestor-simulations,” and then trades on the possibility of a universe-encompassing mega-civilization running just any simulations. But that’s precisely the problem. There actually isn’t any use for “ancestor” sims. And the process of willfully recreating endless horrors, from holocausts to wars to plagues and everything else, would be irredeemably evil. Such monstrosities are unthinkable to any species ethical enough not to destroy itself before acquiring the ability to make them. In other words, unlike options (1) and (3), option (2) to be false requires an extremely unlikely conjunction of events: inexplicable goals plus extraordinary villainy. (2) is therefore far more likely than (1) and (3). There will be more sims than real universes; but almost none of those sims will look like our world, so we know we aren’t in one.

Bostrom acknowledges that running ancestor sims is unethical. In fact, it’s so massively unethical it would require the most evil fuck imaginable even to contemplate running one. And we know a lot about evil fucks. They don’t do things they don’t benefit from. That’s why they’re evil. Why would Hitler, for example, run a sim he couldn’t live in and wouldn’t meddle with? Indeed, why would he run a sim filled with Jews? Remember the problem at the top: designers have goals; they wouldn’t make universes that didn’t realize their goals; and if a universe they built started going off mission, they’d intervene and fix that. Designers far more likely will make universes that exhibit their goals. That problem iterates for every other conceivable Evil Villain. Their goals would be evident. And we just don’t see that (see How Not to Live in Zardoz and Will AI Be Our Moses?).

Until you get to Evil Villains so incredibly bizarre that the assumption that every universe would contain one, who would then conquer and unstoppably control that entire universe, starts to sound a lot less likely than that most universes never have sim-verse capable civilizations—or don’t waste resources on useless ancestor sims but task them all on games and paradises, neither of which we live in. It’s improbable that even the imagined villain would exist. It’s vastly less probable still that they will always somehow thwart The Rest of Intergalactic Civilization, who would sooner destroy them even for trying to create even one horrific ancestor-sim. And it’s vastly less probable still that this will happen in every universe. And remember, even one such sim would require more processor space than one hundred billion galaxies…in case you didn’t realize the size of computer we were talking about, just to run a single ancestor sim that would look like this one, much less the “many” in every universe Bostrom’s argument requires.

Bostrom says “there are certainly many humans who would like to run ancestor-simulations if they could afford to do so.” But that’s not true. Not even one tiny bit true. In fact, I don’t know anyone on this planet who really does. Not least because it is so horrifically unethical. But not even Hitler would see any use in them. Indeed, Bostrom concedes “maybe the scientific value of ancestor-simulations to a posthuman civilization is negligible (which is not too implausible given its unfathomable intellectual superiority).” Right. A posthuman civ won’t have any use for replicating the Holocaust. Or the Plague. Or millions of years of animal disease and predation. Or Donald Trump. It would serve no scientific function to them. It wouldn’t answer any questions they’d need answered. It would just waste processor space they could more productively retask into games and paradises.

And remember, you couldn’t discover anything about your own world’s past by running an ancestor sim; because beings within a universe can have no access to the exact initial conditions of their universe, or indeed the exact conditions at any point in their universe’s history, so they could never replicate their universe and history. They also can never have the requisite processor-space—since every sim must be smaller than the world running it, no world can run a sim of itself. That’s logically impossible. But more importantly…if you already know how to sim a universe, there isn’t anything about the laws of that universe you need the sim for. You already know everything there is to know about it. Whereas, for testing contingent histories (which you could only ever do for universes that can never be identical to your own anyway), you have to be as evil as fuck to run such a mass-murderous human experiment. And most people simply won’t let you. They’ll fucking D-Day your ass.

To answer this by saying “maybe aliens would have some reason to do the most immoral and useless thing imaginable, a reason so unusual you can’t think of it” is to do exactly what Christians do to argue evil is not evidence against God. It’s proposing a hypothesis-rescuing auxiliary assumption that is itself massively improbable—and therefore so is any theory that depends on it. Because unusual = infrequent = improbable. Unless you can show the “reason” you have in mind is probable. Until then, positing that hypothesis-rescuing excuse will always lower, not increase, the probability of the overall hypothesis. This is why all Christian and creationist apologetics fails logically. And it’s why you can’t get simverses back into being “probable” again by positing things you have no evidence are even likely.

Does this mean the question of whether we are in a simulation remains unsolved? I’d say it’s as solved as its likely to get, absent some future evidence that alters the conclusion as to what’s more likely: that we are just in a universe—and not in a universe, inside another universe, built by complex aliens, who for some unfathomable and unconscionably evil reason want to run a sim of some other alien universe they already know everything about because they designed it. We can be almost as certain that we aren’t the victims of that crazy Simulator as we can be certain we aren’t the victims of an equally-immoral God. And for pretty much all the same reasons. Excepting only that, on background evidence so far, alien engineers are vastly more likely than perfect magical mega-ghosts. But either is still just another conspiracy theory, as crank to advocate as any other.

Indeed, strong Theism is less probable only because it posits an entity of infinite specified complexity; but the alien engineers thing is less probable than weak Theism (a theism allowing god is limited and flawed), because it requires positing an entire other complex universe besides this one, and containing extremely bizarre entities besides, just to explain this one. That’s as much a problem for Simulationists as it is for all Theists (see A Hidden Fallacy in the Fine Tuning Argument). Bostrom’s argument fails to get around this because it confuses the fact that sims will exceed universes, with “ancestor sims” will exceed universes, which is almost certainly never going to be even close to true.

The George Hotz Problem

Another enthusiast is George Hotz. His apologetic is to bite the bullet and admit our being in a sim “may be unfalsifiable,” without realizing that all unfalsifiable theories are vanishingly improbable. There are infinitely many such hypotheses, and “unfalsifiability” entails no evidence to distinguish any as more likely than any other, which means each is equally likely, which means, therefore, infinitesimally likely. That’s why no honest epistemology allows any warrant to be assigned to non-falsifiable propositions. This is the point of my previous analysis: Simulation theory requires such an enormously convoluted Cartesian Demon, with no evidence elevating any of that convoluted apparatus in probability, leaving it with a vanishingly small prior; and, Hotz admits, no favoring likelihood ratio; and hence, it has a vanishingly small posterior. It’s close to the least likely hypothesis imaginable. It is literally more likely that alien lizard people control Earth’s governments; and relative to all odds, it’s only barely more likely than that God runs the universe with a fleet of angels in a hidden war with Satan.

What distinguishes Hotz from Bostrom is that he drops the implausible ancestor sim idea and tries instead to equate our existence to a video game, suggesting “NPCs don’t know they’re in a simulation.” But this is a false analogy. The NPCs (“Non-Player Characters”) he is talking about are not sentient. If they were, trust me, they’d figure out pretty quickly that they were NPCs in a game—which is why the film Free Guy had to invent a convoluted conspiracy theory as to how NPCs were forcibly kept ignorant of their status. Indeed, this is one of the possible pathways to producing general AI: building an NPC so well that it actually starts thinking for itself and assessing its situation. This was literally the plotline to the recent Westworld series—involving an even more convoluted conspiracy theory within a conspiracy theory, as the Engineer is secretly trying to use a game investors paid him to design to keep NPCs in the dark, specifically in order to develop NPCs that would be smart enough to figure that out. In every case, the measures the Engineers have to go to to keep the “NPCs” from figuring things out is a perfect example of the convoluted conspiracy theory you have to cling to to keep claiming we are in Westworld or Free Guy. It’s literally more likely alien lizard people rule the Earth.

Hotz also relies on bad philosophy of free will to argue for his conclusion. The reality of deterministic free will no more supports us being computer programs in a sim than just people in a world. Because the alternative doesn’t exist—contra-causal free will is an incoherent impossibility, and thus will not exist in any possible world, simulated or real. Whereas real free will exists in all worlds with sentient beings (see Free Will in the Real World … and Why It Matters). It therefore can never be evidence for a sim. The same goes for positing evil Engineers to get around the problem that only villains would make a mere “game world” like this world, pointlessly mass murdering needlessly sentient NPCs for fun. That, too, requires bad philosophy. Because when you run that thought experiment correctly, any expected villainous world would still be self-evidently to its inhabitants a simulation. Just like Zardoz was self-evidently a built world and not some accidental natural one. Hence even evil sims won’t be useless “ancestor sims,” and won’t look at all like the pointless world we live in. Because evil people act completely differently than that. Their worlds will look like sims, literal playworlds for psychos. Not random, ungoverned worlds, inherently purposeless, and massively wasteful of data resources. For example, their villains would be immortal and all powerful. Ours are neither. That is simply improbable on sim theory. It is, rather, what we expect if sim theory is false. Likewise all the waste. Why gazillions of quark-gluon interactions just to make coffee? Why trillions of vacuous galaxies? And so on. “Evil aliens did it” is simply a really, really, really bad explanation.

And that’s just one example of what I mean. Designed worlds would be radically different from undesigned garbage heaps like ours, in thousands of ways, no matter what the objectives and desires of the Creators—unless you gerrymander their interests to fit exactly our world (just as Christians do with God), which requires positing such an unusual and hyper-specific set of conditions as to tank your theory’s probability before you can even get to the desired conclusion. Indeed, when you are talking about villains who can build sims like this, entirely different powers are implicated: literal gods exist in your scenario, and thus everything the existence of gods entails will be true in sims—evil gods or otherwise. But different limitations are implicated: sims have a resource limit in terms of processing space, and thus designs will be efficient in processing, not wasteful; hence we would not have granularity down to the Planck scale or trillions of dead galaxies and vast interstellar vacuums, as all that processing power could be repurposed more efficiently to the function of the sim.

So, I don’t think sims would be betrayed by glitches. Any there are would be smoothed out by the simulation’s scale and operation, just like glitches in our own consciousness are—it’s own kind of simulation—and thus they’d never be visible to us. Rather, I think they will be betrayed by the needs and intentions of their designers. And no intelligible need or intention can be discerned from our world’s design. It appears to have none (neither good, nor evil). It looks exactly like a randomly selected universe would look, with entirely random, non-intelligent processes driving its every development. So that’s most likely where we are.

The Moons of Cheese Problem

That leaves two other apologetics I have encountered to date: one bites the bullet and tries to keep running with the “evil alien conspiracy” version of Simulation theory; the other tries to get around it with a kind of “but won’t someone think of the children” reductio ad absurdum, whereby “bringing children into this world” is supposed to be the same thing as “making worlds like this,” which therefore “cannot” be evil. Both miss the point of everything I’ve just explained.

I’ll start with the first one: the evil alien conspiracy theory. My point already has been that even evil people have more intelligible designs than what we observe our world to have. Even Zardoz was recognizably the product of an intelligence. Even Tron’s inhabitants (eventually) knew they had Users on the Outside. Even Free Guy looked and worked like a game. Even The Matrix didn’t bother with simulating a vast astrophysics and particle physics. In every way, “evil aliens did it” simply predicts yet still different observations than we are making here. So you have to come up with an even more convoluted conspiracy of bizarre motives to explain why even evil aliens made a world like this one. And that’s simply not even remotely probable. It’s going to be among the least likely sets of motives you could pull out of any hat of all the random motives evil Simulators could have. And it is a logically necessary fact that any theory you have to gerrymander like that is going to be far more likely false than true—until you have good, specific evidence it’s true. And you don’t.

To illustrate what I mean, consider what one commenter figuratively suggested: that “certainly in some universes, felines are the top evolved being, and have no problem creating universes to torture mice,” and therefore we could indeed expect our simulators to be evil and doing something as banal as that. We should not take their suggestion literally; it’s a metaphor for just any banality of evil that could arise, from whatever alien form. That won’t help Bostrom, though. A fringe weird case does not get his conclusion. He needs this to be the case for most civilizations—and not “just” civs, but civs that survived themselves (nuclear eras, eras with even deadlier existential-threat weapons, and so on). “Felines torturing mice” do not even make a civilization, much less one that can stay alive all that way, to godlike power (and survive that as well). And even then, they’d just fill the universe with “mice torture sims,” not the “ancestor sims” Bostrom needs. See what I mean? His premise is wildly implausible, to the point of being almost self-contradictory. It’s nonsense on stilts.

It’s the same for Hotz: games look like games. Again, “mice torture sims,” for our hypothetical evil Gamers. This world lacks any such features. There is no analog to the lovely mice torturing these hypothetical evil Engineers want to efficiently simulate—and devote all the universe’s resources to simulating. “But what if they like holocausts?” Then it would be holocausts all the way down. “What if they want a very hyper-specific congeries of banality mixed in with occasional tsunamis and holocausts, all simmed down to a Planck scale, across billions of lightyears of impertinent star-junk?” Okay. And how likely is that? See what lengths of absurdity you have to go to to make it work? You have to gerrymander a hyper-bizarre motive just to get anyone to want to sim a world like ours—and you still haven’t given any reason to think it’s even likely that that’s where we are, rather than just another real place (like the one these weird cats, or their analogs, live in), which (contrary to Bostrom) are the vastly more common places to be.

And this assessment is based on extensive observation and logic—it is derived from all the empirical facts and logical entailments of any sentient intelligence with the godlike power to build and run sims. To think the contrary (like Bostrom) is conjecture; and worse, it’s conjecture against all the available evidence of how intelligent agents operate. It’s the same as if Bostrom was going around insisting all civilizations in all multiverses will inevitably convert all the moons in their cosmos, across every galaxy, into cheese. You need evidence to believe that is even remotely likely. He has none. You can’t say, “Well, it’s only conjecture that all aliens wouldn’t use all the resources of their universe to do that.” It’s the other way around. We have lots of data on how rational beings employ their resources when able. And Bostrom’s argument requires not a fringe exception, but a consistent rational behavior, universal to all capable civilizations, across all possible universes. That’s as bonkers for Simulation theory as for the moons into cheese theory.

This is why any sim theory has to implicate wild conspiracies to conceal the truth, and posit whole other complicated universes on top of ours, making for mega-trillions of ad hoc epicycles, including alien conspirators, complete with bizarre motives—and remember, bizarre means unusual; unusual means infrequent; and infrequent means improbable. Like the motive to make massively inefficient, totally ungoverned sims, with no design features of any use. Moons made of cheese. And I must make this clear: we’re still just talking about evil species, horrifically vile psychopathic aliens, who’ve conquered and repurposed their entire universe. Aliens with any ethics would never make a sim like ours, much less allow an entire universe to be converted into them. The probability of that is functionally zero. Any decent entity would sooner nuke themselves into oblivion before allowing that (the narrative metaphor that takes place in John Carpenter’s The Thing comes to mind). So remember, we’re reduced now to defending vile alien horrors just to explain quarks and coffee. Think that through.

And yet, while evil civs at least do have pathways to this result that non-evil civs don’t, and thus “could get there,” they almost certainly won’t bother, because they are evil. Being evil, they would have no use or care for ancestor sims. That’s an objective waste of resources even from a completely coldhearted Machiavellian perspective. Just as they would have no use for converting all moons in their universe into cheese. “But they could! You don’t know! You are just speculating!” is not a logically effective argument against my point here. Sure. They could. But is it remotely probable enough to even consider? No. And please don’t make the mistake of saying that in an infinite multiverse, there probably is at least one such place; that’s moot to the question of the probability that we are in that place, which will be infinitesimal, and thus still functionally zero (I have addressed this already under the separate subject of Boltzmann worlds and Impossible Gods).

Why would a selfish being burn server time on useless ancestor sims down to Planck scales across millions of galactic clusters, much less burn trillions of galactic-mass-scale server arrays on that, when they could use all that server time for orgies and Doom trolls? There is no rational answer to this question. You have to spin out ever-more-ridiculous and convoluted conspiracy theories to get a different conclusion, which is identical to claiming Satan planted all the fossils and that, before the Fall of Adam caused carnivorism, dinosaur teeth were for cracking coconuts, and that we’ll all get respawned in a thousand years if our game score is high enough—but no one can ever find out how to score or what their score is, or needs to be, because the game designers forgot to install a dashboard. It’s just not going to be plausible. Any more than the convoluted narrative you’d have to invent to justify the moon-cheese objective.

Simverse proponents are thus just like theists: they know their theory is contradicted by all the evidence, so they have to engage in elaborate, ad hoc, “just so” storytelling to keep their theory alive, all against facts or logic.

The Antinatalist Problem

So the evil aliens conspiracy theory doesn’t hold up. You can’t get any credible Simulation theory that way. So the last ditch left is to try and somehow defend the evil as the good—just like Christians do, once they realize, in a panic, that the evidence establishes their God has to be evil (or else not exist). One way to do this is to argue that, well, we bring kids into this morally ambivalent world, with all its dangers, and that’s not wrong—so why would it be evil to make this world, for that very purpose? This analogy fails at every premise.

We don’t randomly throw children into death camps. Not as a normal behavior anyway. If we could make new worlds for our kids, we’d give them games and paradises, not the Holocaust, or the Indonesian Tsunami, or the Black Death (unless we could make worlds but had no ability to design them—a la Source Code—but that would not be Simulation theory). We have to bring kids into this world, because we need them to make this world better, and currently it’s the only world people can come to exist in at all—so the goods of life are bestowable only through the curated introduction of children into the world we are stuck with (see Antinatalism Is Contrafactual & Incoherent, where I have added two new paragraphs on this very point, as argued in the other direction by antinatalists Häyry and Sukenick). This is not the position Simulators are in. They aren’t “stuck” with holocausts, tsunamis, and plagues. Or the global Vatican child-rape industry. They can make the world any way they want. They would therefore be responsible for threatening our children; whereas we are doing our best to protect our children. You can’t blame Jews for Hitler. There is thus no moral equivalence.

Hence our willingness to bear and protect and raise children does not prove Bostrom’s point. It disproves it. His premise requires radically evil and irrational behavior toward children (indeed, untold trillions of them), which has been empirically demonstrated to be exactly opposite the norm for rational beings, and especially opposite the norm of rational beings capable of not wiping themselves out with their own reckless inhumanity, even after being handed weapons capable of erasing planets (as that is the scale of tech required here). If parents got to make the worlds their kids would live in, they would absolutely not make this world. They’d make one vastly less prone to holocausts, tsunamis, and plagues—and every other impertinent, goal-thwarting evil. Hence my point in How Not to Live in Zardoz and Ten Ways the World Would Be Different If God Existed: good Simulators would make a very different sim than this. That’s how we know we cannot be in any sim they would build. Our making the best of what we got stuck with is no counterpoint to that. It is a confirmation of it.

Hence Bostrom’s premise simply makes no sense when you think it through. We need children. We don’t need ancestor sims. We make children because we believe we can govern their world enough to make their lives worthwhile within the resource-constraints we have. Bostrom is talking about a completely different thing: creating countless children we will never meet, and doing nothing for their welfare, even though we have no resource-constraints requiring anything bad threaten them at all. Simulators would have the choice not to throw those children headlong into holocausts, tsunamis, and plagues—and then do so anyway, knowingly and intentionally. They could stop all those things—but don’t. They let them roll over millions of kids anyway. Hence what is evil is the neglect of moral action from the Simulators; not our continuous effort to protect kids from such dangers. Hence if you could secure all kids into a just and lovely paradise, you would. So why would you waste resources on nightmare worlds? You wouldn’t have to. Nor could you even allow it. That’s why Bostrom’s premise is bunk. Who would choose the Holocaust over Eden? Who would tie up any resources on that?

As I already explained, even evil beings would not waste their time. They, too, would prefer games and paradises. They aren’t going to burn server space on useless nightmare worlds. And even if they did, they’d be game worlds or playgrounds (and thus obviously directed by evil intelligences), not indifferent messes like ours. That requires positing an extraordinarily improbable conspiracy theory, just like the lizard people nonsense—indeed, exactly like. And the thing about “improbable” is, it means “infrequent.” Most civs, by far, simply aren’t going to build such elaborate conspiracies; indeed, almost none will bother, even among the evil, and almost none will be evil (owing to the low survival-rate of any evil species with planet-destroying guns). Yet Bostrom’s argument needs most civilizations to do this. That’s not at all credible.

So you can’t get this world to plausibly be the sim of any moral species, nor even an immoral one, without it being literally the least likely place we would be. It’s thus vastly more likely we are in a real world, than any simulation of one.

Can We Get to the Conclusion Some Other Way?

Everything I have said to this point should have been obvious to anyone who thinks about it. So if you are a Simulation theory enthusiast and didn’t think of any of this, you really need to ask yourself why. At what point did your epistemology produce that failure mode? Why did Bostrom never realize any of this? The indication is that he is operating on motivated rather than critical reasoning; he isn’t thinking any of this through. Which means his end-goal is an emotional desire, and not some result of logic he surprised himself with. And so, too, everyone who buys into any version of this convoluted conspiracy theory.

Theists and Simulationists are thus really the same. They desperately need their conclusion to be true; so once their hopes for it are dashed, they will resort to every apologetic device they can think of to try and claw it back into at least some semblance of plausibility. Some of these efforts are literally insane (see for example The Blondé-Jansen Argument from Consciousness). But others are just erroneous; standard failures of critical thinking, a routine outcome of any motivated reasoning. For example, lattice theory—the idea that if we can prove spacetime is quantized, we’ve proved it’s a simulation—fails at the most fundamental level: not comparing it to any of its best competing explanations. Because, um, the universe could just be a quantized spacetime. Therefore, finding it that way is no more evidence that it’s a sim than that it’s just a quantized universe. Indeed, I think spacetime almost certainly is functionally quantized—as entailed by Superstring Theory; for which far more evidence exists than Simulation theory (see The Argument to the Ontological Whatsit and Superstring Theory as Metaphysical Atheism). There simply are better explanations of such an observation than Simulation theory.

Another example is the notion that quantum indeterminism entails “the universe renders only that which needs to be observed,” as Rizwan Virk puts it, which happens to match an efficiency protocol in computer simulations. The problem is that like most amateurs who don’t think these things through, his picture of how quantum indeterminism works is false, and thus his conclusion does not follow. They act like the metaphor of Schrödinger’s cat is an experiment someone actually ran and there actually was a cat both alive and dead. Um. No. Any real cat would observe itself and thus always collapse its wave state into one or the other. But more importantly, consciousness is not required for “observations” to occur in quantum mechanics. That was just a metaphor that has led countless people to misunderstand what wave-state collapse actually is (or is purported to be: this remains an unproven theory of quantum mechanics, not a fact of it). Any interaction with a system is an observation. Hence, for example, in any real “Schrödinger’s cat” experiment, the very anode that detects the radiation and triggers the lethal gas to release and kill the cat is observing the radiation. It will thus immediately collapse the wave-state. Maintaining pure (uncollapsed) quantum states is extremely difficult—it requires extraordinarily delicate arrangements of systems to prevent the slightest perturbation from an interaction that would collapse it.

This is why quantum states kept on collapsing everywhere for billions of years before we (or life at all) even existed. It’s why we get consistent radiation from events billions of years ago from all directions despite the speed of light being so slow. We cannot have collapsed all those wave forms in reverse time as soon as one of us woke up, because signals from them (light, for example) hadn’t even reached Earth yet, nor could any signal have traveled back everywhere in reverse, so as to go all the way out there and have all the effects there needed, and then back-track all the results to us instantaneously. This is an extremely convoluted—and maximally inefficient—way to run a computer program. Virk references the example of first-person shooters like Doom where every view is rendered only as needed, so as to save processor time and capacity. This is not only not what we observe (the world is not assembled to our perspective; remember, billions of years of shit happening across billions of galaxies before we even showed up, but even in everyday reality our car in the garage is not in a Schrödinger state the moment we leave the garage), but it is not what we would observe in a fully functional world sim.

Doom is a game, designed to solve only problems of the moment, to effectuate the playing of a game. We don’t live in a game. But even more importantly, this is not the most efficient way to simulate a world and not just a temporary play-space. For a world sim, it is actually more efficient to work out the entire system’s contents and let it run. Remember, the most efficient simulation of any thing is the thing itself. Having a computer have to calculate what a room looks like solely to you when you walk into it requires a vast amount of instantaneous math, and only solves that one single problem, which can then be discarded when no longer in use. But having the room already a running program when you walk into it requires a great deal less math to work out—particularly since the room is doing a thousand more things in a world sim than in a game: e.g., the photons bouncing around in it decide the room’s thermal state, which affects the thermal state of environs adjacent, and so on, so that it cannot be put in stasis when you aren’t in it; likewise the condition of what air fills it, and that air’s pressure and contents, and all its associated external inputs and outputs, or the dustiness of the room, and what that dust consists of, and its external inputs and outputs, and so on. It is simply easier to run all this as a coherent system than to customize its appearance to each person only when they walk in—or walk anywhere that is causally chained to it (every room is an inalienable component of the entire thermal system of the Earth, the condition of its atmosphere, and so on). Think of what it would take for a sim to work out billions of years of billions of lightyears of backstory every time each person looks up at the stars, versus simply having done all that once already and referencing the results every moment every person looks up at the stars. One of those methods is far more efficient than the other—and far more to the purpose of a sim in the first place. Probably one of the best examples of this in practice is Dwarf Fortress.

In short, actual game-specific rendering-to-purpose (as in Doom) would look very different to us than the world we actually live in does (and science would thus have discovered this already). Whereas a real “universe sim” would not operate like Doom, but like Dwarf Fortress—but that is also how a real world would work, so that observation no longer can distinguish a simverse from a realverse. Indeed, since a realverse is always going to be a more efficient simulation of itself, Engineers would sooner just make a universe than waste computer processing on it—unless they wanted to do something with that, like meddle or intervene, or test a specific arrangement. Which gets us back to all the reasons the world looks less like such a project and more like a random universe aimlessly plodding along, requiring no extra complex layer of reality to explain any of it. The bottom line is: the universe looks exactly like a non-simmed universe could be expected to look. So why add any epicycles to that? We need specific evidence to warrant such a thing, just as Ptolemy did. Yet like him, we don’t have any. So sim theory is as likely to be wrong as geocentrism was—indeed far more so, given the far greater scale of improbable coincidences required. Ptolemy could pull off his scheme with a few dozen epicycles. Simulation theory requires umteen billions of them. So you’re simply far more likely just living in the universe than a simulation of it.

The irony is that the reason we can increasingly draw parallels between the real world and simulations such as in games now is precisely because both simulations and real systems inevitably seek their most efficient states. Every effort we make to efficiently solve a problem in a sim (like “how light bounces off a wall” or “how thrown rocks behave”) is simply trying to get at the most efficient way a real world would solve that problem; and it always turns out to be the same way the real world does it. Doom renders on demand, but it does so using equations of state that happen to be the same as would describe any real world system. Yet unlike Doom, we know that rendering is not slaved to our perspective but is always running even when we are not around, just like Dwarf Fortress. So all the attraction to the simulation hypothesis is based on a basic fallacy of reasoning: getting the causal direction wrong. Sims are getting more like reality because we are deliberately trying to make them more like reality, and are only discovering all the efficiencies reality already inevitably landed on. No one is asking what sim we would make if we didn’t have to mimic an already existing world.

This illustrates two hidden failure modes within Simulation theory:

• First, if sims have to be steered to look more like an already-existing reality, why would we conclude that already-existing reality is itself a sim? Isn’t it more likely the thing being simmed?
• Second, why would an alien build a simulated world like this, and not like a more intelligent, self-serving construct? Wouldn’t a simverse look and function entirely differently than ours?

To illustrate what I mean, think again of Tron, especially as depicted in Tron: Legacy. That depicted a simverse, but it did so far more realistically than almost any other film. The Matrix, for example, implausibly had the sim struggle to look and operate just like a second, other, real world; but there is no intelligent reason to do that (Agent Smith’s implausible excuses notwithstanding—he wasn’t a reliable narrator anyway). In Tron the sim operated the way we would expect a sim to: it achieved efficiency by not wasting processor power on pointless magnitudes of scale. Our world is needlessly rendered down to units of 10^-35 meters and 10^-43 seconds; as you can see from the way things function and break, Tron didn’t waste rendering anything below 10^-4 meters or seconds, leaving it over 10^-30 times more efficient—nor did it bother simming even an entire planet, much less billions of galaxies and billions of lightyears of intervening space. And everything was designed to purpose: that world’s rules and contents were all intelligently purposed, not random and pointless. Likewise other far more plausible sims, like, say, a Miyazaki world, with the granularity of anime and its far simpler physics. Instead, to try and make our sim look “exactly like” an actual world (an actual “ancestor” sim as Bostrom insists) entails there actually is a real world that already works entirely the same way and contains all the same things. So what’s the point of asking whether we’re in the simmed one or the real one? They both then exist; so we could be in either, right?

Which is the final blow to Simulationist enthusiasts. If the sim is just a perfect replication of an actual world, how could there be any evidence it’s a sim and not the world it is simming? If there were differences, it would fail to succeed as an ancestor-sim (as those differences would veer the sim off any intended course; or certainly ruin any experimental value it could ever have had, by tainting the results with non-conformities); whereas if there weren’t any differences, then it can never be the case that we could ever have any evidence we were in the simmed version rather than the real one—because they would always look and operate identically. That would also mean there was a real world, exactly like this one in every respect (except perhaps in matters of its contingent history). Which of course leaves us with the conclusion that it wouldn’t even matter whether we were in the real one or its sim; it would neither explain nor change anything of significance (see my discussion of this conundrum in “Meaning, Reality and Illusion,” Sense and Goodness without God, I.2.1.2). Every effort we make to find out why quantum mechanics behaves as it does is as much discovering facts of the real world (which must also be governed by all those same quantum mechanics) as of the simmed world—because there is no relevant difference between those worlds.

This is where the Whack-a-Mole apologetic train starts. Bostrom requires the inexplicable necessity of an ancestor-sim. Virk requires Bostrom to be wrong about that. But if Bostrom is wrong about that and we are still in a sim, it should exhibit the goal-directed design and behaviors of a game or paradise; but it doesn’t, it looks inexplicably like an ancestor-sim. And round and round it goes. It’s just always more likely we are in the world Bostrom imagines would be simmed, than in any sim of it. The latter requires infinite epicycles and conspiracy theories, as any Cartesian Demon scenario does. Without evidence supporting the scenario (of a kind depicted in The Matrix and Tron, for example), it’s simply, epistemically, vanishingly improbable. It is thus always far more likely we are in the real world, and not a simulation of it. Because we already know we are not in a game or a paradise (see How Not to Live in Zardoz and Ten Ways the World Would Be Different If God Existed).