“...In time, those Unconscionable Maps no longer satisfied, and the Cartographers Guilds struck a Map of the Empire whose size was that of the Empire, and which coincided point for point with it.”
On Exactitude in Science, Jorge Luis Borges
Nick Bostrom’s groundbreaking paper, Are You Living In A Computer Simulation?, uses rock-solid logic to argue that either the human species will go extinct before reaching a ‘posthuman’ state, that posthuman civilization won't run many simulations of their evolutionary history, or that we are likely living in one of those simulations. His reasoning was strong enough to convince the richest man on Earth to claim the chances of us living in a baseline reality were one in billions.
Now to reach these conclusions we have to accept a few preconditions, namely that an accurate simulation of a conscious mind interacting with its environment is itself conscious, that a ‘posthuman’ state is a resource-rich post-scarcity environment, and that the evolution of computational technology will continue without natural constraints, giving our descendants near godlike simultaneous processing power —all of which, to be fair, many of us are totally on board for.
But what the author doesn’t examine is to what extent ‘ancestor simulations’ would correspond to our own lives, and if so, whether there might be telltale indications of the constraints they would require.
What even is an ancestor simulation and how would you create one?
If you're a fan of old-school science fiction or the Bible, perhaps you'd start with Adam and Eve, somewhere on a simulated Euphrates river valley. This would be a charming idea, but a nightmarish mistake.
The International Union for the Conservation of Nature has assessed that to maintain the vitality of any population over time, a minimum group of 500 reproductively effective individuals are required. This is the baseline, whether your DNA is simulated or not. If you start with Adam and Eve, before just a few generations have passed, you'll have a pack of inbred dysmorphic hemophiliacs, all with the same last name.
But it is fairly easy to imagine our posthuman descendants spinning up 500 genetically varied hominids on a simulated 50,000 BCE Earth. As these simulations progressed, their creators would gain a solid understanding of human psychology, sociology, and the evolution of culture. Maybe the simulated humans would develop societies reminiscent of ancient Mesopotamia, China, or Central America.
But that would not be our history.
Each simulation would be one of the trillions upon trillions of paths that humanity could have taken. Very useful for a general understanding of human nature, not useful at all for studying the ‘authentic’ timeline experienced by those who build the simulations. To do something like that, you are going to need to enforce very specific conditions that would be significantly more difficult to generate, regardless of your processing power.
Let’s say, for example, that our post-human descendants want to study the rise of Fascism in 1920s Germany. Where do they begin? How, from their thousand-year distant perch, do they accurately spin up a weak-willed, communist sympathizing railroad porter or a laudanum-addicted Bavarian housefrau?
If the internet is any indication, it should be clear by now that all the processing power in the world won't make you a good writer. Those self-aware simulations need to have believable backstories, memories, and connections. At any point in time, culture is a complicated web of interrelationships. Your simulation’s predictive qualities will be meaningless if you just pull those details out of your ass.
To better understand your civilization's own past, you'll need a way to create a reasonably authentic timeline, not just spin up a hacky self-aware version of Weimar Sims.
One way to achieve this might be to start with an early man and only follow timelines that lead toward your own history.
Bostrom gives a lower estimate of 10^14 operations per second for an accurate simulation of the human mind. He also posits that a planetary-mass computer of the posthuman era might have a computational power of 10^42 operations per second which would allow it to simulate the entire mental history of mankind by using less than one-millionth of its processing power for one second (this would be 100 billion humans x 50 years per human x 30 million secs per year x [10^14, 10^17] operations in each human brain per second = [10^33,10^36] operations).
Now, a million human histories a second is a lot for sure, but it might not help us address the underlying issue.
Say you’re a male hominid slouching near your cave on a sunny morning in 500,000 BCE. You might pick at your lice, rub your butt, look around for a snack, beat your children, climb a tree, think about sex. Sociology tells us that on the large scale our choices are somewhat constrained, but surely we could say that for every human being there is at least one mindstate option a second, at least one branching of the timeline for that simulated brain. This means that if you hope to chart the entirety of the human developmental space through time to better locate and understand your own culture, you’re going to need to process 10^14 operations a second per 500 simulated minds at the second one.
The second two would be 1000 mindstates to attend to, then 2,000, then 4,000… After 60 seconds you’ve got roughly 10^20 mindstates times 10^14 processes per second. You’ve created a calculational velocity. After less than two minutes we’ve massively exceeded the capacity of Bostrom’s planetary sized computer (and we’re not even taking into consideration limitations caused by the immense cooling fans that will be needed at the 3.6 million atmospheres of pressure going on in the core of a machine with the mass of the Earth).
In other words, an Earth-sized computer would be unable to calculate its way through the timeline space of an actual Earth, in order to locate any particular timeline if they were to begin with original conditions. It becomes a near-immediate clusterfuck if you're trying to just power through it.
I know what you’re thinking though, the simulation of a human mind might take 10^14 operations a second, but over the course of two seconds, how much has really changed?
So let’s take as a guide the compression algorithms that we use for video. Perhaps our Earth computer only really needs to chart out the differences between mental states over time, maintaining their similarities in a single repository. The processing load of the branching timelines could be significantly reduced. But even with tenfold savings, it would take less than five minutes to overload the Earth-sized computer of our evolutionary descendants. A hundredfold or even thousandfold reduction of processing needs would similarly time out before the passing of a single simulated day.
Just as the crypto wallet app in your phone can randomly create an address without concern that it’s already being used —because the information space of possible wallet addresses is so astronomically high— our particular human timeline is so outnumbered within the information space of possible timelines that it would be virtually impossible to locate it through brute force processing in any manner that obeys our current understanding of physics.
But you might say, let’s just prune that simulation back as we move forward. If our simuloids don’t start speaking Proto-Indo-European after a few thousand years, we'll turn that branch off. The problem is that as soon as you start from scratch at a single point in time you’ll encounter the same branching issue if you want to explore the entire information space. And you’ll never really know on the small scale which path is the right one if you want to constrain it.
Let’s say one of your hominids walks off and starts bashing his head against a tree for hours at a time. You might be inclined to turn that timeline off ...or if you wait around, you might discover that he has been so frustrated by having to lug animal carcasses across the savannah that his inner angst leads him to invent the wheel, setting in motion the flowering of civilization. Who knows what tortuous path we followed to arrive at our own point in time?
So the first constraint of any pseudo-authentic timeline is that you'll need to spin it up relatively near the period that you're interested in studying. Just like predicting the weather, initial conditions are extremely important and small changes grow quickly over time.
But how do you manufacture that first timeline with any sort of authenticity?
There are only a couple of options. You can begin your timeline in medias res, crafting human minds the way writers do with a streaming series on Hulu, which would inevitably introduce assumptions and prejudices from your own time period and make your simulation worthless for anything beyond casual entertainment. You would be in serious danger of your creations realizing that they are only poorly constructed cliches, leading to a civilization-wide panic attack.
Alternately, you could start from scratch with a group of humans in early prehistory and influence your evolving timeline through external means as it develops, treating your simulation as a bonsai tree that needs to be sculpted, probably with a lot of crisis actors and mysterious interventions. Or finally, you might begin with an accurate simulation of your 'current' posthuman time period and run your simulation backwards, generating a single best guess of all the individuals who ever lived, using historical data to optimize your approximation as it constructs history in reverse chronological order.
However this initial 'best guess' timeline begins, our descendants could optimize it, running the simulation forwards and backwards until they had a fairly realistic simulacrum of human history. It wouldn't be the real one, but it would come close at all the major points. This would be a monumental feat of engineering, but once they achieved one of these 'canonical' timelines, researchers of the future could then spin up a simulation beginning at any point in time and watch actual random chance and real free will exert their natural impact as it progressed.
Which means that the vast majority of ancestor simulations would fall into one of two categories. The most scientifically rigorous sort would begin with simple, culturally undeveloped prehumans, and by the time any such civilization evolved to a level of complexity approaching ours it would be completely unrecognizable to our descendants (or us). The other option would be to spin up a simulation that branches off one of the canonical 'real history' timelines, and which would bear a strong resemblance to the pre-history of the posthumans who generated it.
I would argue that when you take scientific interest, relatability, and entertainment value into consideration, the latter type would quickly become the dominant form of ancestor simulation.
This means, that if you're playing the odds according to Bostrom, you may certainly be living in a simulation. And if so, it's likely based on the 'original' timeline of our posthuman descendants, and in all likelihood began no more than a few generations ago (because otherwise, it would be of limited utility having veered off in its own direction).
But what are the telltale signs you're living in one of these simulations?
You'd have to ask --what time periods would be most interesting to future researchers? Perhaps they'd want to study periods of rapid technological change? Or the point at which mankind failed to prevent climate catastrophe? Or the beginning of the end of democracy?
There are plenty of reasons to think our time would be of interest. What keeps me from complete solipsistic paranoia though, is how unimportant I am. Because of my complete lack of cultural distinction, I'm pretty confident that any simulation I'm a part of would have thousands or millions of others just like me. Probably some kind of massive 'Decline of America' reenactment.
And that's comforting, because if I were someone who could expect to be remembered into the distant future, the sort of individual our descendants might want to study again and again through any number of generations, then I'd begin to wonder if perhaps I was living in a simulation of one. Because the simplest and perhaps easiest artificial reality to generate would have only one mind with free will, surrounded by heavily scripted non-player characters. And wouldn’t that be a nightmarishly lonely realization?
Personally, I wouldn't make my odds at living in a simulation higher than 33%, but can you imagine, waking up, looking in the mirror, and seeing Elon Musk?
Because the chances of him living in a simulation are literally billions to one.