Anyone who has played even a little of modern video games would be impressed at their realistic quality. The avatars on the screen have been programmed to look almost human. Now what if the designers could program the avatars to have thoughts as well, and that we make them do things by affecting their thinking and putting thoughts in their heads. So rather than making them do things, we give them the thought that makes them do things, as if they had decided to do so themsleves. Then the avatars might think that they are real and acting on their own volition and not being manipulated by an outside agency.
This is the basis on an idea that I keep running into from time to time, that we ourselves are the products of a computer simulation that has been created by an advanced post-human race. Philosopher Nick Bostrom explains this idea in a paper whose abstract reads as follows.
This paper argues that at least one of the following propositions is true: (1) the human species is very likely to go extinct before reaching a “posthuman” stage; (2) any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof); (3) we are almost certainly living in a computer simulation. It follows that the belief that there is a significant chance that we will one day become posthumans who run ancestor-simulations is false, unless we are currently living in a simulation. A number of other consequences of this result are also discussed.
The opening paragraph of his paper lays the situation out quite clearly.
Many works of science fiction as well as some forecasts by serious technologists and futurologists predict that enormous amounts of computing power will be available in the future. Let us suppose for a moment that these predictions are correct. One thing that later generations might do with their super-powerful computers is run detailed simulations of their forebears or of people like their forebears. Because their computers would be so powerful, they could run a great many such simulations. Suppose that these simulated people are conscious (as they would be if the simulations were sufficiently fine-grained and if a certain quite widely accepted position in the philosophy of mind is correct). Then it could be the case that the vast majority of minds like ours do not belong to the original race but rather to people simulated by the advanced descendants of an original race. It is then possible to argue that, if this were the case, we would be rational to think that we are likely among the simulated minds rather than among the original biological ones. Therefore, if we don’t think that we are currently living in a computer simulation, we are not entitled to believe that we will have descendants who will run lots of such simulations of their forebears. That is the basic idea.
Via Machines Like Us, I came across this 50-minute documentary that walks the viewer through the history of science that looks at successive answers to the so-called fine-tuning problem. i.e., how could it be that the universe seems to have just the properties necessary to create and sustain life? It takes us through the various alternative naturalistic answers to the ever-popular “God did it”.
I felt that documentary maker went too far in trying to make everything all mysterious and deep and solemn, with the choice of music and backdrops. Astronomer Martin Rees, dressed all in black and looking like a vampire while also reminding me of the master of horror Vincent Price, poses and speaks in a melodramatic way that I found both annoying and risible. But if you are willing to overlook this cheesiness, the program is pretty good at laying out the issues involved.
The show also has mathematician John Conway speaking about his game Life that shows how computers programmed with a very few simple rules can over time create things that behave a lot like living things, suggesting that they way real life came about could be by similar simple rules. As he says, “We are atoms, time, and mathematics”.
Nick Bostrom also appears, explaining how the computer simulation idea emerges out of the multiverse concept.
The idea that we may be simulations, the imaginings of an advanced post-human simulation is pretty interesting, if mind-boggling, stuff. It is hard, maybe impossible, to show that it is false, the same way it is hard to argue against solipsism. But that does not mean it is true, either.
The biggest problem I have with this idea is the notion of running a very, very large number of simulations at the required level of detail… Even if you assume that you have access to infinite computing power, the problem is that you’re just going to end up generating a ridiculously large number of practically identical simulations, with no means to choose between them. There are a more-or-less infinite number of possible histories which could lead to now, mostly differentiated by things like what some completely unimportant and unrecorded contemporary of Julius Caesar had for breakfast on any given day -- what possible use can it be to simulate even a meaningful fraction of them? But somewhere amongst all that noise, there are an unknowable number of minor events which could change the course of history in significant ways -- such as Adolph Hitler failing to look before crossing the road one particular day in 1904 and getting fatally run over by a carriage. The signal-to-noise ratio is awful.
In short, I’m deeply unconvinced that this level of simulation actually serves any sane purpose, so I’m going for option 2.
To quote a character from an Iain M Banks novel in which this belief is the dominant religion: “Any theory which causes solipsism to seem just as likely an explanation for the phenomena it seeks to describe ought to be held in the utmost suspicion.”
Then there’s the problem that you’re trying to simulate the evolution of a posthuman society which can simulate its own evolution, so you’re going to need a simulation which can contain nested simulations of an equivalent level of complexity. Therefore it seems to me that you’re going to need a computer which is substantially more powerful that itself… I suspect that (much like the halting problem) this is actually impossible.
Whether or not this is true or possible, I still find it to be a bizarre idea, very interesting to think about.
This sort of reminded me of this game from way back when that I played in middle school -- Final Fantasy X, haha but anyway it turns out the main character is a memory of a person who lived 1000 years ago, or something to that effect. Anyways it was pretty cool and sad (when his love interest runs to him and runs through him…buh-huh-huh!!), definitely makes you think.
Why assume that the simulation is run by the descendants of humans? It could just as well be run by a completely non-human race or entity. There might well be no humans at all outside of the simulation, now or ever.
Maybe Douglas Adams was closer to the mark, and the simulation is being run by the mice.
I don’t understand why anyone cares. If I am a good enough simulation that I mistake myself for real (or am programmed not to doubt my reality) then for all intents and purposes, I am. Real enough for me, anyway.
It’s as if a couple philosophers decided, “wait wait they’re taking us seriously. we can’t have that! let’s come up with something really silly to argue about!” I think Bostrom’s also one of the “we can solve death” crowd. (eyeroll)
Why assume that the simulation is run by the descendants of humans?
Correct. We’re just a crowd-scene in some hypertechnological society’s version of “Monsters, Inc.” or a first person shooter -- their equivalent of the zillions of BEMs I’ve killed in Halo.
And one of the things they simulated was philosophers who ask silly questions to wank over. They did that part exceptionally well.
What bothers me most about this “hypothesis” is the extremely lazy thinking on the part of people like Bostrom.
For example, the “logical conclusion” Bostrom reaches in the abstract is a non sequitur. It assumes many facts not in evidence: that simulations are capable of experiencing consciousness, that computing power will ever reach the level that an entire universe can be simulated (see Dunc’s post for why that’s almost certainly false), that a civilization capable of creating simulations capable of consciousness wouldn’t find it ethically abhorrent to do so.
Moreover, let’s suppose some advanced “posthuman” civilization is actually capable of creating simulations of the entire universe containing conscious beings. Given that such a civilization would have presumably advanced moral philosophy centuries beyond what we have today, as well as the science behind consciousness, it is logical to assume that they would a) know whether their simulations were capable of consciousness, and b) understand the ethical problems inherent in intentionally subjecting other conscious beings to pain and suffering. Therefore, if we were living in a simulation created by posthumans, there should be little or no pain or suffering.
That’s a far more logically sound conclusion than Bostrom’s (or any of the other pseudophilosophers who believe in this tripe).
For example, the “logical conclusion” Bostrom reaches in the abstract is a non sequitur. It assumes many facts not in evidence
Well, yes!!
We could just as easily assume that we are all, in fact, in a gigantic bowl of jell-O. Of course it has been produced in such a way that we can’t tell that we are, yadda yadda. Take the old brain in a box argument and add a bit of high tech and a whiff of immortality wish-fulfillment and presto!!! This is exactly why philosophers of science try to build barriers to prevent ideas that are unsupported (and unsupportable) from gaining ground: we’d be overrun with them because there’s no reason not to believe everything that is possibly believable. We should ask Bostom: “what reason do you have to believe that this is actually the case?” and when he says, “none” we reply “maybe you should stop spewing hypotheticals until you have some evidence supporting them? that’s why scientists do it that way, rather than the way philosophers do it.”
If you take Bostrom’s argument and replace the simulation/computer stuff with gods that poofed us into being yesterday, how’s it any different? And, therefore, just as believable.
While someone running a simulation would have that ability, it’s irrelevant to the argument. The simulated beings could have their own thoughts arising from their simulated physics, just as our thoughts arise from our physics (simulated or not). Beware the trap of dualism.
Why wouldn’t we be able to solve death? Sure, we’ll eventually run out of negentropy, but otherwise it seems like solvability should be the default extrapolation of our current knowledge, even if it is a way off.
I ask myself a similar question every time I stand up; how could it be that my legs are exactly the right length to reach the floor?
He acknowledges that there is some dispute on this. It’s an assumption, but explicitly so, and is supported by its own arguments elsewhere.
Did you read the whole argument? “Simulating the entire universe down to the quantum level is obviously infeasible, unless radically new physics is discovered. But in order to get a realistic simulation of human experience, much less is needed…”
These ethical concerns are also addressed: “However, from our present point of view, it is not clear that creating a human race is immoral. On the contrary, we tend to view the existence of our race as constituting a great ethical value. Moreover, convergence on an ethical view of the immorality of running ancestor-simulations is not enough: it must be combined with convergence on a civilization-wide social structure that enables activities considered immoral to be effectively banned.”
Since they are more advanced than us (hence different from us), they will therefore agree with us on this point? Your conclusion might be true, but it doesn’t follow from your premises. And as Bostrom points out, it isn’t sufficient to refute his argument.
I don’t think Bostrom understands computer simulation. As we currently know it, it is the execution of a mathematical model which is accurate only to some degree or other. As others have pointed out the degree of detail is a major obstacle to this suggestion. So is the actual form of objects within the simulation -- definitely not 3-D thinking, feeling, growing, etc. And as someone else noted above how can there be a computer big enough to simulate a universe? It would have to be bigger than the universe. So if we are in a multiverse…..?
Philosophers must be really grabbing at straws in their attempts to simulate relevance if they can propose this kind of idea and expect to be taken seriously.
What I’ve pointed out is sufficient to demonstrate that Bostrom’s chain of deductive logic introduced in the abstract is not justified.
Okay, so he admits that he’s made assumptions. Fine. Now he should be a big boy and phrase his arguments in such a way that they recognize that, rather than making absolute statements such as, “It follows that the belief that there is a significant chance that we will one day become posthumans who run ancestor-simulations is false, unless we are currently living in a simulation.”
That statement, as worded, is unambiguously and unavoidably wrong. Yet it’s the statement he’s chosen to allow to stand out on the very front cover of his paper, and it’s the statement that reporters and bloggers will slap onto the top of their articles (even our gracious host included).
A realistic simulation of human experience does require less computing power than simulating an entire universe, but if he wants to argue that simlations can have consciousness, they need to be extremely detailed simulations. At the very minimum, an entire brain must be simulated. Plus the entirety of experiences, observations, and memories of that brain.
And then, since Bostrom’s argument is that it is virtually certain that we are in a simulation, computing power must be able to simulate an entire world’s population of brains over the entire course of human history, reiterated many times.
We’re probably talking about a quantum computer with sub-nanometer process that is the size of a planet.
And seriously, Bostrom thinks the ethical issues can just be waved away? Any human with half a brain can see the suffering inherent in this world, and the problems with forcing trillions or more of conscious minds to endure that suffering are obvious to anyone who isn’t a sociopath.
A society with advanced moral philosophy would recognize the duty it has to stop an immoral simulation at all costs. If simulations mean that all the suffering in the universe is multiplied millions of times over, then a civilization capable of running simulations would have a moral obligation to destroy itself if it were the only way to prevent those simulations.
And the moral argument does refute Bostrom’s argument. Remember, his argument isn’t just that simulations are theoretically possible. It’s that he can know with (near) certainty that we are actually in one.
@Dunc -- I don’t think the “complexity” argument holds much water. You’re assuming that the universe in which the simulators exist is of the same size and complexity of the simulation itself (us). This is an assumption that is probably not true if we are indeed in a simulation--surely the Simulator’s universe must, by necessity, be larger than the Simulation itself.
Perhaps our universe is incredibly simple by comparison, in which case it is of no consequence how complex our Universe appears to us.
It wouldn’t be necessary to simulate the entire universe — just one planet, or parts of one planet. Or maybe just one brain (mine, of course, since I know I exist; I’m not so sure about you folks).
Surely such advanced technologists would be familiar with the optimization technique known as lazy evaluation. You don’t compute a result until you need it. That way if you never need it you never have to compute it at all.
If you play a video game like Oblivion, for example, you’ll wander through a vast and incredibly detailed landscape. But the game engine doesn’t have to continuously keep track of every twig of every tree, even when they’re off-screen. It needs only to generate the local landscape wherever your character is at a given time.
In a multiplayer game, the different landscapes local to the different players need to be kept consistent with each other. That fact introduces complications, but the principle of lazy evaluation still applies.
You don’t have to simulate the moon’s appearance unless some character is looking at it. You don’t have to generate the smaller craters until one of your players invents the telescope. In fact your simulation engine can ignore most of the universe until simulated telescopes come along. Then you can generate the stars and galaxies as needed, but not in much detail.
The resulting world is something like Schrodinger’s cat. You don’t have to model the simulated cat until somebody opens the simulated box.
Except that the argument isn’t whether it’s possible to be living in a simulation, but whether it’s almost certain that we all are. And that’s an argument of numbers: there must be overwhelmingly more simulated consciousnesses than there are real consciousnesses, so computing power does become an issue.
Furthermore, while a simulation wouldn’t have to render every cat in every box in the whole world, it would have to have indefinitely long memory of every little detail. If I put a cat in a box, it must be there when I open that box decades in the future. And if a leaf falls off a tree, it had better still be on the ground tomorrow.
My problem with the argument is of the same sort as to why I object to ‘God’ being the creator of the universe: okay, but what created god? It doesn’t answer the question, it just moves the goalposts.
If we are elements of some sort of ‘artificial’ simulation, then that simulation is functioning within a wider reality. We may not know anything about it, but we would actually be built out of that in the same way that a game character is made out of matter and energy. So what created that?
I’m not trying to address all the objections. I’m addressing only the objection that a simulation of the entire universe would require a simulator with more complexity than the universe that it’s simulating. On the contrary, with the right optimizations the simulator could be less complex than what it’s simulating.
The simulatior wouldn’t have to retain a detailed memory of all of history. It would need to keep track only of the current state of the simulation. (Due to relativistic considerations the notion of a “current state” is a bit slippery: two points in space-time may or may not be simultaneous, depending on the reference frame. I assume that the uber-geeks would be able to find a way to handle that.)
If a leaf falls off a tree, it doesn’t have to be still on the ground tomorrow unless somebody noticed it there today, and somebody would notice its absence tomorrow, and its absence couldn’t be accounted for by a gust of wind. If necessary, the simulator could back up for a redo in order to resolve these kinds of issues.
Conveniently, the Second Law of Thermodynamics reduces the need for the retention of state. An increase of entropy is a loss of information, and losing information means that you don’t have to store it any more. In the computer world we call it garbage collection: the recovery and reuse of memory previously allocated for some purpose but no longer needed for that purpose.
The universe-as-a-simulation is like a multiplayer version of the brain-in-a-vat. Neither notion is useful unless it makes testable predictions, or accounts for otherwise inexplicable observations.
For example: maybe the Planck time reflects the limited precision of the uber-computer’s floating point processer, whose word size is only 10^39 bits.
Likewise, maybe quantum uncertainty reflects certain computational limits, or the use of numerical approximations when solving differential equations. Or maybe it reflects the need for some slipperiness when reconciling the potentially conflicting perspectives of different players in the game.
If someone could explain the one in terms of the other, he or she might be on to something. Until then, the whole idea is just an amusing speculation — and the notion that it is almost certainly true is preposterous.
Agreed — moving the goalposts. Let me express what amounts to the same objection a little differently.
invivoMark indicates that Bostrom regards it as almost certain that we are living in a simulation. I don’t see that conclusion in the brief excerpts quoted here, but I haven’t read the original paper, so I’ll accept invivoMark’s interpretation for the nonce.
On that basis, Bastrom seems to be saying that, because of reasons A, B, and C, we are almost certainly living in a simulation created by a post-human civilization.
If the post-human civilization is descended from a race of proto-humans like us, then reasons A, B, and C apply with equal force to them.
Therefore there was almost certainly a race of proto-proto-humans. And so forth, ad infinitum. Reductio ad absurdum.
The first story based on this premise was by Stanislaw Lem in the 1960s, I have forgotten the title. I think the story is in the anthology “Nacht Und Schimmel”.
Are you a Boltzmann Brain?
Nope. His conclusion is that one of three things is very likely. You’re ignoring two of them.
No, he doesn’t. Have you still not read the whole argument? Overall, we’re morally advanced enough to know that we should not keep people chained in our basements. But a few of us still do that. Bostrom is saying that even if simulations are generally considered unacceptable, some people may do it anyway.
If it were only the suffering that would be multiplied, maybe. But the simulations would contain lots of worthwhile things too. Maybe not enough for society to generally condone them, but it could easily be enough to convince at least some people that living is still better than dying.
That is one possible solution, but it only makes my main objection (“what actual purpose does all this detail serve?”) even more acute. If you’re trying to simulate your own history and evolution, why would you need to do it at a sufficient level of complexity to simulate an entire planet full of living beings, but not at anything like the actual level of complexity of the world you’re trying to simulate? The massive amount of arbitrary, useless detail becomes even more arbitrary and useless than it would be if you were actually simulating the world fully.
I’ve always felt that if god existed, it’d be in the form of the programmer of our universe.
and our programmers work in mysterious ways.
it’s clear many of my fellow readers here do not have a background in modeling/simulation. it’s impossible to tell what their purpose is, much like how it’s difficult to determine the purpose of a research study by looking at a single rat that was used.
the objections that small random factors can have a big impact is in fact a non-issue, just simulate the universe many times over, and in fact is very useful to programmers in determining things like probabilities/stability of a model.
the objections that our universe cant contain a similarly complex universe should remember that for decades already, we’ve already use simulations that are no where near as complex as what they are trying to model, yet still provide very useful information. imagining how much more complex that universe is is like a rat trying to imagine the world outside of its cage.
and ethically speaking, again one should look at what we already do, rather than trying to imagine what’s being done to us. we already acknowledge animals as alive and capable of suffering pain, although at a lower level of consciousness than us. we already experiment on and kill and eat (and let our sadistic children torture as housepets) countless animals. a similarly higher level of consciousness beyond our comprehension would similarly view us, assuming that we were anything more to them than video game sprites killing each other.
What if it is not a computer at all that controls the simulation? What if our ignorant way of thinking is prohibiting us from opening our minds to any and all possibilities? What if this is a simulation and it is designed in such a way to hide or make you ignorant of certain truth’s so that you can never figure it out? When we make games, rarely do we choose to make games about our past. Why is it you think that is what a simulation would be for?
Let’s say you are highly advanced race, you have reached a point where learning new things is hard because you think you know everything. You get bored, so you decide to create a simulation. The rules of your simulation allow you to live an entire life time in 1 month of your true reality. You set it up so it is based on the same laws of your reality. Your eventual goal is to make it so your simulation reaches your present day in your reality and surpasses it in a fraction of the time it took your reality to do so. You want to learn more but you want to learn it faster. You learn things in the simulation that would otherwise take you generations in your own reality. Does it have to be exact to your reality? NO! You wish to learn new things, so you make it its own allowing you to learn things that you might have missed in yours.
Its really not that hard to imagine, so why would it be impossible to create?
I am no physicist, scientist or anything really… I am just a man with limited intelligence. I do have one advantage though.. I have always had an open mind and a great imagination. Ignorance is a word that is foreign to me.
See my blog post on this topic:
Why You Are Not Living in a Computer Simulation
M Mahin, I do appreciate you correcting your “Why you are not living in a simulation” topic with “I will now discuss some reasons why such a possibility is extremely unlikely.” before you go into reasoning. I am going to do my best to reply.
Technical feasibility
The reasons that you and Bostrom both give are pure speculation. The numbers are based on the knowledge the two of you have and nothing more. To pretend that the few answers that our reality has given us is enough to understand what it would take for an advanced alien race or even computer race that ‘created’ us is foolish. It reminds me of the earth is flat argument, everyone was an idiot for saying it wasn’t. To your point about “We have no idea whether it is actually possible to produce human-like consciousness in a computer, so this is a very questionable assumption.” You assume that the consciousness in the computer was based on a consciousness that exists in the creators reality. You also talk of the consciousness believing it is living on earth but really it lives in a simulation. Living is a word you cannot use because every consciousness in the simulation is simulated to act like a living being, not be a living being.
Motive of it’s creator
I don’t really need to say too much here, you once again assume to have knowledge of what a being in this position would want or need from a simulation like this. Basing it on examples of what you or others in your reality would choose to do with a simulation is pointless.
Realism imperfections
If the simulation had imperfections and we noticed them, do you not think it logical of the creator to ensure we don’t catch on? When games glitch in our reality, we get expansions to try and fix them. Not that our reality is anything close to what we are discussing, it is just a simplified example.
Moral issues
Why do you assume that you are not also a simulation? Why do you assume observer means the observer must consciously take part in the simulation? As for you becoming a bad participant in the simulation if you found out you were a simulation.. you are not meant to find out. At least, I would assume that we are all not meant to find out. What if the point is for us to simulate our lives so the observers can gain whatever they need to gain from it?
The earth turned out to be a sphere and not flat like everybody once thought. You would figure that would have taught us something about ignorance…
To me there are innumerable benefits to running a simulated world if such technology was available now, today.
I think most people would say that they get wiser as they get older and by the time they are 57, like me, they are generally much wiser than when a teenager. How many of life’s pitfalls could we evade, people we could avoided accidentally hurting and opportunities that could be taken, if we were able to start our lives again.
Imagine a key person plugged in to some virtual world computer with their present (real-world) knowledge disabled, or have their neural network coding downloaded to the virtual machine instead. Here they could live out a 70-year life over a single real(mmm!) day. Everything in the virtual environment (including other people) being simulated for the coded person. Think of the wisdom of life that you would have gained once that acquired knowledge was uploaded back to your brain and the real-world knowledge switched back on, and this is before even setting out on the start of your university/college career.
Sometimes we go to university doing subjects we realise later on in life were not meaningful and not what really stimulated us. Wouldn’t be good to run such a world simulation for each and every youth at 18, or even younger, before they are unleashed into the world sometimes like a loose cannon. This would allow every youth to be wiser and more able to see who they really are and thus better prepared to decide what direction they wish to take in life. For some it would be to avoid the mistakes of their virtual experience, for others it might be an ability to better hone their career decisions to gain a more purposeful life. Most now would be highly motivated in the real life, making one grateful for in effect a second chance to do the things that really mattered. The benefit to society motivationally and thus economically (also through reduced crime) might well be considerable.
You could even imagine companies wanting candidates coming for job interviews at some point be immersed in a virtual experience in order to see if the person is suitable for a job. Or perhaps the law enforcement agencies using such virtual experiences to help adjust a criminal’s wayward behaviour.