Are we all living in a computer simulation? This bizarre question seems to be growing ever more popular, and understandably so. For how can we exclude that we might? Indeed, given a few assumptions that most of us accept, does it not actually seem almost certain that we are?
The following is an excerpt from my latest book, The Meaning of Life: An Examination of Purpose, where I argue that it is far from certain that we are living in a simulation. On the contrary, it is almost certain that we are not.
The simulation hypothesis was put forth by philosopher Nick Bostrom, and it states that the world we live in may be a simulation in a computer constructed by an advanced human civilization, which is to say that the world we live in is purposefully created. The simulation argument, not to be confused with the specific simulation hypothesis, states that if we are not very likely living in a simulation, it must either be likely that humanity goes extinct before we become technologically advanced enough to make such simulations, or it must be true that we will choose not to make such simulations. The argument rests upon the assumption that it will become possible to create simulated worlds with simulated, conscious people, and if such simulations will ever be made, the number of simulated worlds and people might easily outnumber the number of real, non-simulated worlds and people, and it will therefore, given that we make such simulations, be likely that we live in such a simulated world according to the argument.[i]
The first thing to note about this argument is that its premises are based on the world we observe. It does not refer to a magical notion of purpose that is detached from what we know about it, and, furthermore, it is based on premises that can perhaps reasonably be considered plausible given what we know about the world. While this is a strength of the argument – basing our arguments on the data we have available to us is ultimately the only reasonable thing to do – it is also one of its greatest weaknesses given what the argument is about. For if the world we live in indeed is a simulation, then how can we know that the conclusions and worldview we may be able to reasonably draw based on what we see in this world, including conclusions about simulations, all of which is pure simulation according to the argument, have any validity beyond our own simulation? If we are living in a simulation created by simulators, what we think we can reasonably say about our simulators can easily be wrong to an unimaginable degree. What the simulation hypothesis secretly assumes is an enormous degree of similarity between the simulated world and the simulating world. This assumption is sought justified with the suggestion that a technologically advanced human civilization likely will want to simulate a world close to its own, a so-called ancestor simulation, and hence the great similarity. This is, however, merely to make the same assumption again: that the simulators resemble us in the first place. If we really live in a simulation, how likely is it, given how many different kinds of “worlds” that can be created in a simulation, that the world in which our world is run as a simulation is anything like our own?
As a general matter, speculations about a world beyond, or behind, our own are bound to be highly speculative. Because, first, we do not know whether such a world exists, and, second, even if it does, many things we know from the world we observe cannot necessarily be applied to say anything about other worlds. This is one reason to seriously doubt the simulation hypothesis: it rests on the assumption that the world in which the simulation is supposed to take place is much like our own, and given the vast space of possible simulations, this seems fantastically unlikely. So, strangely, if we accept the hypothesis that we are living in a simulation, the purported basis of the simulation hypothesis itself seems of questionable validity, and that is rarely a good sign for a hypothesis – when it seems to pull the rug under its own feet.
There are better reasons to doubt that we should be living in a simulation, though. For the simulation hypothesis also rests upon the assumption that we will one day be running realistic simulations of our past. Yet just how likely is it that we, ourselves and our descendants, will run such functional, conscious copies of our own past? This is finally a question that does not borderline on the extremely esoteric, as it relates directly, and exclusively, to the world we know and can know, and we should therefore – unlike when it comes to answering the question about whether a world wherein our world is simulated would resemble our world the slightest – at least be able to provide some sort of hint as to what the answer to this question might be.
Whether we will ever run such simulations first of all depends on whether it is even possible to reliably simulate conscious minds as our own on a computer, and it is not at all clear that it is. Even if our minds are so-called substrate independent, meaning that they need not necessarily be implemented in neurons or any other specific material, we do not know at what level our mind emerges, or whether speaking about a single level, or just any number of levels, even makes much sense in relation to how our minds emerge. We do not know what physical processes, or “information processes,” that are necessary in order to accurately emulate our conscious minds, and it is not clear how we can come to know this in the future.[ii] Furthermore, even if we overcome epistemological challenges of this sort and somehow manage to acquire knowledge about the specific physical processes that are sufficient in order for a conscious mind like our own to arise, this does not imply that we will ever be able to actually simulate such a mind. Even our potential future brilliance considered, there could easily be insurmountable practical and theoretical problems. For instance, some functions of conscious minds may be impossible to simulate, just like some mathematical equations, in fact most real-world related ones, have proven impossible to solve analytically. Or, to make a much closer analogy: just like computer simulations of certain continuous functions cannot be exact simulations of these, but only discretized approximations of them. Might our conscious mind not be such a continuous phenomenon – or at least a phenomenon that depends upon underlying, ultimately “unsimulatable,” continuous phenomena? It is not easy to see how we can gain certainty on this question.
Another objection to the simulation hypothesis concerns our technological and moral future. When we consider what running a functional and conscious simulation of our own past in fact entails, and when we consider the likely future of our own species, it seems incredibly unlikely that our descendants or future selves will ever run such a simulation. A fully conscious simulation of our own past up until this point entails an unspeakable amount of horror and suffering: wars, famines, torture etc. – an ocean of suffering in which the horrors brought to the world by Hitler and Stalin are mere droplets. Any individual or civilization that would intentionally create all this suffering – holocaust upon holocaust of suffering – for the mere sake of curiosity would be evil to an extent that is unmatched in human history. Given our technological and moral progress up until this point in history, it seems very unlikely that we will ever create such an atrocity that dwarfs all other humanly created suffering by recreating it all, and far more on top of it.[iii] It does not seem remotely consistent with our history of moral progress, or even with our embarrassingly primitive moral stance of today, that we would decide to unleash such horror.
But who is to say that a lone madman could not create this unedited copy of the world with all its unspeakable horrors? This question relates to the future of technology, more specifically, the technological power of our civilization to prevent lone madmen from creating and proliferating suffering. Unfortunately, it is not guaranteed that we will be able to effectively prevent people from being bad in the future, yet given the trend of our moral and technological advance, it does not seem entirely unlikely that we actually will be able to prevent people from bringing about immense amounts of suffering. This trend seems to converge toward a morally and technologically advanced civilization, and such a civilization seems, and hopefully will be, increasingly competent at stopping bad agents from being really bad (and, again, we are talking really, incomprehensibly bad, as this deed of simulating our past would be worse than any genocide; it would be the recreation of all genocides in the world to date, and they would still only comprise a tiny fraction of all the suffering unleashed). And in the case that we will not manage to prevent madmen from being mad, simulations of our entire history full of suffering actually do not seem likely either, because if we will not be able to prevent madmen from creating such moral atrocities in the future, it does not seem improbable that we will also be unable to prevent them from destroying civilization – an atrocity that one could actually argue is less bad than recreating all the past horrors of the world. The same level of moral scrupulousness and technological skill seems required. In fact, it may even take less of both in order to destroy civilization, since, again, one can argue that the destruction of civilization is the more merciful action, and since the destruction of the world as we know it might well be an easier task than the functional recreation of it.
So given that some people will be able, both morally and technologically, to make simulations of the original world with all its suffering, the fall of civilization could also be dangerously likely, and perhaps far more likely, which would in turn render such simulations rather unlikely themselves, as they require civilization to exist – at least highly advanced computers provided with massive amounts of energy. Either way, whether civilization persists or not, simulations of worlds with great amounts of suffering, as ours indeed contains, do not seem likely to be performed, at least not for long, and therefore a world full of suffering does not seem likely to be a world created by a human civilization like our own that wanted to simulate their past, but rather the original one itself.[iv]
The final blow to the simulation hypothesis is provided by the impossibility of simulating anything in the world to total precision. The world is chaotic and contains far more information than what can ever be included in any simulation contained within the world, and all its information is crucial if one is to make an exact copy of the universe and its past. Bostrom argues that this does not undermine the simulation hypothesis, as one need only simulate the minds of our ancestors, not the entire world, or at least not the entire world to any great precision. This is wrong, however, as this defense undermines the very premise of the argument: that we will run ancestor simulations. Because if one only simulates a simplified model of our past, then one is not in any meaningful sense running an ancestor simulation, but instead something completely different. Even if one had near-perfect information about a given state of the entire world, which we do not have, and cannot possibly attain, a simulation based on such incomplete information would not result in a functional simulation of our past. It would result in something completely different, and it would be increasingly different for each iteration. One small deviation, and the world falls out in a completely different way – one small change 2000 years ago, and a Caesar is not born; one small change a million years ago, and the entire human species never arises; one small change in the early universe, and the Milky Way Galaxy, including our sun and earth, never forms. Those who run the simulation could then continually try to realign the simulation with what they know about documented history, but that information would still get them nowhere in terms of making an accurate simulation of our past. First, because they will have limited information to correct the simulation with – they may know the historical name of Julius Caesar (or maybe not?), but that will hardly get them far in simulating his mind, much less all the other people who lived in his time, most of whom we have effectively no information about – and this is an example from recorded human history; good luck with assembling our galaxy in the first place. Second, even if the simulation could be continually edited to get back on track, the problem would reappear as soon as the simulation runs again: because it is not an exact copy of our actual past, it will quickly unfold differently from our own. It would be a never-ending editing job that would lead to nowhere. And it would certainly not be a simulation of our past or anything close to it. More objections can be, and have been, leveled at the simulation hypothesis, but we have put enough forth at this point.[v]
The objections to the simulation hypothesis brought forth here of course do not definitively disprove that we may live in a simulation, in the same way that facts about the origin of the moon and cheese do not definitively disprove that the moon could contain significant amounts of cheese; it is just an incredibly unlikely hypothesis that is too unlikely for it to be taken seriously. Each of the objections above seriously question the truth of the simulation hypothesis, and, collectively, these objections do make a strong case that the simulation hypothesis is almost certainly false.
[i] This is a short summary of the argument. For the full argument, see http://www.simulation-argument.com/simulation.html.
[ii] This is not to say that we will not be able to know a lot about how physical states relate to conscious states – it seems clear to me that the ultimate goal for a science of consciousness must be to make such correlations and to establish the relationship as accurately as possible. However, when it comes to what the physical processes, or perhaps information processes, that are necessary and sufficient for conscious minds to arise – which aspects can be left out and which cannot in relation to maintaining the same, or just any, conscious experience – the problem of other minds becomes more than just an annoying mind game.
[iii] The same point has been made by David Pearce: http://www.hedweb.com/hedethic/hedon4.htm (No 35).
[iv] This point has also been made by David Pearce: http://www.hedweb.com/hedethic/hedon4.htm (No 35).
[v] An examination of the simulation argument that also seriously questions its likelihood, partly based on some of the same points made here along with others, can be found in the following piece by David Pearce: http://www.hedweb.com/hedethic/hedon4.htm (No 35).