Our species is not going to last forever. One way or another, humanity will vanish from the Universe, but before it does, it might summon together sufficient computing power to emulate human experience, in all of its rich detail. Some philosophers and physicists have begun to wonder if we’re already there. Maybe we are in a computer simulation, and the reality we experience is just part of the program.
Modern computer technology is extremely sophisticated, and with the advent of quantum computing, it’s likely to become more so. With these more powerful machines, we’ll be able to perform large-scale simulations of more complex physical systems, including, possibly, complete living organisms, maybe even humans. But why stop there?
The idea isn’t as crazy as it sounds. A pair of philosophers recently argued that if we accept the eventual complexity of computer hardware, it’s quite probable we’re already part of an ‘ancestor simulation’, a virtual recreation of humanity’s past. Meanwhile, a trio of nuclear physicists has proposed a way to test this hypothesis, based on the notion that every scientific programme makes simplifying assumptions. If we live in a simulation, the thinking goes, we might be able to use experiments to detect these assumptions.
However, both of these perspectives, logical and empirical, leave open the possibility that we could be living in a simulation without being able to tell the difference. Indeed, the results of the proposed simulation experiment could potentially be explained without us living in a simulated world. And so, the question remains: is there a way to know whether we live a simulated life or not?
At some point in the future, humans as we know ourselves will cease to exist. Whether we become extinct with no evolutionary descendants, or leave one or more post-human species as our inheritance, we humans will eventually be gone. But if we do leave futuristic descendants, those descendants might be quite interested in creating ancestor simulations, virtual universes populated by conscious humans. And if the technology to craft such simulations was sufficiently popular, they could proliferate so widely that the first-person experience of such simulations would outnumber the first-person experiences of humans who have actually existed in fundamental reality.
This presents an interesting problem if you happen to find yourself having a first-person conscious experience: how do you know whether you are one of the original humans, or an ancestor simulation, especially when there are many more of the latter? The philosopher Nick Bostrom has provided a framework for thinking about this problem. He argues that we have to conclude one of three things is true. Either humans or human-like species become extinct before they achieve simulation-producing technology, or ‘post-human’ civilisations have little interest in making or using this technology, or we ourselves are probably part of a simulation. I say probably because, all things being equal, the odds would be greater that a conscious experience is a simulated experience. There would just be way more of them around if the other two conditions (extinction or lack of interest) fail.
Bostrom is certainly not the first to examine the possibility that our perceived reality is virtual, though the proposed nature of the simulator varies a lot. In addition to philosophical and scientific ruminations, the idea that human consciousness is simulated is a staple of science fiction. In the movie trilogy beginning with The Matrix (1999), the world we know is a computer simulation to keep humans’ brains busy while their body chemistry was harvested for energy. In The Matrix, humans experience the world as avatars in a fully immersive virtual reality environment. However, the simulation was sufficiently flawed that some prepared minds could see its glitches, and people from the ‘real world’ could hack into the Matrix.
Bostrom’s idea is somewhat different: in his picture of things, the whole Universe is a simulation, not just humanity. Every aspect of human life is part of the code, including our minds and interactions with the non-sentient parts of the program. However, Bostrom recognises that a complete emulation of reality on every level is likely to be impractical, even for powerful computing systems. Just as our scientific simulations involve levels of abstraction where excess detail isn’t required, simulations would probably make use of certain rules and assumptions, so that not every detail would have to be simulated. These would come into play when we performed experiments: for example, ‘when it saw that a human was about to make an observation of the microscopic world, [the simulation] could fill in sufficient detail in the [appropriate domain of the simulation] on an as-needed basis,’ Bostrom writes in the paper ‘Are You Living in a Computer Simulation?’ (2003). That way, the program wouldn’t need to track every particle or galaxy in every detail, but when those data are called for, enough of the cosmos is in the program to provide a completely consistent reality. Even humans need not be emulated in every detail at all times; our subjective awareness of ‘self’ varies depending on circumstances. Unlike Linus in the cartoon strip Peanuts, we are not always aware of our tongues, so the simulation need not keep the ‘tongue’ subroutines operating in the foreground.
it could be the case that one planetary civilisation is all that can be simulated, without running into computational capacity issues
Beyond these philosophical implications, the simulation hypothesis could help answer some scientific problems. Since Earth-like planets are not terribly rare, it’s possible enough civilisations have arisen in the Universe that they would be able to communicate or travel between stars. Yet we have not seen any so far, leaving us to wonder: where are the aliens? However, if we live in a simulation, aliens might simply not be part of the program. In fact, it could be the case that one planetary civilisation is all that can be simulated, without running into computational capacity issues.
Similarly, the failure of physicists to find unified theories of all the forces could be due to an inadequacy in the simulation. The simulation hypothesis could even resolve the ‘fine-tuning’ problem: that the parameters of our Universe allow for life, but changing them might result in a lifeless cosmos. A simulated Universe could be designed for the eventual rise of life, or alternatively could be the outcome of a successful experiment in which many possible parameters were tested before life was possible. Cosmologists perform similar (albeit simpler) simulations now to see how likely our particular cosmos is from random starting conditions.
Bostrom goes a step further in his simulation argument: ‘Should any error [in the program] occur, the director could easily edit the states of any brains that have become aware of an anomaly before it spoils the simulation. Alternatively, the director could skip back a few seconds and rerun the experiment in a way that avoids the problem.’ However, if the simulation in which we live has real-time error correction, it’s troubling from several points of view. Indeed, it could potentially throw the whole enterprise of science into question. What would prevent the simulator from changing the laws of physics on a whim, to test parameters or simply to mess with our heads? In that scheme, the programmer becomes a capricious and possibly malicious god, whose presence can never be detected.
While Bostrom is interested primarily in showing that we’re more likely than not to dwell in a simulation, scientists who confront this problem have a different set of questions to answer. The primary contrast derives from the fact that science is concerned with what can be tested by experiment or observation. And, as it turns out, there are a few things we can infer from any simulation we might inhabit.
First, if we live in a simulation, it obeys a set of well-defined laws, and any dynamic changes to those laws are relatively small. That’s based on the overwhelming success of the scientific approach over centuries. In fact, the simulation hypothesis has some potential explanatory power: the reason our Universe obeys relatively simple laws is because it was programmed to do so. As for changes the simulator makes as the program runs, that was one proposed solution to the ‘faster-than-light’ neutrino results from 2011: the program contained an error, and we measured something based on that error, and the bug was subsequently fixed. (There’s currently no reason to think the faster-than-light result was real, since the anomaly has a prosaic explanation, requiring no dramatic alternative ideas.)
The truth of the matter might be that we dwell in a simulation but, like the existence of an impersonal god, this fact has no bearing on how we conduct our lives
However, there’s nothing in this cosmic lawfulness to tell us whether we’re in a simulation or not. If the program is good enough with no obvious ‘Easter eggs’ or hidden messages left by its designers, then any experiment we perform will return the same results whether we’re in a simulated cosmos or not. In this scenario, there’s no way we can ever tell we’re in a virtual world, no matter how convincing our favourite philosophers are on the matter. The big-T Truth of the matter might be that we dwell in a simulation but, like the existence of an impersonal god, this fact has no bearing on how we conduct our lives.
We should also consider the possibility that we live in a simulation, but that the laws governing it are different to those of the world of the programmers. After all, scientists generate models all the time that don’t correspond directly to the real world but help refine our theories. And if such a simulation is an imperfect emulation, there might be places where the computer code shows its presence. If the Universe is a numerical simulation similar to those run by modern nuclear physicists, then there might be a point where the program’s necessary simplifications are at odds with the predictions of fundamental physics.
Consider atomic nuclei, which are made of protons and neutrons that are themselves made of quarks. The whole mess requires understanding the nuclear strong force that binds everything together, but the complex interactions have no consistent treatment of the kind for free particles such as electrons. However, it’s often difficult for physicists to calculate interactions between more than two particles at a time, especially at the high energies involved inside nuclei.
Instead of allowing them to move just anywhere, nuclear physicists act as though the particles reside on a three-dimensional lattice, like atoms in a solid crystal. Because energy increases as the quarks get closer together, forcing them to stay apart by a fixed distance keeps the numbers manageable — and still reproduces the behaviours we see experimentally. This type of numerical calculation is known as lattice quantum chromodynamics (LQCD).
While the simplifying principle in LQCD is the only consistent way they’ve figured out how to describe quarks, it violates the principle of relativity as set out by Albert Einstein. Spacetime in relativity is a continuum, with no special directions defined. On the other hand, a lattice such as the one in LQCD has special points and special directions (along the connections between the nodes). If high-energy collisions such as those produced by cosmic rays exhibited behaviour more like LQCD than like the predictions of relativity, it could be a sign we’re in a simulation where the programmers cut the same corners as modern nuclear physicists do.
Silas Beane and colleagues at the University of Bonn in Germany considered other testable deviations along these lines (including some anomalous behaviour by the electron’s heavier cousin, the muon). However, there are several possible ways their scheme won’t work. Whoever wrote the simulation might not use the same type of code nuclear physicists do, meaning that the predicted deviations won’t show up. The deviation might also happen at such high energies that we won’t discover them in the foreseeable future. Lastly, spacetime might behave like a lattice for reasons other than living in a simulation, a possibility seriously considered by a number of physicists.
In fairness, Beane, Davoudi and Savage, the nuclear physicists who proposed a way to test the simulation hypothesis, know all this, and it would be a mistake to think that this is the focus of their life work. If you look at Beane’s bibliography page on the INSPIRE repository (the high energy physics information system), you’ll see that this paper is the only one he has yet written on the subject; the rest involve standard LQCD research. While I’m sure he and his colleagues take the cosmic simulation work they did seriously, they’re likely typical of most researchers: they might find these questions interesting, but they won’t devote their lives to investigating the answers.
Partly that’s pragmatic: you can get funds for working within the standard paradigms of modern physics, but it’s harder to pay for research into what could be construed as open-ended philosophical questions. However, the problem itself is far too slippery to offer a tangible pay-off. Despite the impression one can often receive from reading popular science accounts, there’s little chance of success in devoting your life to the biggest questions about life, the Universe, and everything. The reason major breakthroughs (like the quantum mechanics revolution of the 1920s) are rare is because they’re hard. Science is mostly incremental progress, and that’s not a bad thing, even if it might seem unglamorous.
The difficulty with probing into the cosmos-as-simulation is finding the right scientific questions to ask: the ones that lead to testable consequences. In a hypothetical simulated Universe where the program manager can step in and fix problems in real time, we might not be able to distinguish between a real cosmos and an emulated one. The same applies to a simulation without any detectable imperfections. Even a compelling philosophical argument in favour of us living inside a computer program seems empty if we can’t obtain experimental evidence to back it up.
Do we live in a simulation? My gut feeling is no, and not just because I don’t want to believe in the existence of an intelligence who is either indifferent or who programmes beings to suffer needlessly. (Why not simulate a paradise?)
The power of science often lies in its generalisations, its abstractions, and even its simplifications. Simulating an entire Universe with sufficient detail to include conscious minds will be complex, even if the fundamental rules underlying the program are simple. It seems needlessly baroque to programme something as complicated as that, when you can learn just as much from something simpler.
However, those are intuitive musings, which might or might not prove valid. A better refuge is empiricism, unromantic as it is. From a scientific point of view, if we cannot distinguish between a simulated and real Universe, then the question of living in a simulation is moot: this reality is ours, and it’s all we have.
21 January, 2014