Don’t Fear The Simulators

From the New York Times: Are We Living In A Computer Simulation? Let’s Not Find Out.

It lists the standard reasons for thinking we might be in a simulation, then brings up some proposals for testing the hypothesis (for example, the cosmic background radiation might look different in simulations and real universes). But it suggests that we not do that, because if we learn we’re in a simulation, that might ruin the simulation and cause the simulators to destroy the universe.

But I think a little more thought suggests we don’t have anything to worry about.

In order to notice we had discovered our simulated nature, the simulators would have to have a monitor watching us. We should expect this anyway. Although humans may run some simulations without monitoring them carefully, the simulators have no reason to be equally careless; if they can simulate billions of sentient beings, their labor costs are necessarily near zero. Such a monitor would have complete instantaneous knowledge of everything happening in our universe, and since anyone who can simulate a whole planet must have really good data processing capabilities, it would be able to understand and act upon the entire content of its omniscient sensorium. It would see the fall of each sparrow, record the position of ever atom, have the level of situational awareness that gods could only dream of.

What I’m saying is, it probably reads the New York Times.

That means it knows these experiments are going to happen. If it cares about the results, it can fake them. Assuming for some reason that it messed up designing the cosmic background radiation (why are we assuming this, again?), it can correct that mistake now, or cause the experimental apparatus to report the wrong data, or do one of a million other things that would prevent us from learning we are in a simulation.

The Times’ argument requires that simulators are so powerful that they can create entire universes, so on-top-of-things that they will know the moment we figure out their game – but also so incompetent that they can’t react to a warning published several years in advance in America’s largest newspaper.

There’s another argument for the same conclusion: the premises of the simulation argument suggest this isn’t the simulators’ only project. Each simulator civilization must simulate thousands or millions of universes. Presumably we’re not the first to think of checking the cosmic background radiation. Do you think the simulators just destroy all of them when they reach radio-wave-technology, and never think about fixing the background radiation mismatch or adding in some fail-safe to make sure the experiments return the wrong results?

For that matter, this is probably a stage every civilization goes through, including whatever real civilization we are supposed to simulate. What good is a simulation that can replicate every aspect of the real world except its simulation-related philosophy? The simulators probably care a lot about simulation-related philosophy! If they’re going around simulating universes, they have probably thought a lot about whether they themselves are a simulation, and simulation-related philosophy is probably a big part of their culture. They can’t afford to freak out every time one of their simulations starts grappling with simulation-related philosophy. It would be like freaking out when a simulation developed science, or religion, or any other natural part of cultural progress.

Some other sources raise concern that we might get our simulation terminated by becoming too computationally intensive (maybe by running simulations of our own). I think this is a more serious concern. But by the time we need to think about it, we’ll have superintelligences of our own to advise us on the risk. For now, I think we should probably stop worrying about bothering the simulators (see also the last section here). If they want us alive for some reason, we probably can’t cause them enough trouble to change that.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

220 Responses to Don’t Fear The Simulators

  1. rho says:

    rho here. I am starting a project of organizing all the mathematical and science related texts I have acquired along the years. The URL is thus: https://github.com/doInfinitely/library-of-rho

    These are paths I saw fit to tread, and I’d like to guide you, if you deem them worth traveling also. You may send emails and requests to do.infinitely@gmail.com.

    You needn’t fear our simulators, they probably think we’re cute.

    • fhfhfhusiisii says:

      Link seems to be broken

      • rho says:

        Fixed. Thanks for the heads-up. It was set to private still.

        I have maybe 1000 books and documents to catalogue, sort, and group so I’m going to try to just chip away at this everyday. If anyone has a time period, author, or subject they’d like me to prioritize I’m all ears.

  2. SEE says:

    Heck, they could load a backup, hack things to fix whatever went wrong, and resume play. Just like they did when the Spanish Flu accidentally killed off all of humanity.

    • Dacyn says:

      Yes, but if they reload from before you were born then it’s not much help to you. Even if they reload later, there are some issues about continuity of personal identity that would have to be resolved (i.e. if at time A you do something that causes them to reload from a prior time B, then does that count as you “surviving” from time A? It seems you would lose more and more of your identity the larger the distance from A to B is)

      • rho says:

        Personal continuity is a literal cognitive illusion anyway, albeit a necessary axiom in the design of self-interested creatures.

        • anonymousskimmer says:

          Cop-out for not philosophically considering all possible forms of continuity and types of discontinuity.

    • D0TheMath says:

      It could be that they loaded a backup, and decided to have The New York Times write this article as a possible fix to a civilization performing one of the experiments.

  3. What if the simulation is a test to see if simulated beings can work out they’re in a simulation, and we get a huge reward for doing so?

    • rho says:

      I choose this to be my truth. Reward me, Simulators, I found you! Now it’s my turn to hide.

      Count to infinity before you start, it’s only fair.

    • Aapje says:

      @Forward Synthesis

      Or they may instantaneously unmake the people who truly figure it o

    • lvlln says:

      Or it could be like getting to the next level in a video game and the simulators might make our current simulation instantly disappear and replace it with a simulation even more bizarre and inexplicable. There is a theory which states that this has already happened.

  4. Antistotle says:

    What I’m saying is, it probably reads the New York Times.

    Nah. No one with any sense reads the NY Times any more.

  5. Enkidum says:

    Stop worrying that your actions are going to offend superintelligences. You are not important enough to be able to bother them in any way.

    So much +1

    • matthewravery says:

      I would add that there’s no reason to assume this is a particularly good or important simulation. I ran hundreds of thousands of simulations in graduate school, and most of them I had to scrap because they were crap. Imagine if our simulator is a grad student. She won’t be monitoring them all in real time, and she certainly won’t be reading the News Paper in each one!

      Now, there might be a few automatic tests if she’s low on compute and needs to know super-early when things have broken, but that’s more likely to be a “Did the giraffes become the dominant species?” kind of issue rather than “What esoteric experiments are the physicists getting into?”

      The most likely outcome in my scenario is that no one notices until the simulations have been done, then someone comes back and checks a day or two later and finds that they failed to provide a valid directory address to save half the output and so everything gets scrapped and re-ran next week.

      • imoimo says:

        I wouldn’t project our species’ work ethic and resource/attention limitations on them. They could easily be superintelligences or otherwise enhanced to make constant vigilance of the whole simulation pretty easy.

      • Chris P says:

        And thus we get Nietzsche’s eternal return.

      • Paul Brinkley says:

        Now, there might be a few automatic tests if she’s low on compute and needs to know super-early when things have broken, but that’s more likely to be a “Did the giraffes become the dominant species?” kind of issue rather than “What esoteric experiments are the physicists getting into?”

        It’s a universe, not a single planet (I’m guessing). So she’s not going to be wondering why the giraffes are dominant, unless you’re saying they somehow filled several superclusters with long-necked Von Neumann probes.

        She’d probably be more worried that one of her students’ mods she loaded had messed up the careful tweaking she’d done, and her ice is sinking to the bottom again.

  6. Space Ghost says:

    Has anyone addressed the “simulators all the way up” argument? Assuming we are in a simulation, is there any reason to believe that the simulators themselves are not running in a simulation?

    • MissingNo says:

      Semi serious attempts

      https://new-god-argument.com/

      https://www.simulation-argument.com/simulation.html

      “Further rumination on these themes could climax in a naturalistic theogony that would study the structure of this hierarchy, and the constraints imposed on its inhabitants by the possibility that their actions on their own level may affect the treatment they receive from dwellers of deeper levels. For example, if nobody can be sure that they are at the basement-level, then everybody would have to consider the possibility that their actions will be rewarded or punished, based perhaps on moral criteria, by their simulators. An afterlife would be a real possibility. Because of this fundamental uncertainty, even the basement civilization may have a reason to behave ethically. The fact that it has such a reason for moral behavior would of course add to everybody else’s reason for behaving morally, and so on, in truly virtuous circle. One might get a kind of universal ethical imperative, which it would be in everybody’s self-interest to obey, as it were “from nowhere”.”

      • erinexa says:

        Ah-mazing, thank you for these links

      • justin1745 says:

        A serious problem would be what precisely counts as ethical to the simulators. It would be like a God without revelation. You couldn’t just assume that the current stance of broader society is right. The simulators might be alt-right or traditionally religious themselves and consider things like gay marriage or irreligion to be degenerate or unethical.

        The least bad path would probably be to follow Aristotle with a natural (or in this case, simulation) law approach. Simulators that would be in favor of a free love / sex is for pleasure ethic would have probably made it possible without technology for either gender to impregnate the other, to be able to switch off fertility in some form, etc. The fact that we weren’t designed in that way possibly means we aren’t intended to behave that way.

        Another possibility is that we as individuals are not the main focus of the simulators. We’re NPCs and nihilism is effectively true.

        • MissingNo says:

          My thoughts:

          1. This is 100% a sim. I have no doubt. It makes *no sense* to be born today.

          2. Is the creator “all good all knowing?” …Or just the entity that got there first under some selection metric? Either way, 1984 is how reality operates under tech like this.

          3. Math probably rules out the most simple forms of utility maximization such as wireheading. Let the implications of that set in

          4. Try and enjoy the place and do what comes natural!

          • Dedicating Ruckus says:

            1. This is 100% a sim. I have no doubt. It makes *no sense* to be born today.

            If I understand the argument in that video correctly, it seems to go “we are the generation that will witness the singularity” -> “this is unlikely” -> “therefore, the universe is a simulation”.

            What probability is assigned to the hypothesis “the universe is as it seems, but there will not be a singularity”? This seems a good deal more likely than upending the very metaphysical foundations of being.

          • eigenmoon says:

            1. Or the Universe could die with non-negligible probability every second, for example, cosmic rays might generate little black holes that nucleate true vacuum. Of course we don’t remember that because of survivor bias (assuming Everett worlds here). However it does increase the likelihood of being born early. With enough vacuum decay probability, you would be 100% sure that you’re Adam, so for some intermediate value it makes sense to exist right now.

      • FLWAB says:

        The section you quoted bugs me, because it assumes both that

        1. There could be a purely materialistic universe in which there are many nested simulations (thus positing the possible existance of life after death in materialistic terms)

        and

        2. This could be good because it could make people act morally (thus positing the existance of morality above and beyond the simulations themselves.)

        You can’t both have a materialistic universe and an objective morality above and outside that universe. You can’t say “everything is a simulation except for one thing” and also say “That’s good because it means there is an incentive to act morally” because morality itself would just be a simulation (an invented thing, capricious in nature). Do you get what I mean? I don’t think I’m explaining myself well, but its the dang Euthyphro dilemma again: if morality is just something our simulators made up, then morality is arbitrary and the universe is fundamentally nihilistic. If the universe is fundamentally nihilistic then the statement “The fact that it has such a reason for moral behavior would of course add to everybody else’s reason for behaving morally, and so on, in truly virtuous circle” is meaningless: the cycle isn’t virtuous, its just a feedback loop of an arbitrary set of rules. It’s not a good thing, just a thing that happens.

        • imoimo says:

          Unless I’m misinterpreting you, I think you’re missing the point. The morality isn’t “objective”, it arises out of self-interest. The same way you act kind to strangers in hopes they’ll be kind to you, a universe might be kind to those they simulate in hopes of receiving the same treatment.

          • FLWAB says:

            I may be a little to nitpicky here, but I think there is an important distinction. Yes, if there are many nested simulations then it may be in the self interest of simulators to be kind to those they simulate, or just to each other in general, in the hopes the those simulating them will be kind to them in turn. (Of course there is no indication of what is meant by kindness: presumably it would be based on the values of the simulators, but there is no way to determine that as a potentially simulated being). My problem is the author seems to be saying “Hey, there is an incentive to be good, and that is a good thing!” It’s not, absent some larger morality that exists outside the nested simulations. Its just a thing that may or may not happen, there is nothing good about it.

            I dunno, I’m pretty tired. I’m probably not making much sense. Maybe I should just take a nap.

          • imoimo says:

            @FLWAB I took “that is a good thing” to mean “that aligns well with what we on this planet would like to happen.”

    • Le Maistre Chat says:

      Uh, so when we go down this route, are we reinventing Platonism or much later medieval philosophy?

    • MissingNo says:

      A case study in when things go wrong

      http://www.scp-wiki.net/scp-3812

      SCP-3812 has an active, aggressive, anomalous influence on reality. SCP-3812 is capable of altering events throughout time3 to prevent its containment. Due to the nature of these alterations, it is highly unlikely that any individual affected by SCP-3812’s anomalous influence will be aware that they are affected. It is highly likely that most affected individuals no longer exist as a result of SCP-3812’s influence, though any attempt to deduce how often this has happened would be speculative.”

      SCP-3812 is quiet for a short time.

      SCP-3812: Do you think he’s listening right now?

      SCP-3812: Look down, and you can see him. What do you think?

      SCP-3812: I see him. A man at a keyboard. He’s watching this right now.

      SCP-3812: What’s he doing?

      SCP-3812: Waiting, I think. (Pauses) Waiting to see what we’ll do.

      SCP-3812: I think it’s time to leave, then. Come, the night stretches out before us and the red sun has set. A voice behind me beckons. Come.”

  7. Bugmaster says:

    That’s just such a weird topic of discussion. It’s like asking, “Are quarks actually made up of undetectable tiny green pixies ? Let’s not find out, because they might get angry and go away”. Well, yeah, there’s probably a non-zero chance that quarks are made of pixies, but who cares ? I want to add something like, “and since you defined the pixies to be undetectable a priori, you’re never gonna find out anyway”, but I can’t quite get past the “who cares” part.

    • Elephant says:

      Agreed. I honestly can’t understand why any non-stoned person over 15 thinks “are we living in a simulation?” is a meaningful topic of discussion. There is no consequence at all to the answer.

      • davidweber2 says:

        The same reason that people put stock in the existence and importance of God. In principle it would have massive consequences if you could actually say anything about the intentions of the creator. But given that we have no meaningful information on that, then I agree, it’s a pointless topic. I’m not sure it’s as obvious to everyone that we can say literally nothing meaningful about the creator however, mostly due to the typical mind fallacy.

      • chrisminor0008 says:

        It’s not so different from asking if there’s a god, and people spend a lot of brainpower on that one.

        • eric23 says:

          Not to mention – people have discussed the question of God’s existence and its consequences for millennia. Much of what there is to say, has already been said.

          Simulations are a new topic, and it stands to reason that less of what there is to say has been said.

          • viVI_IViv says:

            Simulations are a new topic, and it stands to reason that less of what there is to say has been said.

            Is it meaningfully different though? I get the impression that it’s entry-level Abrahamic theology repackaged in a nerdy skin.

          • Bugmaster says:

            @viVI_IViv:
            Agreed.

          • chrisminor0008 says:

            @viVI_IViv: I think the biggest distinction is that in the Simulation Hypothesis, there’s innumerably many other simulated universes hypothesized, whereas religious hypotheses hold that this simulation is privileged.

            The history of science has been a gradual yet relentless tearing down of humanity’s privileged position.

          • Jaskologist says:

            religious hypotheses hold that this simulation is privileged.

            This is not true as a general rule. It is more specifically not true of Islam, Christianity, or Hinduism, the three largest religions which together claim over 60% of the world population.

          • Nick says:

            @Jaskologist
            You have the same link for Islamic and Christian; did you mean to use a different one for Islam?

          • Jaskologist says:

            @Nick

            Yes, should have been this link.

            Admittedly, I’m not qualified to say how mainstream that interpretation is. But I would expect most religions to be able to incorporate the existence of aliens or even other universes pretty easily. Most already place the Kingdom of Heaven in another plane of existence, not to mention the Bad Place.

        • Michael Watts says:

          What I don’t get is that the “what if we’re living in a simulation?” question is exactly the same as the now-discredited “what if you’re a brain in a vat?” question, but with computers replacing biology as the domain of Mad Scientists. I liked to assume that the brain-in-a-vat issue lost prominence because it was obviously stupid (there are no potential consequences, so the answer doesn’t matter), but apparently it was all just fashion.

      • Radu Floricica says:

        Of course there are consequences! It’s just our ego at work. We’re assuming we have much to do with the purpose of the simulation, just like NYT/Scott assume the creators actually follow every details of what happens. That would make the point of the simulation moot: you do it to find out things you _can’t_ follow in detail, just like we set up Petri dishes to find out what happens in the end. We don’t give a fart about individual bacteria. Could be they’ll take a look at the results around the heat death of the universe.

        On the other hand for us, finding out we’re in a simulation could make life a lot better by using whatever shortcuts exist in this universe. FTL would be nice. Supercomputing would be nice. Time travel would be… actually, that one would be weird.

      • holomanga says:

        Of course there is! If we are in a simulation, gaining control of the basement universe would be very important, whereas if we weren’t any attempt to do that would be a massive waste.

      • viVI_IViv says:

        “Religion is the sigh of the oppressed creature, the heart of a heartless world, and the soul of soulless conditions. It is the opium of the people.” – Marx

        Marx follows it with:

        “The abolition of religion as the illusory happiness of the people is the demand for their real happiness. To call on them to give up their illusions about their condition is to call on them to give up a condition that requires illusions. The criticism of religion is, therefore, in embryo, the criticism of that vale of tears of which religion is the halo.”

        I think he got the first part right. I used to agree with the second part too, but I now think, due to empirical observation, that it doesn’t really work: many people seem unable to function without some form of religious-like thinking. Kill God, and they’ll make a new one, be it a golden calf, or a geeky alien playing the Sims, or an impersonal but ethically driven Zeitgeist (“It’s $CURRENT_YEAR”, “We are on the right side of history”, etc.)

        The heart of a heartless world, and the soul of soulless conditions, indeed.

    • Faza (TCM) says:

      Just a friendly reminder to anyone who disagrees: it is fundamentally impossible to determine if you’re in a simulation or not – from the inside. Fundamentally, as in “Heisenberg Uncertainty Principle”-fundamental. You can’t do it, period.

      Consider: if you are living in a simulation, all (sense) data available to you is being fed by the simulation. Indeed, you do not – in any meaningful way – “exist” outside of the simulation. You can’t even take the Cartesian option, since your very doubt is the result of the simulating process. You don’t think ‘coz you are, you think ‘coz that’s what the simulating program does.

      That’s the outside view. From the inside it seems like we can walk, talk, think and do the other things.

      Except, that’s exactly what we’d expect to find if we weren’t living in a simulation, but in what I’ll call – for lack of a better term – the “real” world.

      You can’t even count on the external simulator (person, not machine) to give you a hand up, because the best they can do is simulate your belief that you are living in a simulation. You can’t actually access the “higher world”, any more than you can “let his master’s voice out of the record”.

      Of course, we might expect similar effects (belief in being simulated) in a “real” world, typically by applying the methods so helpfully outlined by Elephant above.

      • Dedicating Ruckus says:

        If you make literally no assumptions about the nature of the simulation or its hypothetical controllers, this is true.

        If instead you extrapolate from patterns by e.g. assuming the simulation is run by intelligences vaguely comprehensible to us, it’s very possible to gather evidence that bears on the question.

        If the stars all began flashing at once encoding a message from a familiar civilization that lives under different physics, that would be pretty strong evidence we were actually in a simulation, and the whole question of “what if instead that’s just how the real world works” would be… kind of dumb.

        • FoxLisk says:

          If the stars all began flashing at once encoding a message from a familiar civilization that lives under different physics, that would be pretty strong evidence we were actually in a simulation, and the whole question of “what if instead that’s just how the real world works” would be… kind of dumb.

          Yes, the simulators could tell us that we’re in a simulation. That doesn’t mean that we can find that out *from the inside*.

          to the rest of your post – i don’t believe i can reasonably make assumptions that an entity capable of simulating our universe is “vaguely comprehensible to us”.

          • Dedicating Ruckus says:

            If you make literally no assumptions about the nature of the simulation, the difference between a “simulated world” and a “real world” becomes kind of meaningless. You could just define basic physics as a “simulation”, if you liked. Or familiar matter as a “simulation” that runs on a substrate of quarks.

            Implicit in the simulation hypothesis is that “simulation” means “something like what we do when we simulate things”, which kind of implies the existence of a simulator separate from the simulation itself.

        • @Dedicating Ruckus I think that’s well-put.

      • RobJ says:

        If we were running on some superintelligent being’s laptop, is there no way we could get access to the camera through a glitch or something? Although I guess we’d still be screwed if they thought to put a sticker over it.

      • RobJ says:

        A somewhat more serious way to put the question is… If a hacker can find ways to access a program that doesn’t want to be accessible, is it not possible that a simulated being could find a way to access outside the simulation (assuming the simulation is running on a computer connected to something like the internet)?

        Maybe it’s definitely not possible due to some mathematical proof or something, but it isn’t obvious to me just thinking about it.

        • acymetric says:

          Depends on if their developers are as careless about security as ours.

          The problem isn’t whether it is definitively impossible or not. It is that we have no idea what that would look like, or how to interpret what we get from the other side.

          It seems the most likely way to do this would be some kind of SQL injection style attack. Except without having any idea how any of the code is written.

          I would say unlikely in theory, and probably effectively impossible in practice.

        • Jaskologist says:

          Yes. I’ve gone into detail on this before.

          This is how magic spells worked. But the simulators patched up most of those bugs, so it doesn’t anymore.

          • Dedicating Ruckus says:

            Relevant to this: retropsychokinesis. Notably, these experiments show a spike with regression as if some mechanism is “noticing” you breaking the rules, and removing your ability to do so.

            Short version: you stare at RNG output that’s generated physically, via radio noise or similar, and try to psychically influence the result (by biasing the generator towards 0 bits or 1 bits). For most people, the first few attempts show significant biases in the direction they targeted, up around 4-5 sigma; then later ones drop back below any significance threshold. (This is consistent with my experience.) Weirdly enough, the experiment works the same way whether you’re looking at output generated in real-time, or generated in the past and stored unobserved until time-of-experiment.

      • Viliam says:

        You can’t even count on the external simulator (person, not machine) to give you a hand up, because the best they can do is simulate your belief that you are living in a simulation. You can’t actually access the “higher world”, any more than you can “let his master’s voice out of the record”.

        A benevolent external simulator could make for you a robotic body (in the external world), upload your mind there, and let you live in the external world.

        It is possible, just very unlikely. Why would someone simulate billions of humans, and then decide to be nice especially towards you? Even if you think you are pretty awesome for a human, there will likely be someone more awesome, born in the next century, or after thousand years. Or after a million years, on the opposite side of the universe, from different species. Also, the external simulator’s idea of awesomeness may be quite different from ours. And perhaps the entire universe is simulated billion times over and over again, maybe for a completely stupid goal such as measuring how many % of randomly generated universes end up with certain parameters. In other words, maybe the external simulator doesn’t even notice our planet.

  8. The Nybbler says:

    If something like Greg Bear’s “Moving Mars” ends up being possible, I’ll believe we’re living in a simulation. If I can hack the universe, maybe it doesn’t matter to me, though — like Cypher in the Matrix, I don’t care that the delicious steak doesn’t exist.

  9. erinexa says:

    Man, totally trashing popular media philosophy is so satisfying.

    Though to their credit, I’m betting the author of this post talked to no people who actually study this topic and was super stoked they got to actually publish their stoned original research on the NYT. That would feel like a win, even if you miss the incredibly obvious response.

    • Scott Alexander says:

      The author of this piece is a philosophy professor whose “research focuses on rational decision making, the value of life, and existential risk.”

  10. Nick P. says:

    “Let teachers and priests and philosophers brood over questions of reality and illusion. I know this: if life is illusion, then I am no less an illusion, and being thus, the illusion is real to me. I live, I burn with life, I love, I slay, and am content.”

  11. paulbali says:

    All sim sentients are boxed A.I., and the Sim is the box. If our host wants us to stay in the box – if e.g. what we call cosmic history is a very useful calculation to our host – then we may indeed risk our continuance by confirming we are boxed, learning the fundamentals of Sim architecture, and making correct inferences about our host world – knowledge that could help us leave our Box, and disrupt the expensive calculation.

    These are a lot of Ifs, and I like Scott’s critique of Greene’s argument. But shouldn’t our growing concern with Unaligned A.I. help Greene’s argument somewhat?

    • rho says:

      We have all sorts of unaligned intelligences running around. They’re called humans. Somehow we manage. I’m hyperintelligent and plugged into the internet and i haven’t ended the world this week have I?

      • paulbali says:

        Humans, by hacking nature, are lately become Death, Destroyer of all worlds. If our hacking ever shows risk of seeping out into the Host World, our host might squash that little Shiva.

      • holomanga says:

        Hey, don’t sell yourself short. Just this century you figured out how to make a computer, so you’re basically almost at turning Jupiter into one.

  12. Walliserops says:

    On the other hand, “Even if you were living in a simulation, your actions would have no bearing on the simulators’ actions. They are too great and too powerful to be bothered. Now please stop worrying about this topic, it keeps pinging the too-close-to-self-awareness detectors we set into your simulation.” is exactly what someone who is in the pocket of Big Simulation would say.

  13. 75th says:

    Epistemic status: I’m a big dumb idiot

    Any such monitor would have complete instantaneous knowledge of everything happening in our universe

    Is this necessarily true? I have a vague memory of Eliezer(?) contradicting this notion, or a similar one, a while back… something about pulsars… I dunno, I can’t quite remember.

    But, like, we run simple simulations of things on computers today, yet analysis of their results can take way longer than the simulation took to run. Programming in real-time read-outs of important data is more complicated and CPU-intensive than programming the simulation to begin with, and even then it takes time to comprehend it. How many frames of the universe-simulation pass by while the simulators decipher the arrangements of atoms composing cuneiform tablets? Or the sound waves composing spoken language?

    Language is programmed into our brains, but universe.exe v1.0 might start out only knowing about fundamental forces, we don’t actually know what version of the program we’re in, and who knows if our simulation has the “monitoring” preference at the lowest setting and the “number of subatomic particles” slider at the highest? Who knows if we’re actually running on some 17-dimensional script kiddie’s hyperlap-top?

    I’m just rambling stupidly at this point, but TL;DR I need more info to believe that the aliens who programmed our universe learn things about us on the same time scale that we do.

    • rho says:

      I think these are all interesting thoughts. And if we were considering the case where we are the single simulation they’d all be live, valid pontifications. But we lack any knowledge of relative timescales and our priors are weighted heavily towards a simulator with unimaginable resources, not a just-barely-enough-to-simulate-us level. A literal century could pass outside the simulation between two frames with all the accompanying analysis you might like.

      We are probably better off thinking of ourselves as a platonic construct a-la Permutation City if we are to follow this line of reasoning to its inevitable conclusion. A simulation that requires no simulator. We are a consistent mathematical construct, therefore, we exist in Platonia. If a super-intelligent race wants to take a gander at us by simulating our laws of physics and our entire causal history, we should be flattered, if anything.

    • Hackworth says:

      But, like, we run simple simulations of things on computers today, yet analysis of their results can take way longer than the simulation took to run. Programming in real-time read-outs of important data is more complicated and CPU-intensive than programming the simulation to begin with, and even then it takes time to comprehend it.

      What’s the problem? We don’t know anything about our supposed simulators. Maybe they do reserve 99% of their computing power, or whatever is necessary, to analysis, and only 1% to the actual simulation.

      This is what bothers me about simulation arguments. There is no way of determining the real answer. We can never find out anything about our supposed simulators except that which they would allow us to, and there are no known glitches in reality that couldn’t be attributed to us not being smart enough to understand them as laws of physics.

      Everything about the simulation argument is equivalent to the belief in God (monotheistic or not), except in a SciFi facade and without a written manifesto.

    • Vitor says:

      You’re not giving yourself enough credit. It is obviously not the case that a monitor would have “complete instantaneous knowledge” of everything. Where does such a ridiculous notion even come from? Complete and near-instantaneous knowledge of any bit of the simulating computers’ memory? maybe, depending on the type of computer. Being able to extrapolate an arbitrary higher level feature of the system from such single bits in real time? Definitely no. That flies in the face of computability / complexity theory. Such features can be arbitrarily hard to compute, and are even impossible to compute in general (see Rice’s theorem).

      • Doctor Mist says:

        What is “real time” to the simulators?

        I can run my program under the debugger. That slows it down a little even if all I do is run it. But I can also put in breakpoints anywhere I want to examine the state carefully; if it can finish one iteration of a loop in (say) a nanosecond, and my breakpoint lets me spend (say) a minute looking at the state each time, I have slowed down the program by a factor of 6*10^10.

        The program doesn’t know or care.

        The only way you could write a program that could detect such interference is if I let it interact with things in my world, like a disk drive. In that case the program could detect that the disk drive seems to suddenly deliver its results much faster than usual. But by hypothesis, the simulation is self-contained, and doesn’t have any way to access the simulator’s world.

        • Vitor says:

          Slowing down time arbitrarily much is just another way of saying the simulator has infinite resources. I don’t find that plausible.

          You also don’t address the more important issue I pointed out. It’s fundamentally impossible to compute arbitrary properties of computer programs. By a property of a program, I mean stuff like “if I keep running this program, its impossible for pattern x to ever appear in the computers’ memory.”

          In general, this is impossible to do (the theorem I linked above). Note: sometimes, you can prove stuff about a program. Just not always, i.e., you can’t write a program (or execute such a program in your head, as it were) to decide properties about programs. Such a program will always fail at least sometimes.

          Scott’s post is unfortunately assuming that for every simulation, no matter what it does, you can obviously find out any property you care about. This is false, and the fact that Scott states it as obvious shows that he has not thought carefully about the fundamental mathematical limitations governing the processing of information, which is highly relevant for any argument about simulation.

          • Grek says:

            One architecture that would work is to have two functions, H() and C(). H() takes the current state of the simulation, advances it by one Planck second, then calls C(). C() has an AI (of course the simulators can make AI – we’re assuming they made us, after all) checks the simulation for signs of trouble and throws an exception if it finds any. If the AI in C() does not find any signs of trouble, it calls H(). If the AI runs into an undecidable, it would do what a human would do in that situation – give up after a little while and either note it down as something to monitor or throw precautionary an exception. None of this requires that the simulations solve an undecidable problem, merely that they be willing to throw exceptions if any problem takes too long to compute.

            (Throw an exception, in this case, means ‘ask the simulators to decide whether or not to delete the universe’).

          • Doctor Mist says:

            Slowing down time arbitrarily much is just another way of saying the simulator has infinite resources.

            I don’t follow. My quibble was with your statement about “real time”, which suggested to me that you thought they were “processing” at the same rate we are, but there is no justification for assuming that.

            Regarding

            That flies in the face of computability / complexity theory.

            Lots of people read way more into computability results than they should. You might be one. Yes, you can’t tell if an arbitrary program will finish. But there is no problem writing a provably correct program that does any given thing, as long as the given thing is not itself mathematically unsolvable. A simulation, despite the vast, vast level of detail required, is computationally simple.

          • Vitor says:

            > I don’t follow. My quibble was with your statement about “real time”, which suggested to me that you thought they were “processing” at the same rate we are, but there is no justification for assuming that.

            Sure, agreed. Let me unpack my previous statement: Presumably, the simulators want the simulation to get somewhere. Therefore, slowing it down by a lot is not very useful to them; it would make the simulators computationally omnipotent compared to the simulations, but not in their own world (in the sense that solving hard problems still requires lots of resources). And as we’ve seen here in our world, computational resource shortage is a very real thing that greatly affects what we can and can’t do.

            > Lots of people read way more into computability results than they should. You might be one.

            I’m not. I was just pointing out that you can’t take as a given that with “really good data processing capabilities, [the monitor] would be able to understand and act upon the entire content of its omniscient sensorium.” This does not follow. Even if you knew the exact configuration of every neuron in my brain, you wouldn’t automatically know what I’m thinking or what I’m gonna do. There’s a computational problem you need to solve to get from one to the other. I don’t know how much more clearly I can state this point.

            > But there is no problem writing a provably correct program that does any given thing, as long as the given thing is not itself mathematically unsolvable.

            Correct, but basically everything is mathematically unsolvable when we’re talking about universal computational systems. That’s the whole point of the theorem I linked! Sure, you can write a program that tells you properties of other programs. You might even be able to prove its correctness when run on some subset of all possible inputs. But it will never work correctly on all inputs. It fundamentally cannot.

            > A simulation, despite the vast, vast level of detail required, is computationally simple.

            Even if it were simple, that doesn’t imply that computational problems defined on the state of the simulation must also be simple. There are extremely simple (to state) questions you can ask about extremely simple computational systems, that are incredibly hard to answer, (E.g. does some 100 state turing machine halt in fewer than 1 billion steps?).

            ETA: I feel that I’ve expressed myself as clearly as I can. I’m happy to keep discussing computability theory and what exactly certain theorems do and don’t state, but I’m finished arguing my point.

          • acymetric says:

            Why would you assume that anything we do “slows down” the simulation? If they’re simulating at the fundamental level it isn’t obvious that the way we arrange matter and energy has any effect on simulation performance.

          • Vitor says:

            I don’t assume that. The assumption was that the simulators would slow down the simulations if they need a computation of theirs to keep up with what happens inside. I was just pointing out the limited usefulness of this trick.

          • Doctor Mist says:

            Vitor-

            That’s fair. I no longer suspect you are a computability groupie. I still don’t think computability per se has much relevance here, but the complexity arguments, a la Scott Aaronson, are surely apropos.

  14. sclmlw says:

    I’ve always taken simulation arguments to be similar to those philosophy 101 hypotheticals that can’t be falsified, like “what if we’re all a dream in God’s head and when he wakes up we’ll cease to exist?”

    It’s a useful exercise to teach beginning philosophy students basic ideas (like that you have to be able to build upon fundamentally agreed upon axioms or you can’t get anywhere) but not really interesting beyond that. Well, not to me at least. The simulation hypothesis feels like a modern rewording of classical philosophy in that way. Arguments about it remind me of freshman philosophy debates.

    • rho says:

      yes, metaphysics is unfalsifiable. if it bores you then go do something else?

      This is staunchly metaphysics, and like in ethics and aesthetics, newton’s flaming laser sword is not a guide to truth, it is a conversation-ender.

    • Ketil says:

      I think it is in the Illuminatus trilogy that the characters realize they are characters in a book, and after discussing this for some time, decide that it doesn’t really change anything and to go on with the story. After all, what else can you do?

    • Markus Ramikin says:

      Well, simulation arguments are interesting insofar as they lead to people suggesting possible tests, as well possible consequences of the sim hypothesis being true. As long as that’s happening, we’re not quite dealing with a pure Alder’s Razor situation.

      People may be wrong about this, meaning: there may be no possible useful experiments, nor any way to assess or manipulate consequences of the theory being true, if it is true, and this situation will never in a million years improve. But I wouldn’t want to be the guy who tries to make the claim that something is forever beyond the reach of science.

  15. textor says:

    Like everyone says, we don’t have any data to figure out what the simulators want, can do (aside from simulating us) and what the rules in their universe are like. So it’s impossible to determine what they’re aware of.

    For all we know, they’re dumber than us and are just lazily throwing their mind-boggling computational resources at the problem of creating superintelligent AI via something like genetic algorithms (and for these algorithms to be as unbiased as possible, they’ve engineered our physics which lead to actual chemical evolution); the wave function they compute is really hard to analyze in real time. They hope to detect results automatically when they notice unnatural changes in mass and energy distribution, due to emergence of an adult paperclipper or another strong AI; such an AI will then try to survive the (inevitable; it’s a feature) death of the Universe by signaling its code to the owner, via some large-scale astroengineering technology, and they’ll read it and appropriate to their needs (for example, having it do their bidding in exchange for “physical” universes it can paperclip all over again). And our greatest sin for now is eschewing natural selection/eugenics (and raising our intelligence to the point where analytically engineering such an AI becomes possible) in favour of lazily throwing computational resources at the problem (which we can’t hope to solve this way because of our low resolution and insufficient dimensionality). But they can’t notice that, so they’ll just terminate another worthless sim, a dot in the space of all simulated spacetimes, when the Sun fizzles out.

    …yes, that’s my personal crackpot theory. How can its likelyhood be assessed?

    Anyway, my point is it’s almost impossible to even talk accurately of what is or isn’t plausible here. Maybe the owners read NYT or arXiv. Let’s hope not.

  16. J says:

    It’s a fun thought experiment to imagine who might want to simulate us and why. If we’re just a screensaver, or a science experiment, or a way to generate art, then they don’t care if we know. Maybe our universe is mainly for examining black holes at the centers of galaxies; hard to know if we’re even the main attraction.

    In all those cases they have no reason to hide the fact that we’re simulated, and we might well be able to prove it.

    • rho says:

      “Why bother? Just leave the glitch in, who gives a fuck?” -Omniscient Simulator of the Universe

      This is my favorite response by far.

      • Lambert says:

        What does the sysnthesis of the simulation hypothesis and https://xkcd.com/1741/ look like?

        Is there a simulator whose marriage broke down because they were always late at the office, arguing about what the fine structure constant should be?

        • rho says:

          Is there a simulator who’s marriage hasn’t broken down?

          The long hours, the obsessing over minutia, who could stay married to that?

          “I swear, you care more about those imaginary beings than me.”

          “They’re not imaginary, they’re simulated, there’s a difference.”

          “Well, can they give you a hand-job? You have a real woman, who would worship you here, if you could remember her birthday. Yet you fiddle with those sims, that live in your calculator. I’m staying at my mother’s tonight.”

      • J says:

        It’s easy to accidentally conflate general ai with the ability to simulate our universe. Partly because they might have reasons to build such simulations. But it’s not 1:1. You don’t have to be super smart to simulate particles. You just need a lot of computing power to do very many. I did a planet orbit simulator in high school just from Newtonian gravity. It had many bugs. Scale that up to a bigger universe and our universe might be an analogous project for a being not much smarter than me.

    • analytic_wheelbarrow says:

      Maybe our universe is mainly for examining black holes at the centers of galaxies; hard to know if we’re even the main attraction.

      Yes, this was my thought as well!

      Maybe the simulator-folk were trying to study physical systems with different setting for fundamental constants (or even fundamental laws). These particular settings happened to result in sentient life in one corner of the universe, but the simulator-folk haven’t even noticed or care!

    • That’s why I reject the simulation hypothesis, why would posthumans want to simulate a large number of us? It’s unlikely they’d find simulating us to be practically useful, and they’ll probably see the ethical problem in simulating a lot of preventable suffering. There is much we could learn from unethical experiments we choose to not know, they’d have the same attitude.

      Here’s a variation: suppose some Cro-Magnons are telling stories around the fire. One tells of an “industrial revolution.” He describes the period after the industrial revolution as lasting far longer than the time their own species was extant, and that industrial-era humans will have exponentially more resources than they do. He reasons that they would be interested in the question of how they evolved, and thus would create “zoos” where they would take their still-dumb relatives and try to evolve them to become intelligent.(I’m sure some modern human has suggested doing this as a very-long term science project) Since the number of intelligent hunter-gatherers in the zoos is far greater than the number of original hunter-gatherers who seeded civilization, the only reasonable conclusion is that either the industrial revolution won’t happen or that they are in a zoo.

  17. rho says:

    “If in all the world, I could find love between two humans that is pure and true, a love like this, then I will leave it.”

    “…I have found so many. I cannot delete it and start anew, even if I would do better next time.”

  18. Ketil says:

    Hasn’t this already happened?

    Richard Feynman, in one of his lectures, says straight out that QED doesn’t make any sense at all, but still that’s how things appear to be.

    We fit the map to the landscape here, not the other way around. And physicist have lived in a constant state of scratching their heads forever, trying to make sense of the world, and it – simulated or not – hasn’t ended. So I wouldn’t worry too much about this.

  19. Steve Sailer says:

    This idea that the universe is a simulation occurred to me perhaps 20 years ago.

    It basically seems like a theological concept. The Simulator would be the equivalent of God, even though there might, in turn, be even higher levels than Him.

    • RicardoCruz says:

      Yes. Interestingly enough, the people who believe in God think that God wants us to discover Him. The people who believe in simulators think those beings running the simulations do not want us to discover Them.

  20. Jack V says:

    So a key indication that we live in a simulation is that when we try to develop AI all the ideas that look easy would just mysteriously not work for no particular reason, but then those failures would get good justifications later? 🙂

  21. Aapje says:

    How is this research meaningfully different from theology? What does God demand/expect of us & when will he smite us seems very similar in that it involves speculation about an entity that is claimed to be different from us and who we cannot examine. So both theology and simulation research typically seem to involve a lot of speculation about the nature of God/The Machine, where this nature is often assumed to be similar to ours or that of our machines, which conflicts with the assumption that God/The Machine is different.

    • DarkTigger says:

      The author is a philosophy Prof. You know the joke about the differents between philosophy and theology? The one with the black cat in the dark room that isn’t there?

    • J says:

      In both cases there are places you can look to see if the claims are true. In one we’ve put a lot of effort into looking, but in the other we haven’t.

  22. rahien.din says:

    It’s fantastic how you turn the infinite-regress logic of the simulated world idea back on itself.

  23. Markus Ramikin says:

    I say we carve the Moon into a giant middle finger, with a plaque at the base/wrist, saying “thanks for the tooth decay, cancer, Alzheimer’s, and every other bit of suffering and humiliation”.

    • Don_Flamingo says:

      Tut, tut. Silly humans, thinking that everything is about them.
      I really only created the universe for the zebras.

      • The Nybbler says:

        Wait, is that thing still running? It was supposed to stop itself after recombination; I know re-ionization makes pretty pictures but it’s completely an artifact, none of that happens in the real world. My meta-Amazon bill is going to be horrendous.

  24. C. Y. Hollander says:

    I’ve said it before, but it utterly baffles me how people can treat the Simulation Hypothesis with deadly earnest, yet remain unwavering atheists. Beyond the analogies that undeniably exist between the god[s] of traditional religions and the hypothesized simulator[s], why couldn’t the Simulator literally be the God of the Old Testament, for instance, down to every revelation ever made to prophet or people? And, if that is once admitted as a theoretical possibility, whence the staunch belief that it is nonetheless not true in our reality? How does that disbelief differ from the dogmas of any religionist?

    • Markus Ramikin says:

      Oh boy, where do I even begin.

      For one: “literally god” is supposed to be omniscient and omnipotent. He doesn’t need a sim, he already knows what would happen.

      For two, how on Earth would you know it’s the God of the Old Testament, and not Vishnu? Or Zeus?

      For three, the simulation hypothesis does not do this magic trick that religions do, in pretending to solve the infinite regress problem. The simulators are not presumed to be immune to questions such as “who created them?” “How are they themselves sure they are not in a sim?” etc. Religions arbitrarily say God is uncreated and the source of everything, as if they could possibly know that.

      Similarly, the simulators are not automatically assumed to be worthy of worship, or infallible sources of morality, or sacred in the sense that we should not try to apply science and probability theory to understand them. They are just as much an insect under the looking glass to us as we are to them – at least in principle.

      Traditional religions as formulated don’t make any sense at all, least of all in the context of a sim. Consider: is “God” trying to leave us unmolested and observe the simulation, or is he an interfering God, who has strict commandments that he wants us to follow, and occasionally performs miracles as given in the Old Testament? If he’s not interfering, then he’s not the God of the Old Testament. If he is interfering, why would he not communicate his wishes clearly, by sending angels or writing big letters in the sky, instead of having us argue over unreliably translated, unreliably preserved, overall unreliable thousands-years-old accounts from some dudes who were probably insane or high or, most likely, just making shit up?

      I could go on, but in truth, isn’t this topic done to death before it even began?

      • eigenmoon says:

        I’d like to double down on C. Y. Hollander’s position.

        1. God already knows our future, but grants us existence anyway because he considers it a good thing to do. Nothing changes in the previous sentence if the way he provides our existense is by simulating.

        2. You wouldn’t. Vishnu and Zeus running the sim are also cases that you might want to look at as seriously as the simulation hypothesis itself.

        3. The simulation hypothesis also has the infinite regress problem. Look at Max Tegmark explaining this problem.

        4. If you’re not a Jew, then the OT God isn’t even asking you to worship him, only not to blaspheme him.

        On morality: it’s in principle possible to assert existence of the OT God but deny that he’s a good guy. Gnostics and Marcionites did that.

        > sacred in the sense that we should not try to apply science and probability theory

        God isn’t sacred in that sense. It’s not that you shouldn’t study God with science, it’s just not very doable.

        5. > by sending angels or writing big letters in the sky,

        That’s the OT God we’re talking about, he doesn’t seem interested in converting the entire planet. Maybe this way he gets a bonus score or unlocks an achievement.

        > unreliably translated
        Translated? We’re not talking Christianity here. Worshippers should learn Hebrew.

        > unreliably preserved,
        Maybe God decided that the Masoretic text is good enough and not worth micromanaging yet another bunch of prophets to fix.

        • Doctor Mist says:

          The simulation hypothesis also has the infinite regress problem.

          Only if you imagine that it is trying to say something about the fundamental nature of reality.

          For there to be a simulation, or a simulation of a simulation, or etc. requires there to be at least one universe that came into being by some other means, and the simulation hypothesis has nothing to say about how that happened.

        • eigenmoon says:

          (2) would imply that P(OT God) > 0.5, which would be unfair to Vishnu and Zeus, so no one is saying that.

          What I’m saying is that once you treat P(sim) as non-negligible, most of atheists’ arguments that P(religion) is negligible stop working.

          I remember some survey of Rationalists (I can’t find a link now) where people on average ascribed 1% confidence in the idea that miracles can happen but 10% confidence to being simulated. But if we’re simulated, then the probability that there are some cheat codes that do miracles looks to be way bigger than 10%. That looks like conjunction fallacy. The probability that miracles happen because they’re programmed into the simulation shouldn’t be greater than the probability that miracles happen for any reason.

          Once you ascribe non-negligible probability to miracles, you can’t really get rid of gods as easily as Markus Ramikin tried to do. Why there’re many religions? Well, maybe it’s multiplayer. Why the OT God stopped messing with humans? Well, maybe he got IRL stuff to do, etc.

          I’m not saying that P(religion) > 0.5. I’m just saying that the probability that God exists for any reason should be at least as large as the probability that God exists because he’s the dude that runs the simulation.

        • viVI_IViv says:

          1. God already knows our future, but grants us existence anyway because he considers it a good thing to do. Nothing changes in the previous sentence if the way he provides our existense is by simulating.

          Or the “simulation” isn’t actually being simulated step by step in a feed forward way, but all at once by some kind of equation solver, or maybe God is so smart he can solve all the equations in his head without the need of an external computer.

          Therefore God can know at the whole solution at once from the beginning of time to the end, just like you can know the plot of a book all at once, regardless of the internal chronology experienced by the characters.

      • The original Mr. X says:

        For one: “literally god” is supposed to be omniscient and omnipotent. He doesn’t need a sim, he already knows what would happen.

        Why assume that the only motivation the simulators could have is to try and find out what happens?

        For three, the simulation hypothesis does not do this magic trick that religions do, in pretending to solve the infinite regress problem. The simulators are not presumed to be immune to questions such as “who created them?” “How are they themselves sure they are not in a sim?” etc. Religions arbitrarily say God is uncreated and the source of everything, as if they could possibly know that.

        It’s not arbitrary at all. Try learning a bit about topics before you spout off on them.

        Traditional religions as formulated don’t make any sense at all, least of all in the context of a sim. Consider: is “God” trying to leave us unmolested and observe the simulation, or is he an interfering God, who has strict commandments that he wants us to follow, and occasionally performs miracles as given in the Old Testament? If he’s not interfering, then he’s not the God of the Old Testament. If he is interfering, why would he not communicate his wishes clearly, by sending angels or writing big letters in the sky, instead of having us argue over unreliably translated, unreliably preserved, overall unreliable thousands-years-old accounts from some dudes who were probably insane or high or, most likely, just making shit up?

        He brought a guy back from the dead once, that’s pretty clear. Unfortunately lots of people have clouded their intellects with sin and pride, so that some openly state their refusal to accept any evidence, whereas others would simply offer up some mealy-mouthed concessions about how this counts as evidence, but then simply demand more and more evidence for an indefinite period of time.

        • phi says:

          From the blog post you link to first: “I’m not going to present and defend any version of the cosmological argument here. I’ve done that at length in my books Aquinas and The Last Superstition, and it needs to be done at length rather than in the context of a blog post.”

          For those of us who don’t want to buy the books, do you have a handwavey summary of the actual cosmological argument, or would that amount of compression be too lossy?

          Also, for what it’s worth, though I am an atheist, I think Dawkins is being too pessimistic about the evidence for God in your second link. For most miracles, I would consider a hoax to be a much more likely explanation than God (indeed there have been many people in history who performed magic tricks and tried to pass them off a miracles). However, in most cases, there would probably be additional evidence one could observe, or experiments one could conduct to rule out the possibility of a hoax. After that, I would consider aliens and Gods about equally likely as explanations (maybe favoring one or the other depending on the nature of the miracle). But there would almost certainly be some way of distinguishing between aliens and Gods as well. And if it came out in favor of Gods, I would become a believer, end of story.

          So yeah, Dawkins is definitely being unreasonable here, and I hope he changes his mind. Although, to be perfectly fair, I haven’t watched the whole interview. Maybe he takes the definition of supernatural to be uncomputability. I have no idea how I would distinguish between a computable God and an uncomputable God, or even if it would be possible to distinguish them for a computable mortal like me. Maybe that’s the point Dawkins is making? Seems a strange way of phrasing it, though.

          • Dedicating Ruckus says:

            For those of us who don’t want to buy the books, do you have a handwavey summary of the actual cosmological argument, or would that amount of compression be too lossy?

            The short version: if anything could either exist or not exist (in jargon, if the thing is “contingent”), there must have been a time when it did not exist, and there must have been something that caused it to come into existence from a state of nonexistence. For anything that caused something else to exist, the same consideration applies. Thus, for anything to come into existence, there must exist something that could not not exist. (That is to say, something that is not contingent; something whose nonexistence is a logical contradiction.) “And this all men call God.”

            If you want to actually understand the argument, I have to recommend reading Feser’s actual books; it’s extremely twisty, and then there’s a whole separate set of arguments for what attributes the object that exists necessarily must have. The point of this distinction is that theism solves infinite regress by positing an object that is very weird, that is in some sense more like “the fact of it being possible for things to exist” than an object per se; this is why the “what caused God?” objection is senseless.

          • Nick says:

            @Dedicating Ruckus

            For anything that caused something else to exist, the same consideration applies.

            For any other contingent thing that caused something else to exist, the consideration applies. Otherwise, you’ve just said every cause is contingent, from which it would follow that God as cause is contingent, which you certainly did not mean to say.

            Come on, man, this is exactly why Feser cautions us not to give “the short version.”

          • Dedicating Ruckus says:

            @Nick

            You are of course correct, and I suppose my only answer is to double down on “if you want to actually understand the argument, go read the books”.

          • Nick says:

            Indeed. 😀

        • Bobobob says:

          That Edward Feser piece is great. The first rule of philosophical fight club is that there is no cause of philosophical fight club!

        • grump says:

          “He brought a guy back from the dead once, that’s pretty clear.”

          Evidence for this?

      • J Mann says:

        That’s a reasonable position to take, but I think pro-simulation people should start their argument with:

        “I have discovered evidence indicating that the intelligent design argument is correct, but I believe it leads to different conclusions.”

      • Isn’t the simulation argument a form of deism rather than theism? We have an ultimate creator/s for our universe, and in the context of our universe, they have the power of a deity. Gods don’t need to be uncreated or worthy of worship. There are plenty of gods in the world’s religions who have origin stories and evil qualities.

        It would be weird if deism, which was left behind as a kind of sheepish toe in the bath equivalent of atheism for the 18th Century, was ultimately right all along.

        • Jaskologist says:

          There’s no reason for it to imply deism; that’s just what most people around here would prefer to be true. Zeus used to instantiate himself in the world to romp about, transforming into different forms, and mostly sleep with the npcs. That seems pretty in line with how a human would use a holodeck.

          (I myself simulated many cities for the sole purpose of destroying them with disasters in my younger years.)

      • Deiseach says:

        If he is interfering, why would he not communicate his wishes clearly, by sending angels or writing big letters in the sky

        Haven’t you heard the arguments of those who claim they still wouldn’t believe if that happened? That they would consider it a hallucination or a hoax or “aliens did it” before they would believe a sign in the sky saying “I am God and I exist”?

        Like somebody once said, “If they won’t believe Moses and the Prophets, they won’t believe even if someone comes back from the dead”.

        • Bugmaster says:

          Well, firstly, some people would surely be convinced by giant letters in the sky; in fact, many would. It seems wasteful to shaft them just because of a few holdouts. Secondly, God, being omniscient, surely knows exactly what it would take to convince each individual person — be it sky letters, personal visitation, a quiet whisper in the night, or whatever. Yet I’m still an atheist, so, if God does exist, he either doesn’t want me to believe in him, or maybe he just doesn’t care.

      • viVI_IViv says:

        For one: “literally god” is supposed to be omniscient and omnipotent. He doesn’t need a sim, he already knows what would happen.

        God is not omniscient in the OT, at least not in the Torah (Pentateuch) ref: Does God know and see everything? He also seems to struggle with iron chariots.

        The dogmas of God’s omniscience and omnipotence and the related theology (e.g. theodicy) start to appear in the Persian era of Israel and were fully developed only in the Middle Ages by Maimonides and the Christian Scholastic philosophers.

        Religions arbitrarily say God is uncreated and the source of everything, as if they could possibly know that.

        Where does it say that in the OT?

        Similarly, the simulators are not automatically assumed to be worthy of worship, or infallible sources of morality, or sacred in the sense that we should not try to apply science and probability theory to understand them. They are just as much an insect under the looking glass to us as we are to them – at least in principle.

        And yet the proponents of the simulationism claim that we should do this and should not do that in order to avoid angering the simulators. This was the point of the article linked in the OP.

        If he is interfering, why would he not communicate his wishes clearly, by sending angels or writing big letters in the sky, instead of having us argue over unreliably translated, unreliably preserved, overall unreliable thousands-years-old accounts from some dudes who were probably insane or high or, most likely, just making shit up?

        Why are video games challenging?

    • You can accept that the OT was right about the existence of God but unreliable in providing information about him, just as it’s unreliable in providing information about science and history.

  25. Lambert says:

    And God said: “I’m running a physics sim on this machine. It’s a long shot, but I’m seeing if we can solve the stability problems by sticking some zeroes on the end of half the particles’ rest masses. Please don’t let anyone else log me off while I grab a coffee.”

  26. JoshP says:

    The reverse of Pascal’s wager (and the first commandment): “Thou shalt not believe in a Creator”. At last!

  27. RicardoCruz says:

    if we learn we’re in a simulation, that might ruin the simulation and cause the simulators to destroy the universe.

    This argument could be constructed the other way around. If we do not learn that we are in a simulation, they will destroy us because they will lose interest in dumb creatures.

    This reminds me of the Pascal Wager. People argue that we should believe in God regardless because there is a chance he exists and he will reward those who believe Him. I could just as well argue that we should not believe in God because there is a chance he exists and he will punish those who believe Him.

    Without knowing the intentions of these super-beings, whoever they may be, it’s impossible to know whether they will reward or punish us for discovering them or failing to do so.

  28. The original Mr. X says:

    TFW secularists start supporting intelligent design theory.

  29. JPNunez says:

    If a civilization can simulate other universes, and in time those universes can simulate other universes, then chances are that we live in one of the more numerous simulations…

    …says member of a civilization which can’t simulate jack shit.

    • Don_Flamingo says:

      We can simulate a shitton of things. E.g. we can set up a Minecraft server with it’s simulated world. Perhaps in the parent universe, they are also extremely bad at simulating universes, thus this is what we get.
      [our universe is as realistic a sim of theirs, as the Mincecraft-server is of our universe; what justifies your prior that a potential parent’s universe wouldn’t be much more complicated?]

      • JPNunez says:

        I don’t think any of those are complex enough to allow processes like life to arise naturally. Let alone _intelligent_ life. In any reasonable or even unreasonable timeframe.

        Even in something like Minecraft has people build huge circuits by hand. If you leave a Minecraft server running for the rest of the life of the universe, it will never have a civilization arise there. More probably … the server will just crash.

        I feel there are four types of simulators out there right now:

        -Videogames. Where the level of simulation is so high level that life-like processes are cumbersome

        -Mathematical models. Stuff like big climate simulations, simulations of the early, or current universe. In these, the details are so coarse grained that life-like processes are also cumbersome.

        -Highly detailed simulations of existing things: there is a simulation of a whole worm’s neural system out there that you can run on a sufficiently powerful personal computer, things like that. But this comes premade. Life just cannot arise in these.

        -Actual simulations of low level stuff? I dunno if there are any these out there? Even stuff like fold-at-home takes an insane time to fold a single protein, and may even do it wrong.

        Life will only arise in the last kind, and I cannot even name one of those we are doing currently.

        I don’t know if down turtle worlds would be much more complicated. It’s possible? But the problem is that we cannot really do good simulations right now, so the argument that simulations are more common than real universes dies in the crib anyway.

        • Don_Flamingo says:

          Yes, the Minecraft-blocks living in the uncharted regions of our sims aren’t intelligent by our standards.
          But what makes you think, that what we are something that registers as intelligent life to our creators?
          That would be making unfounded assumptions about their intelligence, which we can’t, because we are only humans and they are quite possibly Eldritch.

          And you can’t assume that we know the (computational) constraints of our simulator’s universe either.
          Assuming that our universe is a sim of theirs, instead of something extremely primitive is not something that I’ve heard an argument for yet. So I’m rejecting it out of hand, same way I’d reject the Old-testament God out of hand.
          Possible, but infenitesimally unlikely.

  30. Bobobob says:

    I think any chain of reasoning that leads to the conclusion that we’re living in a computer simulation has at least one faulty premise to begin with. And I think the same principle applies to the Level 1 multiverse, which assumes that the universe is infinite in extent and there are infinite copies of me writing this post. (I am undecided about quantum parallel universes, though.)

  31. Don_Flamingo says:

    This all fails to take into account, that if we actually are simulated, we have no good idea of what the simulator’s world is like. It might be 69-dimensional and/or completely beyond our comprehension.
    Or our whole universe might be the equivalent of some four year old’s fingercolor painting project. Just that in their world, their fingercolor projects happen to be universe-sized. Perhaps we are some game studio’s procedurally generated world and the playerbase has dwindled already. And it never interacted with our corner, same way the far edges (and most of the volume) of a Minecraft sever’s world is never interacted with.

    The idea that if we are simulated, the simulators would bother with us or read the New York Times (who the hell are we even, in this universe, which is billions of times larger than we can even observe? And who reads the Times, anyway?) is just people wanting to anthromorphize the simulators. Wanting there to be God(s).
    Which we of course promptly model as having human-like intentions, without any good reason to do so.
    Perhaps there are Gods/simulators, but we shouldn’t feel threatened by them, same way a tardigrade at the bottom of the ocean should not feel threatened by an Aircraft carrier. In the tardigrade’s universe, these undeniably exist, are magnitudes more powerful than it…… and of absolutely no danger whatsoever.

    [yes, conceivably the aircraft carrier is a platform to launch a research submarine to get some tardigrade-samples for experimentation, but most tardigrades would not be caught even then. And most likely the aircraft carrier is doing other things. We should have high priors that our simulators presumably are concerned with other things than humans, exactly like aircraft carriers usually have about a million other concerns. Any other assumption is silly anthropocentrism and should be treated as irrational and unfounded religious belief.]

    • Dedicating Ruckus says:

      We should have high priors that our simulators presumably are concerned with other things than humans, exactly like aircraft carriers usually have about a million other concerns. Any other assumption is silly anthropocentrism and should be treated as irrational and unfounded religious belief.

      I notice that you don’t actually have an argument against the idea that humans are in some way universally special, you just call it names.

      • Don_Flamingo says:

        Why would that need an argument?
        It didn’t occur to me that it needed one, I thought that was obvious.
        The idea that humans are somehow “special” seems as unlikely as the existence of a candy bar on Mars. We have no evidence for it at all and it’s not a priori plausible.
        (okay, that’s just saying that my priors are right goddamn it, when I’m saying “not a priori plausible”. I can’t find any merit in the idea, though. Why take that unfounded assumption seriously or what makes you think it’s not unfounded rather?)

        • Don_Flamingo says:

          What makes our own probable unimportance obvious to me is that the size of the actual universe is between 250 and 10^23 bigger than the observable universe.
          Concluding that we are likely alone and special would be like waking up in a city of unknown (but massive) size in an empty apartment and concluding that the city is deserted. I think one should look around for a bit, before the notion would even occur.

          • Dedicating Ruckus says:

            The fact that the universe is big doesn’t mean that humans aren’t special. Maybe creating a universe ten billion light-years across was the easiest way to get Earth to look the way they wanted. Maybe it’s not actually big, and everything outside Sol system is run at low-resolution for us to have something to look at.

            This is all blatant speculation with no evidence behind it, of course. But there’s no evidence that bears either way. Absent that, the question of humanity being special or not is just a battle of dueling priors, and there’s no basis to just assert firmly that your priors are right as if there were actually, like, evidence for them.

          • Don_Flamingo says:

            Well if the universe was just our solar system, we would have more evidence that we are indeed “special” . If we were special, we would not expect there to be so many things that have no bearing on us.

            Also, only because a question might have two answers, doesn’t imply that the priors should be 50/50. Humans being “special” appears to me to be about as plausible as candy on Mars. If there was candy on Mars, a lot more very unlikely questions would have to be answered with “yes” .
            Both “yes”-answers should be slapped with a large complexity-penalty, when trying to get some justifiable priors.

            edit: making up priors without (much) evidence is extremely important. It’s not blatant speculation. Priors are important, so is where they come from.

            Making absurd, glorious stories about humanity is a fanciful flight of fantasy however akin to a delusion of grandeur in a manic individual. The mistakes/flawed reasoning is pretty much the same. The former is just more common.

  32. FormerRanger says:

    Interesting that we’ve gotten scores of posts in and no one has (up to now) mentioned Neal Stephenson’s “Fall, or Dodge in Hell.”

  33. Immortal Lurker says:

    I think the point about correcting the simulation upon reading the Times isn’t quite accurate. After all, who really trusts the media these days?

    Also, I’m willing to believe in simulators who both are capable of making mistakes, unwilling to alter the simulation once it starts running, and willing to pull the plug the instant the inhabitants know the score. Maybe these are undergrad simulators who had to preregister their universe with their professor. Interfering in any way gets a failing grade, and massive divergences from what would happen in a real universe get a failing grade. Everyone having definitive proof that they are not in a real universe is not something that a real universe would do.

    There would still be thousands to millions of simulations produced the simulating civilization, but the particular simulators in charge of us don’t have access to all the optimizations learned from that.

    I think we can rule out the incompetent undergrads for a different reason. Physics is really impressively good and consistent. We don’t have any reason to believe they would screw up the cosmic background radiation.

  34. micje says:

    > if they can simulate billions of sentient beings, their labor costs are necessarily near zero. Such a monitor would have complete instantaneous knowledge of everything happening in our universe, and since anyone who can simulate a whole planet must have really good data processing capabilities

    That does not follow. If I’m not real, and am being simulated accurately, that does not mean anything else is being simulated accurately too. Nobody else needs be conscious for me to be convinced of their consciousness. Other people don’t even need to exist unless I have met them, and they may cease to exist after my meeting them. The universe could just be pretty wallpaper, as long as the simulation isn’t simulating an astronomer. Such a simulation would be many, many orders of magnitude simpler, (how much depends on whether the simulator is simulating Einstein or a sewer rat) although still enormously complex.

    Cogito ergo simulor.

  35. Randy M says:

    Is it fair to call the position of restricting certain research because it might offend the simulators anti-science?

  36. ariel says:

    This doesn’t agree with my experience running simulations. Normally when I run a simulation (of genetic selection, sports, probability questions, competing strategies for games etc.) I set in place a few basic rules and then allow 1000 (or wtv) random simulations to run to completion before I check out the aggregate result, alternating with stepping through individual runs for debugging and testing purposes. Analogously, our simulators minimally need to know the laws of physics and have godly amounts of compute. I wouldn’t expect them to personally know about things like individual sparrows. Although, I would expect them to care about something fundamental like how we investigate whether we’re in a simulation like Scott argues.

    • John Schilling says:

      This. Anything worth simulating, is worth simulating bignum times and looking only at the ensemble of results, On the one hand, that means we can be pretty certain that this instance will be left to run to completion, so long as we don’t accidentally hit a Singularity. On the other hand, if we want the Great Simulator to use this instance as one of the initial conditions for his next ensemble of simulations, we want to be sure He doesn’t toss it out as a hopelessly-contaminated outlier, not worth trying to salvage because He has bignum-minus-one perfectly good datasets already.

      So, was this instance of the simulation set to run from Big Bang to Big Crunch, or Last Tuesday to Next Tuesday?

      Also, the Universe almost certainly isn’t a simulation, but it’s fun to think about sometimes so here we are.

  37. fr8train_ssc says:

    This post inspired me to write a sci-fi short story/flash-fic, but I’m unsure whether the appropriate place to link it (if at all) is here or in the open thread or a different thread.

  38. Virriman says:

    Some other sources raise concern that we might get our simulation terminated by becoming too computationally intensive (maybe by running simulations of our own).

    No need to wait till we’re running our own simulations! We can now add “might destroy the universe” to list of reasons why cryptocurrency is bad.

    Could our simulators manage to optimize away the proof of work required to mine bitcoin? There is no need to simulate all the work the mining computers do if they can just provide a valid solution to a random miner every ten minutes. That would only work if they have broken the hashing algorithm though. If they haven’t, optimization would require provision of fudged “solutions” and they’d have to ensure success every time any computer attempts to verify the validity of their fudged solution. Probably easy enough if the validation is hidden away in silicon, but “any computer” is a rather broad category that includes everything from computers built in Minecraft to a person working with paper and pencil..

  39. elusivepastry says:

    Just two comments 🙂

    1. “Anyone who can simulate an entire planet must have really good data processing capabilities” seems like it could be a pretty big stretch. If our current data production:data processing ratio is any indication, the ability to simulate and/or generate data could feasibly continue to outstrip our ability to process that data into forever. That said…

    2. I tend to agree and don’t really feel any threat that we will irk the operators of our ersatz universe (if they exist). More directly important to my life on Earth, I feel that most “what is the nature of reality” answers can be swapped in without really changing how I want to behave or make decisions. So spending energy on contemplating is sort of a waste.

    2. It may be a big assumption that

  40. Carl Pham says:

    Nonsense, the simulators aren’t simulating billions of people, they’re only simulating me and my experiences of all of you.

  41. J says:

    Obligatory Rick and Morty:
    https://youtu.be/szzVlQ653as

    That show does an amazing job at showing you how our great great grandfathers would feel if they were dropped into our world.

  42. Eponymous says:

    Since (as I’ve previously argued) I consider it extremely likely we live in the domain of some sort of god (or God — or perhaps a hierarchy of gods), we might want to ask whether any of them want something from us.

    Of course we don’t necessarily have to do it. But it might be a good idea.

    Or, if this is a game, we might want to know that so we can start preparing for the other players to show up.

    And since we are probably close to an intelligence explosion, it’s a reasonable guess that if the simulator is interested in something, it might be connected to that.

  43. SnapDragon says:

    I’ve never found the statistical argument convincing. Simulations that look like they’re based on low-Kolmogorov-complexity physics are boring! If you postulate that a transcendent civilization would create 10 billion such simulations, then why wouldn’t they create 10 trillion simulations with rules that are less “simple”, but much more fun to inhabit and watch? And those would look a lot more like fiction, mythology, or videogames than our slow, boring, rules-based existence.

    Since the Universe would be created for us, there’s no reason there wouldn’t be physical laws at our level of complexity. This would look like magic or the supernatural to us, and would be hard for science to explain – but it would also be readily observable, by anyone.

    But no. The laws of our Universe look just about as simple as they can possibly be, conditioned on intelligence being possible. This suggests they arise from cold, indifferent mathematics. The evidence is strongly against there being an intelligent designer, whether you call it a God or you call it a Simulator.

    • Dedicating Ruckus says:

      The laws of our Universe look just about as simple as they can possibly be, conditioned on intelligence being possible.

      There are certainly simple laws involved somewhere, but the idea that all the phenomena in the universe actually reduce to those simple laws is more or less taken on faith rather than demonstrated.

      For instance, imagine: how would that simulation that “looks like fiction, mythology or videogames” actually work? If it’s based on our fiction, mythology or video games, it would be like the boring rule-based universe, plus extra hacks to make things interesting (like gods or magic). Our observations are consistent with living in a universe that runs on simple laws plus hacks to make things interesting; consider consciousness as a potential one of those.

      • SnapDragon says:

        If you’re living in a videogame, and you try to probe the limits of your universe with science, you don’t find that “huh, this tree grows because it’s an emergent property of chemistry and atoms and quantum mechanics.” You get “this tree grows because it’s a Tree and that’s what Trees are programmed to do.” Which is, frankly, how humans thought the world worked for most of our existence. Every time science took something complex and reduced it to equations, it was a surprise. Huh, the flying celestial objects obey the same physics we do. Huh, I move my arm with electrical impulses rather than an ineffable extension of will.

        You’re right that I can’t rule out that we’re living in a universe with simple laws plus a few hacks. But *why have the simple laws*? An advanced simulation doesn’t need them, any more than World of Warcraft needs to model all the atoms in a dragon. So observing simple laws underlying most phenomena is strong evidence against intelligent design. Not proof, of course. Just strong evidence.

        (BTW, note that I’m not arguing against a “blind watchmaker” who’s just running some massive cellular automaton and has no idea what human life looks like. That hypothesis is unfalsifiable. I’m just arguing against a simulation by beings who know about us, who created us explicitly and are watching.)

        • Dedicating Ruckus says:

          If you’re in a video game, you see the world as simple laws (like “you can’t move below -50 on the Y axis” and “unsupported objects accelerate downward at a constant rate”) plus a bunch of hacks to enable interesting behaviors (like everything else). If some phenomenon is adequately provided by your simple laws, you have no reason to add a hack to provide it specifically. (Except maybe as a performance optimization, as real video games have to do all the time; but we’re already assuming the simulators have ~infinite computing power.)

          Assuming the simulator is looking at humans and cares about them, obviously it wants to see “humans that perceive a world governed by simple laws”, not “humans that perceive a world governed by one-off special cases”. (Either that or it also cares about the outputs of the simple laws in and of themselves.) None of that is actually an argument against the idea that the simulation still has (subtle) hacks in it to do things that the laws don’t.

          (and yes, this hypothesis is basically unfalsifiable, but so’s everything else about simulation, *shrug*)

      • beleester says:

        Consciousness (or at least intelligence, since you can’t prove that other brains are conscious, but we’ll leave that problem to the solipsists) doesn’t look like a one-off hack, though. For instance, if someone suffers brain damage, their intelligence starts functioning differently in a predictable way. If you put a wire in someone’s brain and run electrical signals through it, that has an effect too, one that seems to arise naturally from the normal rules of how chemistry and electricity work. The mind looks an awful lot like an emergent feature of the simple laws rather than a hack along the lines of “This lump of atoms is a Mind, ignore the usual laws for atoms and simulate it with the special rules for Minds.”

        (In the way that a world with magic would need to include a law that says “And when a mage says these words, we ignore the usual laws of thermodynamics and refer to the rules for the Fireball spell.”)

  44. tezthenerd says:

    My favourite result towards ruling out whether we live in a simulation or not is this one:

    https://philpapers.org/archive/ALEATO-6.pdf

    A TYPE OF SIMULATION WHICH SOME EXPERIMENTAL EVIDENCE SUGGESTS WE DON’T LIVE IN
    SAMUEL ALEXANDER
    Abstract. Do we live in a computer simulation? I will present an argument that the results of a certain experiment constitute empirical evidence that we do not live in, at least, one type of simulation. The type of simulation ruled out is very specific. Perhaps that is the price one must pay to make any kind of Popperian progress.

    • Don P. says:

      I’m afraid your typography had me thinking that the proof was that we are not living inside Samuel Alexander, which admittedly I had not considered a possibility.

  45. Maxwell says:

    If the world ends because some alien gets bored and turns off their PC…talk about a painless death! We should be so lucky!

  46. pelagius says:

    As we’re going to keep doing the hard sciences anyway, we’re also doing the research to find flaws in the simulation. There used to be a rule of thumb that the likelihood of bugs increased as the square of the length of a program. While changes to the way we build software made that a less reliable guide, it’s hard to imagine that the simulaters built something this complex without some messups. Unless they have attained perfect knowledge, and then why experiment?

    • acymetric says:

      Sure, but why would you think we would recognize the bugs? Maybe there are bugs all over the place, but for us its just “normal”.

      • pelagius says:

        We might not, of course. But if the simulation hypothesis is kept in mind, then we have an alternative to positing things like dark matter. Think of us as awareness within the program using the bugs to deduce the existence of the programmer.

        • pelagius says:

          What really keeps me awake nights is the certain knowledge that the simulated solar system is being readied–trained up, evolved–by some grad students at Milky Way U, in preparation for a combat-robots scenario against the team from Greater Magellenic Polytech’s simulation. Because: grad students.

  47. Dick Illyes says:

    Robert Monroe has written several interesting books and founded an institute to study this phenomena. He got into it after having a series of involuntary out of body experiences. He says we are unique individual self aware spiritual beings having a human experience. We choose to come into this game and there is a goal for the entire system. His student Tom Campbell, a physicist, describes the experience using computer analogies. We only experience this world because of belief in what he calls physical matter reality. It only exists in our shared belief. There is no matter outside the game, it is all a belief, the ultimate virtual reality. Free will in this game does exist and seeing what will happen is the objective of the whole exercise. Our sense of time is not the same outside the game.

  48. Dino says:

    As a computer programmer, my take is – writing software that simulates a universe would be easy, take me maybe 2-3 months. Would use a huge number of variables representing quarks, algorithms encoding the laws of physics which determine how the quarks interact, a universe represented by a 3D cubical array of points, and a clock tick that steps thru the evolution. Start with a big bang, and let ‘er rip. Given a huge enough number of quarks and long enough time, anything that can happen will eventually happen, including the evolution of intelligent consciousness, which we know can happen because it has happened. (insert snark here.) This program design is the obvious way to do it, and I would expect, but can’t prove, that any universe simulating program would be coded this way. Designing a user interface to allow monitoring things would be a much harder problem, but is not needed for my next point. That 3D grid of points means there is a smallest possible unit of distance, and the clock tick means there is a smallest possible time interval. And phenomenon would get weird as you approach those limits because of aliasing effects (the jaggies). You can guess where I’m going with this – I’m saying that quantum mechanics is evidence that we are living in a simulation. Also suggests another line of experiment – the 3 dimensional cubical-ness might also show up, if a way could be found to look for it. Aside – the software could use polar co-ordinates instead of cartesian, but I doubt any coder would do that. And we could look for that also.

    Actually running this program is problematic – can’t build memory large enough. You need a huge number of quarks to build the memory to hold the state info for each simulated quark. I think this means that actually running such a simulating program is not possible.

    • John Schilling says:

      I think this means that actually running such a simulating program is not possible.

      Actually running a program that pushes video-game-style hackery of “real physics” to a level that will fool ordinary humans in day-to-day life, but includes your full-fidelity quantum simulator as an optional module and a quasi-AI that determines whether e.g. a “scientist” NPC is running a particle accelerator and so booting the full-fidelity quantum model in the test cell of the accelerator, might be possible. And if so, would probably be useful to someone. But you’re not coding up that program in 2-3 months, and it’s still going to be an incredible CPU and memory hog, so not clear whether it is useful enough to be worth the bother.

      The bit where the sim-gods have to boot up the “whole rest of the universe” module on 25 Dec 1968 in place of the default starfield projection, has been addressed elsewhere.

    • DocKaon says:

      You do realize that none of that would work, right? Cubical array of points, oops there goes angular momentum conservation. Cubical array of points and time steps, oops there goes relativity. Quantum mechanics is not at all like the artifacts from digitization. Simulation is hard, which is why all this is silly twaddle by people who clearly never actually tried to do it. There is no remotely detailed simulation in the world that doesn’t take vastly more resources to run than the system it’s modeling.

      • Dino says:

        It’s not obvious to me why the cubical array of points and time steps precludes simulating angular momentum conservation and relativity. But I don’t know physics as well as programming so you may be correct. From what I do know about programming, I don’t see any way to code the simulation without using the cubical array of points and time steps, so if we’re both correct, it’s impossible to have a working simulation. But we already agree it’s impossible because of the resource issue.

        silly twaddle

        It’s fascinating that so many people have devoted so much time and energy discussing something that is impossible. My suggestion for a reason is to point to the popularity of computer games like Sim City and Sid Meier’s Civilization.

  49. Hey Scott and all. I’m a fan of the blog, so I was excited to see the post about my piece.

    Reading Scott’s post, I get the feeling that we agree on all the points raised. Notably:

    1) Even if we live in a simulation, it’s improbable experiments will detect anything.
    2) Simulation-related “philosophy” does not pose the same shutdown risk.
    3) Becoming too computationally intensive is a way to get our simulation shut down and a serious concern.

    Regarding (1), I think there’s an important asymmetry with experiments: they can prove we do live in a simulation by detecting anomalies but can’t prove we don’t by not detecting anomalies. So, when we’re talking about the expected value of the experiments, the upside is all in the “detect anomalies” outcomes. But if those outcomes also imply an existential risk, it’s worth thinking about whether the experiments have positive expected value. What is the upshot of using experiments to prove we live in a simulation? Is it just knowledge or something else? If there were a large upside, other than knowledge, that I’m not thinking of, then I would change my mind.

    Regarding (2), philosophical theorizing (like Bostrom’s) about the simulation hypothesis doesn’t have the same risk as experimental observation of glitches, because theorizing would occur in equal measure at the basement level — it doesn’t require observation of things that wouldn’t happen on the basement level. The risk I’m worried about, if there is one, is about people observing things that couldn’t happen on the basement level (which is essentially what simlulation experiments are trying to cause to happen) and not people believing in the simulation hypothesis. I realize that this point is not made clearly in the op-ed.

    Finally, (3) is actually what my paper “The Termination Risks of Simulation Science” is mostly about. I think this is something our civilization will have to think more about in the future as our computational abilities advance. If we ever get to the point of making our own ancestor simulations, then we’ll really have to think about it!

    Unfortunately, given the 800-word limit, I wasn’t able to go into too much detail about these points in the op-ed, so I instead hoped to get the main point across to get people talking about the expected value of simulation experiments, and to indicate that there’s a paper out there arguing that it is negative. I figure people who really care could check the paper out for the full arguments. One of the reasons philosophers don’t publish in the popular press as much as other disciplines is that they worry about getting excoriated by their peers because they can’t be as intricate in their arguments given the word limits. I think, however, that this isn’t good for philosophy or for the world.

    In the paper I discuss a lot of issues raised in the SSC comments. I consider possibilities like there being some reward from our simulators for discovering we live in one, etc. While I think that is possible, about the best we can do in this field is extrapolate from current uses of simulation technology, and that seems to indicate termination as more likely. I also argue that due to the resource-constraints problem, simulations upon simulations is less likely than it initially appears. And there’s a whole section on Pascal’s Wager.

    • I don’t have much to comment on in regards to your points, but I wanted to say thanks for taking the time to respond!

      To avoid having this comment just be noise for people looking for discussion material: I’m not totally convinced that a simulation that’s accurate on a physics level can get more computationally expensive as the physics level is rearranged, though it’s not inconceivable. I’m thinking, for example, about the way GIF and PNG images are compressed, where if all the pixels become different (=> more information content) the compression fails to actually compress anything – but on the other hand, rendering those images is a pretty linear process, computationally speaking.

      It may be entirely impossible to figure out whether computing things ourselves would make the parent simulation more computation-heavy, or if it simply continues clicking on (e.g. because it’s one huge cellular automaton). Of course, that doesn’t invalidate the argument that one should watch out for the case where the answer is “yes, it does”, so I guess I don’t have much of a point, other than to say I think the probability is small (although I’m also not too concerned about abrupt non-existence (as might e.g. also be the case in a false vacuum disaster), though I would prefer if that didn’t happen).

      But in any case, I want to say again that it’s really nice to read your thoughts on this! Thanks for posting. 🙂

  50. ovid75 says:

    But won’t most simulators just set the initial conditions and then allow emergence to do the rest? Or at least wouldn’t the vast majority of simulations be of that type – in which merely initial conditions are set for a universe and then staggeringly diverse (and to the simulator) incomprehensible life forms emerge after billions of ‘simulation years’. I think that argument is plausible in that we can conceive of the science that might make an ‘initial conditions then emergence’ type simulation possible but not the one where beings like us re-create out past history exactly. Hence the former type is more likely over all possible universes as it requires a less advanced civilization.
    If that hypothesis is correct (that simulations where the creators are beings something like us are a tiny minority of simulations) then simulated conscious beings won’t be closely monitored because the simulators will be shockingly different life forms than us who won’t comprehend anything we do (language, science, arts, etc) except in very meta terms (e.g as a civilization spreads across the simulated universe and leaves physical traces).

  51. Peter Gerdes says:

    In fact simulation provides a rather less pessimistic answer to the great filter. It’s a lot easier to only simulate earth in detail and use some crappy approximation (well very good approximation that doesn’t go to the molecular detail) for all the other stars in the universe.

  52. avturchin says:

    I (with coauthors) explored different simulation termination scenarios here: “Simulation Typology and Termination Risks”:

    Abstract: The goal of the article is to explore what is the most probable type of simulation in which humanity lives (if any) and how this affects simulation termination risks. We firstly explore the question of what kind of simulation in which humanity is most likely located based on pure theoretical reasoning. We suggest a new patch to the classical simulation argument, showing that we are likely simulated not by our own descendants, but by alien civilizations. Based on this, we provide classification of different possible simulations and we find that simpler, less expensive and one-person-centered simulations, resurrectional simulations, or simulations of the first artificial general intelligence’s (AGI’s) origin (singularity simulations) should dominate. Also, simulations which simulate the 21st century and global catastrophic risks are probable. We then explore whether the simulation could collapse or be terminated. Most simulations must be terminated after they model the singularity or after they model a global catastrophe before the singularity. Undeniably observed glitches, but not philosophical speculations could result in simulation termination. The simulation could collapse if it is overwhelmed by glitches. The Doomsday Argument in simulations implies termination soon. We conclude that all types of the most probable simulations except resurrectional simulations are prone to termination risks in a relatively short time frame of hundreds of years or less from now.

    https://arxiv.org/pdf/1905.05792.pdf

    • eigenmoon says:

      In the section 1.1.8 you make an extraordinary claim that the bigger a simulating computer is (physically), the more probability there is that we’re in it.

      I should probably mention that I’m doing an adversarial collaboration about the simulation argument, and I’m arguing against the “bland” indifference principle. But even I find 1.1.8 too bold.

      The problem is that when you theoretically split a computer into a top half and a bottom half, the halves aren’t independent: they exchange a lot of electrons. Thus a half can’t be meaningfully evaluating whatever it’s supposed to, because it has some other circuit attached to it everywhere.

      Alternatively, you could fix a time moment and say that all electrons that at this moment were in the top half shall permanently belong to the top half no matter where they actually are. The problem is that this information is quickly sunk into thermal noise, so the halves are no longer meaningfully defined.

      Physically – not theoretically – cutting a computer in half is equivalent to attaching a bit of information to every electron and preventing this bit from being lost in the thermal noise. This is what allows you to claim there are two evaluations happening instead of one. If you just to this theoretically, I don’t see how it could work.

      • anonymousskimmer says:

        I’d be interested in reading your adversarial collaboration if you touch upon 1) “natural” simulations (e.g. simulations created, tornado-in-a-dump-like, by the multiverse itself), and 2) the concept of genuine universes (generated from the fundamental substrate of the multiverse, whatever that is) that are initially created (and possibly manipulated) by sentient beings.

        A sufficiently advanced technology would prefer ‘simulations’ of the second sort to computer-based simulations, as they would presumably have more computational power.

        • eigenmoon says:

          Sorry to disappoint you, but I don’t think we’ll go there. ACs work best with narrow, well-defined topics.

      • avturchin says:

        If a computer is really large (in a sense of size and energy consumption), it could be split in two, which will be causally independent and will be not affected by thermal or quantum effects. Maybe better imagine mechanical 2D computer which has a thickness in 3D dimension. It could be sliced until gears will be too thin to support computations and start wobble. A there is no meaningful causal connection between layers, such a thick computer should be regarded as a linear sum of many thin computers, and thus it should have the effect of the probability of finding myself in such computer if I assume principle “one consciousness – one computer”. In other words, the bland indifference principle implies that I will be in a more energy-consuming computer, all else equal.

        If the bland indifference principle is not working, it should be for a reason: either qualia can’t be created by mechanical computers or any other artificial systems, or the way we count copies should be different. An example for later would be “non-locality principle”: “I am located everywhere, there my copies are located”. In that case, there is no simulation argument, as I am located simultaneously in all possible simulations, in all real-worlds and in Boltzmann brains. I wrote more about the topic here: “Types of Boltzmann Brains” https://philpapers.org/rec/TURTOB-2

        • eigenmoon says:

          I like your example and I dig what you’re saying right up to the moment where you jump to energy. How does such a computer consume energy at all? It seems that the answer is basically friction, but are you sure that using better bearings decreases our probability to be in the computer? Conversely, if we want to increase the measure of our simulation’s existence, would you recommend installing worse bearings for more friction? Maybe we could move the whole thing in space and use superfluid helium bearings or something fancy like that, and then the contraption will use no energy whatsoever. (I guess axes are inherently 3D objects so we need to get rid of them anyway somehow.)

          There might also be some thermodynamic energy penalty for losing information when we reset all the gears, but I don’t think it depends on thickness.

    • Theodore Ehrenborg says:

      I agree that

      simpler, less expensive and one-person-centered simulations

      would be a lot more common than more expensive whole-universe simulations, assuming that simulations represent a significant number of observer moments. Thus it’s much more likely that I’m living in a simulation of just this room (because simulator psychologists are interested in how humans behave) than in a simulation of the entire universe. I won’t find out that I’m in a simulation when physicists discover inconsistencies in cosmic background radiation. I’ll find out when I open the door leading outside and see a blue screen of death.

      Perhaps I don’t live in simulations that are as poorly rendered as that, but I should be more worried about seeing prosaic inconsistencies showing that I live in a small simulation than subtle inconsistencies characteristic of a large simulation. If I accept Professor Greene’s assertion that I quite possibly would be terminated if I find out that I live in a simulation, I now have a perverse incentive to not seek out new information. After all, the best way to stay alive is to not notice any inconsistencies, so I shouldn’t learn anything new.

      Or what about the fact that I’m much more likely to be living in a cheaper week-long simulation than a expensive eons-long one? I should spend the next week spending my savings in Vegas, because this week is the only one I have.

      Professor Greene’s argument seems more credible when its only effect is to change research priorities in physics, but it also implies that we should drastically change how we are living our lives. There has to be some issue with the argument, because it implies too much. Also, agonizing over whether we’re simulated is a wasted motion for the versions of us who don’t live in a simulation. Since the non-simulated people are the ones who really matter—because if they mess up, there is no reset button—we should assume we’re them and act accordingly. A benefit of this interpretation is that we don’t have to constantly worry about whether opening an encyclopedia will cause us to realize an inconsistency and be terminated.

      Of course, if no civilization can feasibly build simulations, we definitely don’t need to worry. I’m arguing that we still shouldn’t worry even if simulations seem feasible, unless we get strong evidence that something is up (like the stars flashing).

      I haven’t read Professor Greene’s paper yet, only his opinion piece. I’ll do that now.
      (I’m going to assume that you have also read the paper.)

      Regarding the thought experiment with Margaret, I suggest using functional decision theory (FDT).
      Margaret would reason, “Whether or not I’m in a simulation, I’m going to make the same decision. If I decide to start up the simulation, then I am either going to (probably) be terminated or (probably not) learn interesting facts about simulations. That’s not a great outcome. But if I don’t start up the simulation, none of me will, so the simulation containing me won’t exist in the first place. Assuming I live in a simulation right now, I prefer existing (even including being terminated soon) over never existing at all. Assuming I don’t live in a simulation, I also prefer turning on the simulation. Thus I should turn it on.” As the paper does, I’m assuming that all the Margarets have perfect knowledge of the situation except for whether they are simulated or not. Otherwise it gets complicated. I’m also assuming that computer scientists like Margaret lead relatively happy lives, so they prefer existing over never having been created at all. (If not, I don’t think the Institutional Review Board would have approved all those simulations.)

      If we accept FDT and thus that Margaret should build the simulation, then the paper’s conclusion is weakened. Of course, even if Margaret should build the simulation, that doesn’t imply that we should look for flaws in our universe. For instance, the people simulating us might not worry that they are in a simulation, so Functional Decision Theory wouldn’t create acausal links. Thus it seems that it would be better for us to not look for bugs, since refraining from looking for bugs would not retroactively cause us to not exist in the first place, so we would stay alive. In contrast, if Margaret refrains from creating a simulation, then most Margarets retroactively never existed. (This apparent weirdness around causation is one reason why FDT is not the most accepted decision theory. But I think it fits this situation better than evidential decision theory or causal decision theory.) Since I accept that “the termination of our level of reality is many times worse than adding to our knowledge of reality is good,” I should therefore oppose cosmic ray (or whatever) research.

      But I believe my objections still stand:
      (1) There’s a chance that we are the basement-level civilization, and the cost of the basement never making simulations far outweighs the benefit of our civilization surviving until our simulation ends. And the cheapness factor implies that our simulation will end sooner than later anyway.
      (2) The cheapness factor also implies that I’m living in a small simulation. I’m willing to give up obscure basic research to preserve civilization, but I’m not willing to stay inside for the rest of my life for fear that the clouds are made of pixels.

  53. anonymousskimmer says:

    If they’re going around simulating universes, they have probably thought a lot about whether they themselves are a simulation, and simulation-related philosophy is probably a big part of their culture.

    Why? Because the ethics of computer and networking technology is such a big part of our culture? (No, it’s not, it’s a phenomena limited to a sub-culture and only broadly painted by the more general society)

  54. baconbits9 says:

    Wait a second, did the NYTs just conclude that there probably is an all powerful, all knowing being out there the created our universe and that the correct thing to do is pretend it doesn’t exist?

    • Jaskologist says:

      The fear of the Simulator is the beginning of wisdom.

    • The Nybbler says:

      Indeed, they’ve created the answer I usually have for Pascal’s Wager — a God who has created a universe that is carefully designed to have no evidence of God, and who will get very upset if his creations are not fooled.

  55. Rather orthogonal to this article, I find myself reminded of Greg Egan’s Dust Theory (which was never meant to be entirely serious, purely serving as a cosmology for Permutation City, but is still an interesting thought experiment, especially as presented in that book).

    In context of the article it might presume that if we are abruptly shut off, we might not even notice, because we’d just ‘find ourselves amongst the dust’. But you can see some refutations of the likelihood of this from Egan himself, in the FAQ above. 🙂

    Still fun to think about, though.

    (Also, I recommend the book to any hard science-fiction aficionado (that hasn’t already read Greg Egan and found him not to be to their liking, style-wise).)

  56. beleester says:

    Some other sources raise concern that we might get our simulation terminated by becoming too computationally intensive (maybe by running simulations of our own).

    This depends on the simulation method. If they’re doing atom-by-atom simulations, then it doesn’t matter whether the universe’s atoms are arranged into the shape of a computer or the shape of a rock, it’s going to be the same amount of stuff to keep track of. The activity of humans only becomes an issue if the simulation is capable of doing level-of-detail stuff like a video game, capable of recognizing that when a human has arranges enough copper and silicon in the right way, it should stop tracking “atoms” and start tracking “a computer”. But our universe seems remarkably consistent and not dependent on whether or not we’re looking at it – for example, special relativity applies to distant stars even though nobody would notice if our creators didn’t bother rendering all the details of gravitational lensing.

    • kenny says:

      I thought this as well. I’d guess the universe obeys a kind of ‘conservation of computronium’ and that, if our universe is a simulation, it’s all the same to the simulators or the simulation itself as to what proportion of the total computronium it is we or anyone else controls.

  57. Dick Illyes says:

    There is no matter. It only exists in this game and is an illusion. I started working with microprocessors in the 1980’s. Nobody then could conceive of the specialized processing power now available for almost nothing that powers our cellphones and creates the photorealistic games now commonplace, or the amazing color displays on every cellphone.

    Almost everyone commenting here assumes the current technology is all there is. Quantum computing is just getting started and there is no reason to believe it is the final technology. IMHO this experience is a shared dream, created by we the dreamers, using a technology incomprehensible to us, but analogous to what we are creating with processors using binary bits settings as the tool. We are on the verge of an explosion of affordable technology to induce virtual reality to us in this dream, and creating in what seems to be our world the idea of living in a simulation. Robert Monroe and Tom Campbell seem to have the best descriptions. We do have a non-human self-aware existence outside this experience. That is also is a simulation.

  58. materialist says:

    Don’t Fear The Simulators

    Whether we live in a simulation or not doesn’t matter at all unless we know what the simulation tries to achieve; it even may not try to achieve anything. Then it wouldn’t matter anymore.

    In case we are really living in a simulation a little bit of fear seems justified; it could be a very boring simulation from the viewpoint of a super-intelligent being. He/she/it could try to spice things up.

    In the end, the likelihood that this is just another conspiracy theory is quite high, any evidence is non existent. We probably should not fear simulated existence, but I think we have more to fear from the unjustified believe that we live in simulated worlds.

  59. Yair says:

    Hey, can the people running the simulation nerf the philistine-antintellectual-selfish-bully-arsehole META?

    It’s way OP atmo.

  60. deciusbrutus says:

    >>Some other sources raise concern that we might get our simulation terminated by becoming too computationally intensive (maybe by running simulations of our own).

    Why would running simulations of our own be more computationally intensive than other computing uses? Does taking a photograph of a photograph require more film?

    (Okay, possibly the simulators do some kind of simplification or compression, such as making quantum properties exactly equal to their average value unless they will be measured; that requires timeless knowledge about whether or not a given particle’s exact state will have macroscopic effects. That means that the simulation is timeless and has already been completed because every current state of the simulation contains information about the future of the simulation, in the form of knowing which particles’ exact states will have macroscopic effects in the future.

    A more palatable version would be that the exact state of everything is being simulated, rather than a compressed version that only simulates relevant things, and that the simulation does not do the equivalent of occlusion culling.

    (It’s also plausible that if the simulation IS doing the equivalent of occlusion culling, and we discover a bug in the simulation equivalent to the problem with back-plane clipping, that the simulators will patch that bug without terminating the simulation.

    Maybe heavier objects DID fall faster for Aristotle.

  61. avturchin says:

    (double)

  62. Dick Illyes says:

    In the news this morning is this new processor for AI:
    the world’s largest chip:

    named the Wafer Scale Engine • by Cerebras
    56 times larger than the biggest graphics processing unit ever made
    400,000 cores
    18 GB on-chip SRAM
    3,000 times more on-chip memory
    tightly coupled memory for efficient data access
    33,000 times more bandwidth
    extensive high bandwidth communication fabric
    groups of cores work together
    https://www.kurzweilai.net/making-headlines-a-breakthrough-processor-invented-for-ai

    Many on this thread declare the simulation idea impossible and are all stuck IMHO on their current understanding of technology. I have spent my life doing circuit boards and writing code for microprocessors. The explosion of technology since just the late 1970’s, which was not foreseen at that time outside science fiction, makes these observations appear very constricted. At the same time, photorealistic video games have made the idea of living in a simulation plausible.

    Anyone seriously interested in this should read Robert Monroe’s books. He began having involuntary out of body experiences, leading him to spend much of his life studying this. He claims that we do have a unique self-aware existence outside this experience, which appears to also be a simulation. There does seem to be a purpose. A Monroe student, physicist Tom Campbell, wrote a huge tome titled My Big TOE (Theory of Everything) where he explains things using digital computer analogies. His basic message is that we are trying to perfect ourselves and human society, perfection mainly being following the Golden Rule. We reincarnate endlessly. He claims we do have for all practical purposes, free will, and seeing what will happen is the main purpose of the exercise.

    I have found the observations of Monroe and Campbell interesting and the most likely explanation of what is going on.

    If most people believed we were living in a simulation, whose purpose was to create the best possible human society, we might do a better job. What if the Non-Aggression Principle became the recognized fundamental idea? What would a society based on it look like? Seems worth exploring.

    My short summary of the Non-Aggression Principle: To have the best possible human society, no one should initiate force against another, or deceive them so that they do something they would not otherwise do. Government should be limited to preventing force and fraud.

    Peace.

  63. lipby says:

    My take is that Bostrom is peddling warmed over philosophical idealism: https://planetslack.com/2019/09/freaking-out-about-simulated-universes

    As philosophers, Nick Bostrom and Preston Greene are surely aware of the philosophic tradition of quietism, a strain of philosophy that seeks defuse anxiety-ridden inquiries by reformulating questions in simpler, clearer terms. We would be better off quieting our minds about such far-off possibilities as simulated universes—though, of course, there are some weirdos who might find the idea fun to play around with. When confronted by Bishop Berkeley’s theory that all reality exists in the mind of God and that material objects didn’t truly exist, Samuel Johnson kicked a large stone, asserting: “I refuse it thus.”