Could our Universe be a simulation? How would we even tell?

Ever since Oxford philosopher Nick Bostrom proposed his simulation argument in 2001, the nerdiverse has attempted to assess the possibility that reality is not really real, that what we experience as our Universe is instead the product of a computer simulation. Popular figures such as Elon Musk and Neil deGrasse Tyson have offered their own conclusions, but taking a firm stance was not the point of Bostrom’s argument. Instead, Bostrom’s position is nuanced and careful, and it doesn’t arrive at fixed answers.

I’ll take it for granted that the Ars readership is more sophisticated than the average geek, so let's take some time to dissect Bostrom’s simulation argument, exploring its construction, its implications, its strengths, and its weaknesses.

But I have to warn you: If you’re hoping for relief, one way or the other, from the existential crisis brought about by the possibility that we live in a simulation, you won’t find any comfort here. The firmest conclusion anyone can reach, after examining and re-examining the arguments for and against the simulation thesis, is a profound yet resigned "maybe."

The Universe in a box

Bostrom’s argument relies on a simple observation. We continue to develop ever more powerful and capable computers, and our abilities to simulate the Universe, from cosmic to microscopic scales, are becoming more comprehensive with time. As a theoretical cosmologist who specializes in computation, I’ve witnessed this firsthand in my own field. Decades ago, we could only simulate small portions of the Universe, representing individual galaxies as tiny dots of gravitational attraction. Now, our most sophisticated universes-in-a-box include star formation, magnetic fields, cosmic rays, radiation, and more, and they trace the evolution of millions of galaxies simultaneously through billions of years of cosmic evolution.

At the other end of the scale, we apply the known forces of nature to simulate the behavior of nuclear matter, the interaction of elements and molecules, and even the complex relationship among synapses.

Presumably, simulations will eventually become sophisticated enough that we could re-create our entire experience of the Universe within a computer, with simulated conscious brains experiencing a simulated reality. Crucially, Bostrom’s argument doesn’t depend on when this turning point will happen, only that it does. It could be in the next decade, via some revolution in computing that’s right around the corner. It could be a hundred-thousand years from now, where a post-human society wields a knowledge of physics currently unknown to us to build a planet-sized supercomputer that faithfully recreates our current experience of the Universe, complete with consciousness and exploding stars and the smell of a good camembert.

The second major piece of Bostrom’s argument relies on the nature of consciousness. Specifically, that consciousness is generic, whether it arises from wet, squishy biological synapses or clean, dry electronic hardware. For the simulation argument to work, consciousness has to arise within the simulated universes and be basically the same as the au natural version. It has to be aware and, well, act like we’re used to conscious human brains behaving: capable of self-awareness, independent action, and deep contemplation about the fundamental nature of reality.

Bostrom admits that this portion of his argument isn’t without controversy and is a matter of much debate within philosophical circles. But with these pieces in place, we can move on to Bostrom’s argument: Someday, our descendants, or some very eager alien civilization, will create simulated consciousness and place those digital brains in a simulated universe. Bostrom calls these “ancestor-simulations,” following the line that future cosmologists will want to recreate the entire history of their Universe down to the level of subatomic interactions. In our case, this would naturally include a simulated Solar System with a simulated Earth that evolves simulated people having simulated arguments on simulated Internet forums.

Once this starts, the number of simulated brains will vastly outnumber the organic brains. Think of all the digital creatures that have ever “lived” in all instances of all video games combined: how many NPCs, monsters, and avatars were born with the click of a button or the flip of a switch, followed their programming, and then were just as quickly shut off? With suitably powerful computers at their disposal, the simulation-builders wouldn’t just stop at one brain in one Universe; they would make a bunch of Universes, each containing a bunch of brains. At this point, the vast majority of conscious entities would be simulated rather than biologic.

The trilemma

That’s it. That’s Bostrom’s simulation argument: That once somebody, somewhere, at some time, develops the ability to create faithfully simulated brains, most brains in the Universe will be housed in some computational facility.

Crucially, this line of thinking does not immediately lead to the conclusion that we are living in a simulation. Instead, the argument ends in a trilemma, three possible conclusions that, absent any additional evidence or convincing argument, are all equally valid.

Conclusion #1: For some reason, our future descendants, or any other beings living in the Universe, will be unable to build such a simulation, either because of some limitation in computing that we can’t yet fathom or because all intelligent species manage to kill themselves off before embarking on the journey of building ancestor-simulations. While the idea is a little depressing, it’s not that far-fetched. We’re not even close to the necessary level of computational sophistication, and we’ve managed to develop globe-killing nuclear weapons and civilization-disrupting climate change.

Conclusion #2: For some reason, our future descendants, or any other beings living in the Universe, will be unwilling to build such a simulation. They totally could if they wanted to, but they don’t. Perhaps there’s an enormous amount of social pressure or self-policing that prevents every super-advanced society from building large numbers of simulated brains. I mean, I feel kind of bad when I kill pigs in Minecraft, so maybe as the eons go by, this distaste evolves into a universally consistent Butlerian-Jihad-style proscription against this class of sophisticated simulations. Or maybe we can’t fathom what post-humans want to spend their time on, and nobody in the far future even bothers running such a simulation because it’s not interesting to them.

Conclusion #3: Our future descendants, or any other beings in the Universe, totally can and totally want to build large numbers of simulated brains, in which case the vast majority of all conscious beings in the Universe are digital rather than biological, so the raw probability of us being simulated quickly approaches one. We are likely living in a simulation.

Trilemma laid out, we now get to the fun part of philosophical inquiry: the arguing. But I want to make one thing very clear. Despite the claims of some science popularizers, the simulation argument is not pseudoscience. Pseudoscience is the act of pretending to make scientific claims without the rigor that science requires. But Bostrom is a trained, professional philosopher going about his day job, which is to be philosophical. Bostrom is not making scientific claims or advancing testable hypotheses. He's making a philosophical argument based on clearly defined assumptions and a rational line of thinking. That’s not pseudoscience any more than carpentry or accounting is.

Indeed, Bostrom’s argument stops at the trilemma. His paper continues, assessing the strengths and weaknesses of the three possible conclusions (with a strong preference for #3), but in the end, that’s as far as the assumptions can take you. So it’s up to us to decide where we land.

Task failed successfully

One option is to take the trilemma at face value and accept the uncertainty that comes with it. Our future descendants may or may not build a simulation. We may or may not live in one. That’s it. There are no answers, no big revelations. It's an interesting dinnertime conversation, but it ultimately leads to nowhere we weren’t before; there’s no new knowledge or perspective on reality to be gained here. Shrug your shoulders, have a drink of possibly simulated beer to take the edge off the existential crisis, and move on with your life.

While I won’t begrudge anyone for taking that position, Bostrom’s argument allows us to dig deeper into the nature of computation, consciousness, and even reality, and we can do that by poking and prodding at the assumptions and conclusions of the argument itself. And that’s a lot of fun, so we’ll keep going.

Another approach to the trilemma is to argue that one of the conclusions is more reasonable or more expected than the other. Note that this is not taking bets but advancing rational lines of thinking in favor of a position when there is no clear-cut evidence either way.

For example, perhaps it’s fundamentally impossible to recreate our Universe. We don’t even have complete access to all the laws of physics—the stuff going on down at the Planck scale is still a head-scratcher, after all. And there’s a plethora of mysteries and experimental inconsistencies that keep scientists of all stripes awake at night (but at least employed), so an authentic simulation of the Universe wouldn’t include just the physics we know about but all the physics we don’t. There may not be enough resources available to ever build a computer capable of the feat. And that’s not even including the most complex structure known to humanity: the human brain. We have no clue about the complexity and sophistication required to generate consciousness.

Bostrom argues that a simulation doesn’t have to be 100 percent comprehensive at all scales to work. Clever programmers could ignore many of the fine details (like, say, the majority of quantum interactions) until some simulated brain decides to run a particle physics experiment, in which case the programmers would need to devote some small amount of additional resources to cover the added complexity.

At first glance, this is a compelling argument. Does our awareness of the Universe really depend on what some random hydrogen atom is doing in the Andromeda galaxy? But it’s not necessarily correct. Because we don’t fully understand physics, especially the non-locality inherent in quantum mechanics, we don’t know what matters and when. And speaking from personal experience in developing high-fidelity simulations of the Universe, you can’t just wave away the small-scale stuff with low-resolution models. If you want to get it right, you have to be complete. Most, if not all, of the macroscopic processes we take for granted only arise from the combination of countless quantum operations.

At the very least, an accurate simulation of the Universe that hosts these simulations would need the necessary level of completeness to enable the creation of universe-simulators within the simulation. If it didn’t, it wouldn’t be entirely faithful, threatening the assumption that conscious awareness inside the simulation is just as good as the biological kind. Note that a nested series of simulations-within-simulations isn’t necessary for Bostrom’s argument to work, but we have to consider the possibility as we gauge the complexity of building realistic mock universes.

On the other hand, who knows what future post-humans could be capable of? What we consider complex and difficult calculations today may be accomplished by the machinery of their toasters. They could be so sophisticated in their understanding of physics and computing that they could create mock universes with such granular detail that those mocks could create mocks within themselves—all without the computational hardware even breaking a sweat.

Another angle we can take when picking apart the trilemma is to investigate the construction of the argument itself, making sure that every link in the chain is sound. For example, probability weighs heavily in Bostrom’s reasoning. If our future descendants (or aliens—aliens could always do the same thing) build a simulated universe, then there might be 99 simulated brains for every one organic brain. So if you turn the question around and ask about the probabilities that you are a simulated brain, it’s a 99 percent chance, all else being equal.

But maybe everything isn’t equal. Maybe there is something ineffable about organic consciousness that cannot be completely captured in simulation—maybe true awareness requires meat brains in meat bodies. Maybe we could devise clever experiments regarding consciousness that can turn this into a scientific argument, allowing us to discover whether we do or don’t live in a simulation. Perhaps simulated consciousness doesn’t last long, or it isn’t stable, or there’s some other reason that prevents simulated brains from being truly equal to the organic ones. Any of these could change the calculation of the probabilities.

Lastly, we have to be careful when counting brains. If we were to successfully develop a 100 percent faithful simulation of the Universe tomorrow, then we would know for sure that we weren’t the ones in that simulation. We could create a dozen, a million, a trillion simulated brains, and that still wouldn't answer the fundamental question of whether we are simulated or not. All we could say for certain is that we’re definitely not inside the simulations we create.

Our ability to create simulated universes doesn’t get us closer to the answer because we don’t know if we’re the first organic iteration of humanity to accomplish the feat of building these simulations or whether we’re just running a simulation-within-a-simulation originally created by organic humans. Alternatively, we could be the invention of some highly advanced, ingeniously clever, and profoundly bored aliens—and since we haven’t met any aliens yet, we don’t know what they’re capable of. In other words, we have no way to count the number of simulated versus organic brains, so we can’t assess probabilities.

The blue pill

So here we are. Possibly simulated, possibly organic, with no way to tell the difference and no clear argument one way or another. What are we supposed to do with that?

Upon the realization that we could be simulated, you might be tempted to throw away life—what good is it if it’s not real? But if you were predisposed to nihilism, I doubt you needed an Oxford philosopher to help you along. A simulated universe isn’t a fake universe; it’s just real in a different way than we expect. Simulated pain still hurts. Simulated love is still powerful. We can still strive to be good stewards of our planet and generous toward our neighbors; faithful simulations are miserable to live in if you have neither shelter nor food.

We could spend an entire book dissecting and examining the ins and outs of the simulation argument. Indeed, Bostrom’s conclusions sit in a storied philosophical tradition, a long line of skeptical arguments that question the very nature of existence. This is the joy of philosophy—to challenge our basic assumptions and interrogate this strange Universe that we find ourselves in. Science is a part of that journey, and it's a branch of philosophy in its own right (if you ever meet a scientist who disparages philosophy, ask what the “Ph” in their title stands for and watch them squirm).

Perhaps some of the arguments outlined in this article spoke to you; others may have been less convincing. So now it’s our job to do what philosophy teaches us to do, simulated or not: to explore, to question, to argue, and to learn.