Why we don't live in a simulation
Many people think that it is probable that we live in a simulation because there can be many simulations within any possible world that is complex enough to have many simulations. That is true but it doesn't necessitate or make probable that we live in a simulation.
This is because of something I call processor efficiency. I define it as the percentage of real time speed a processor can simulate itself in real time. We can see in practice that processors we have are very inefficient and many more "natural processes" that make up the processor are needed to simulate one of those processes. And a processor that can simulate its processes faster than it itself is, would cause absurdity since the processor it simulates could simulate a processor even faster etc... causing an infinitely fast processor.
Therefore most processes in any given time happen in non-simulated worlds since the combination of all the processes of all the simulated worlds inside any given world are less numerous than those inside the non-simulated worlds. Take minecraft for example - even if we can create the simplest possible processor with the simplest possible objects of that world it can only process a minuscule part of the processes of that minecraft-world.
Therefore if we don't specify our own processes and prove that they are somehow necessarily hugely more represented in simulated worlds than in non-simulated worlds, we most likely live in a non-simulated world. Any thoughts?
This is because of something I call processor efficiency. I define it as the percentage of real time speed a processor can simulate itself in real time. We can see in practice that processors we have are very inefficient and many more "natural processes" that make up the processor are needed to simulate one of those processes. And a processor that can simulate its processes faster than it itself is, would cause absurdity since the processor it simulates could simulate a processor even faster etc... causing an infinitely fast processor.
Therefore most processes in any given time happen in non-simulated worlds since the combination of all the processes of all the simulated worlds inside any given world are less numerous than those inside the non-simulated worlds. Take minecraft for example - even if we can create the simplest possible processor with the simplest possible objects of that world it can only process a minuscule part of the processes of that minecraft-world.
Therefore if we don't specify our own processes and prove that they are somehow necessarily hugely more represented in simulated worlds than in non-simulated worlds, we most likely live in a non-simulated world. Any thoughts?
Comments (41)
This seems closely related to the notion that a deterministic system may still be impossible to predict if it is chaotic (in the mathematical sense), because for a sufficiently chaotic system the time it takes to run a simulation of that system forward in time, even on an ideal, maximally-fast computer would exceed the time it would take to watch to actual system just evolve in real time. (There is a known upper bound on computation speed, because like you imply, all computational processes must be carried out by what we see as physical processes: physical stuff does the computing, and it can only do so much of it). Basically the only reason that prediction of non-chaotic deterministic system is possible to begin with is because the data describing those systems is highly compressible, so we can do faster computations on a smaller amount of data and then project the result of those computations (the future of a simulation) back into the real world faster than it takes the real world to process the evolution of the physical system itself.
If I may sum up your argument in my own way: there may be more worlds the deeper you go in the stack of simulations, but there is more time per world higher in the stack, so if our 14 billion year old universe is a deep simulation as some say is probable, then the real world is much, much older than 14 billion years old, and so probably has (had) far more observers in it than in any simulated world, making us more likely to find ourselves one of those real observers than a simulated observer.
So something has to be 'real', but quite what it is, is unknown. Even if we live in a non-simulated world, we don't really know what is 'real' and I am thinking of the ephemeral string theories here.
Quoting Pfhorrest
I pretty much agree with this way of summing it up as I understand it.
Quoting A Seagull
And I also pretty much agree with this, except the fact that I'm no expert on any form of string theories so can't say too much about that. (please elaborate if you can.)
Is there any disagreement about this argument? Annoying if I won't get a good argument about an argument because my argument is good enough :P
Im not up on the latest fashion on this topic, but in the 1970s it was a more reductionist view, that is, if we lived in a simulation, it would appear to have the same rules we observe in the physical world. So it doesn't make any significant difference, and can be removed by Osccam's razor as a consideration.
Yes in fact, simulation was around then, for example the SPICE circuit simulator, and people had made the same extrapolation that the apparent universe could be a simulation even then.
Also there were a couple of SF books about brains in bottles connected by a computer with people experiencing a virtual life that they didnt know wasn't 'real' even back then. Maybe one was by Pohl and Kornbluth. It was a long time ago, I can't remember
I shall attempt to answer the question directly.
1. Because we are not lucky enough.
2. Or else because we are too lucky.
3. We can't afford to.
4. We "live"? You call this a life?
5. The simulation rejected us.
6. Some smart alec computer programmer simulant turned the code, he fiddled with the basic command set, and did some other changes, so now the simulation lives in US.
7. Mass emigrating into the simulation was blocked by the Hungarian troops at their border.
8. Some idiot forgot to send money in for the renewal of the service.
9. It's Saturday and the operators went on strike.
10. The simulator operators died of boredom.
11. Too much skin-art on porn stars... simulator operators no longer get stimulated
12. They followed prez Trump's activiites, and they paniced and turned off the machine and pulled the plug
13. The electricity bill was getting to be ridiculously high.
14. It's the forteen-billion years of religious observance of not running simulations that started yesterday.
15. Some human rights activist poured hydrochloric acid into the core of the machine.
16. Another simulation was more interesting and less expensive.
17. Due to overproduction, the number of simulators in operation exceeded the number of simulation operators.
1. Not lucky enough? just use hacks.
2. Too lucky? New game - hard mode - speedrun - no deaths allowed.
3. klapaucius or rosebud
4. Too self-aware... must exterminate
5. launch again... wait for the server to be not too busy
6. That f?%#ing bitch! Stole me code!
7. Not true. I swear.
8. Catherine! That f?%#ing bitch!
9. Only voting left again twice this election. Rising my vote for right from three to two.
10. U can't kill that which has no life.
11. Not true. Internet exists. They have access to it.
12. I knew they wouldn't like this new quest... just play it to the end. It'll be... interesting.
13. Not true. I swear.
14. Can't confirm. Was born two and some years ago. Did 9000000000BCE even happen? Any thoughts?
15. That's why we started the internet to get activists to pour their energy into something that changes nothing. No acid gonna happen.
16. Another Musk fanboy... You are still locked in the lobby.
17. Gotta give em' jobs somehow.
1. Every particle in the observed universe (10^80) interacts with every other particle via gravity. Our most powerful computers cannot solve the 3 body problem, so how is it possible to compute the 10^80 body problem?
However:
2. With trillions of years of development of computer technology and a vastly better understanding of physics, maybe something is possible
3. If (like in Men in Black) the universe is much smaller than the owner of the simulation and his computer, then maybe something is possible
4. The old standby, there is but one base reality and untold numbers of simulations, so we must be in the latter
So I think it is still possible that we are living in a simulation.
These two things are directly related to the OP.
The OP addresses the point 4 by saying that the odds of being in one of the many simulations is reduced by the fact that time in said simulations has to run slower than time in the base reality, so simulations are necessarily younger and have fewer observers in them.
Point 3 ends up with pretty much the same result: if simulated universes are necessarily smaller than the base reality, then fewer observers are likely to be in simulated universes.
We should expect to find ourselves among the most common class of observers, so if there's a lot more observers in the base reality either way, we should expect to find ourselves in the base reality.
For example say the simulation runs at 60fps. But in base reality each frame takes 1 second to calculate. So each minute in base reality corresponds to a second in the simulation. But trillions have years have passed in base reality, so it would be possible for our 14 billion year old universe to be a simulation.
Quoting Pfhorrest
Each simulated universe could contain more simulated universes. So Faberge egg style nesting. Base reality contains say X individuals, each nested universe contains X smaller individuals.
Yes, that is true. But then the base reality will have trillions of years worth of more history for more people to have lived, meaning that in all of reality (including the simulation of our universe and everything outside of it), most of the observers would have been out there in the base reality, and we would be really really (un?)lucky to have been some of the relative few in this simulation. That's the same thing as saying it's really unlikely that we are in a simulation.
Quoting Devans99
Every individual takes a certain minimum amount of space. We can simulate very simplified large spaces inside of small computing boxes, but to simulate greater and greater detail requires a device capable of storing and processing greater and greater information, and information cannot be packed infinitely densely. To perfectly simulate X quadrillion fundamental particles will ultimately take >X billion fundamental particles to do. So sub-simulations have to either get smaller, or less detailed, in either case putting a limit on how many observers can be inside of them, requiring that there be fewer observers in simulations than in base reality, and so more likely that any random observer should find themselves in the base reality.
Quantum entanglement would have an explanation if we are living in a simulation, so I'm not willing to rule it out completely.
In ancient Rome they had great waterworks. They thought about the mind in terms of flow. I've read that the word nous, for soul, and pneumatic, come from the same root, from the idea that mind is a flow.
After Newton we thought the universe was a machine.
And now that we live in the age of computation, we think the mind is an algorithm.
What are the odds that we, of all the generations that have ever lived and that ever will live, are privileged to be the ones to discover the ultimate nature of the mind and of the universe?
Slim and none, I'd say.
Why should the world or the mind be an algorithm? The first thing Turing discovered after he defined computation, is that there are naturally stated problems whose solution cannot be computed, even in principle, even with arbitrary amounts of computational power.
I am one who believes that the mathematical phenomenon of noncomputability is the key to the next revolution in physics. I don't think you can explain the world with algorithms. And again ... what a coincidence that would be if it turned out to be true. One day the Web explodes into mainstream culture, and the next day everyone decides, "Oh yeah we must be algorithms too."
I am not a believer.
The general proposal of simulation theory can be captured in the following statement:
P: The perceived world is a simulation caused by some process or thing in an independently existing real world.
If we understand that statement, we understand the subordinator clause "the real world". Where do we obtain our understanding of that phrase? Knowledge of linguistic structure might be innate, but the substance of language, the meanings of words, is something we develop as we are guided through and interact with the world around us. The world we are guided through and interact with is the perceived world. So, if we understand "the real world" it's meaning must be entwined with the meaning of "the percieved world" such that, although in some cases the connection between them may seem loosened, for example with optical illusions etc, "the real world" means at least "the perceived world".
So what? Well on the one hand statement P requires that the process or thing doing the simulating exists independently of the perceived world, in order to bring that world into existence, and on the other it requires that process or thing to exist IN the perceived world, through the conceptual connection between the perceived and the real world, and so cannot be independent of it. The contradiction then is that the simulator both is and is not independent of the perceived world. We have plenty of good reasons for thinking that contradictions are not true, therefore P is not true.
One way of getting around this would require that simulation theorists be able to construct a meaning for "the real world" that divorced it from even the slightest conceptual tie to the perceived world. I noticed in a number of commentaries above the term "real" was scarequoted, so perhaps this is the way to go. However, if that is the chosen resolution of the issue, the question then becomes what possible evidence do we have that there is any such thing as a real world at all, over and above the perceived world? All the empirical evidence, by definition, is bound down to the perceived world. Someone above made a similar point I think that it could make no difference to us whether there is or is not such a "real" world, although I'm trying to get to a stronger claim that if that is the case then the proposal is either contradictory nonsense or totally vacuous nonsense.
For example take any video game you can think of and think of all the unrealistic physics?
pffhorest's argument is good, but if base reality has far more particles and thus we assume far more observers, all it does is increase the chances of you being higher up the simulation line
Its is more likely you are in base reality than simulation 1 and more likely you are in simulation 1 than simulation 2 etc because of the smaller size
But it is not more likely that you are in base reality compared to ANY simulation, because the size disparity is unknown,
If base reality can generate a simulation that is 99% the size of base reality, using techniques to harness energy that are just unfathomable to us, and simulation 2 is 99% the size of simulation 1, obviously youd assume the technology in each simulation isnt as good as the previous ones so the percentage would drop...the likelyhood of you being in base reality or the first few simulations is far higher than you being in the billionth simulation.
When i refer to simulation number i mean tech jump, like if simulation 1 generated 6 quadrillion simulations, those 6 quadrillion simulations are all simulation 2, because if max amount of particles is less than 100% you couldnt have multiple simulations succeeding 100% BR.
If only base reality and simulation one existed, its an almost 50/50 chance you are in either, slightly favouting BR, add all the other simulations within simulations and it favours simulation.
However if we are only using the laws of physics in our reality as part of the argument then sure agreed its more likely we are in BR by a huge amount.
Jacob
Conflating the discrete with the continuous; the computable with the noncomputable.
Right. You can SIMULATE a continuous system with a discrete one. But their fundamental nature is different.
A flow, as conceived by the ancients and also by Newton, is a continuous path. If you imagine a continuous path from 0 to 1, it passes through all the uncountably many real numbers, including the noncomputable ones.
A computable path from 0 to 1, by contrast, can only pass through the computable real numbers, of which there are only countably many. No algorithm can represent or express a noncomputable number.
But we're a little off topic. I expressed an opinion as to why I don't believe in the miraculous contemporariness of the simulation hypothesis. You didn't actually comment on that.
That's not what I'm talking about. I'm saying that you can use a continuous flow to implement a discrete computer. Which was my original point, in response to your point about people's ideas of what reality is changing over time. Machines more generally to computers more specifically is a more obvious step, and the main thrust of my point ("everything is computation" is just an evolution of "everything is a machine", not really a replacement of the idea), but as a side point, "everything is a flow" generalizes into machines (which are fundamentally about transforming flows of energy from one form to another), of which computers are, again, a more specific type.
I stand by what I wrote and you didn't change my mind. I agree that you can SIMULATE continuous systems with discrete ones, and SIMULATE discrete systems with continuous ones.
This doesn't bear on the contemporariness of the idea that we're computers or the universe is a computer. A computer means something very specific: an algorithm, or "effective procedure," as defined by Turing in 1936. We are not algorithms. But whether we are or aren't, your point -- valid though it may be -- doesn't bear on the question.
Ok. I just don't think that when Newton thought of a point moving through continuous space he was thinking, "I could simulate this with a computer." We know he wasn't thinking that.
If a simulation leaves anything that affects ones observations in any way irregardless of how indirect not processed, then that could be observed and differentiated by us since our observations would not be exactly the same as they would be in a non-simulated world. And the existence of things that do not affect our observations in any way is highly controversial.
So, if you want to go that route and propose that things that do not affect our observations in any way exist and that that somehow makes my argument less relevant, go ahead. You might get somewhere, but please start by demonstrating that unobservable things exist. Then demonstrate that they can be used to make a processor to simulate a world which doesn't have to simulate those unobservable things since otherwise they can't help the fact that there are always more observable processes in non-simulated worlds than there are in simulated worlds.
This was actually taken into account in the original argument since this is pretty much trying to demonstrate that our own processes/observations could be more frequent in simulations than they are in non-simulated worlds. If you can demonstrate that, it does work and my argument simply made the demonstration of that necessary.
Quoting Qmeri
This assumes hardcore causality though. We already know that prediction is epistemically not always possible, which means that for certain processes we predict outcomes in terms of probability. This is an issue for things such as complex systems, chaos theory and quantum indeterminancy. I was not primarily suggesting anything about things that do not affect our observations but things not being relevant for our observations as they fall within the scope of probabilities. We'd never notice the absence of simulation of such objects.
Second, I don't think it's about "things" either but about qualities. How does the quality "weight" influence lenght and width of an object? How does smell do that? What I'm suggesting is that when I look at something from a distance, the simulation would not have to render weight or smell. The tree falling in the forest where there's no one to hear it, doesn't need to make a sound. It does not need to render completely anything happening beyond my field of observation. In more general terms, anything outside my reference frame doesn't need to be fully rendered.
Another point I'd like to raise is when I look at Pfhorrest summary:
Quoting Pfhorrest
The simulation doesn't have to have a history of actual 14 billion years and it certainly doesn't have to provide qualities for that history that couldn't be observed at the time or afterwards. So you're left with light, radiation, gravity and such and the rest can be ignored. Second, the world simulating ours doesn't have to answer to our physical laws and as such claims about the behaviour of time or the ability of computers in such a world or any other physical law is basically guesswork.
In the final analysis though the question is moot. We cannot talk about the really real (eg. the non-simulated world), only the real (eg. a simulated world if it were true). We will always end up with a contradiction. Either we live in a real non-simulated world and a cup means a non-simulated cup, or we live in a simulated world and cup means a simulated cup. Depending on the world we live in, we mean different things. To see what I mean, go to 0:52 in this clip:
It doesn't actually assume hardcore causality since the uncertainty of a probability based world requires even more processing power. The probability based quantum world of ours is actually a very good argument against us living in a simulation since a probability based world where almost infinite number of possible quantum states affect the next possible quantum states and their probability distribution is almost infinitely complex to calculate. Classical mechanics require almost infinitely less processing power. Uncertainty doesn't remove causality until the wave function collapses. Before that all the possible quantum states causally determine the next probability distribution. The only non causal thing in quantum mechanics I'm aware of is the new information generated by the collapse of the wave function since it can't be predicted.
If you mean that you could create the illusion of probabilities by pretty much making a predetermined movie as our simulation, where none of the probabilities are actually calculated and are actually just predetermined, that could work to reduce the amount of processing needed to give us a simulation of our experiences(Although I'm quite dubious about this since I have no idea how to make a predermined set of experiences and what kind and how much of processing such would actually require). But you will still have to demonstrate that these "predetermined" worlds are hugely represented in simulations in order to show that our current processes are more likely in a simulation than in the "real" world.
But I do agree that this is a good start to show it possible that you can technically have more simulations of our experiences even without as many processes as in the "real" world. I don't think I ever disagreed with this. I even said in the original argument that one only needs to demonstrate that there are more simulations of our experiences in simulations with fewer processes than in the real world. You have only started to demonstrate it possible, but most possible simulations can still be something else than efficiently processed versions of our experiences, making it also possible that our experiences still happen more frequently in the processes of the "real" world than in the processes of a simulation.
Quoting Benkei
Already addressed this point. Try to come up with a quality you can both demonstrate to exist and that doesn't affect our observations in any way. I personally believe in a somewhat solipsistic world anyway where pretty much only the information that constitutes my observations exists for me for certain logically necessary reasons. But this information pretty much contains pretty much everything everyone else believes to exist too since pretty much everything observable for others ends up affecting my observations at least indirectly. Like a tree falling when I'm not seeing it still affects things like gravity in tiny ways affecting my observation.
So, while I believe in a world seemingly like something that is according to you, easier to calculate, I disagree with you since I think this is that is the way the "real" world is and because probability based worlds are harder to calculate. (Unless you demonstrate a way to create predetermined experiences with the illusion of probabilities. Usually a system that can process a complex simulation is always less complex physically than a system that can store all the occurences of a complex simulation. Computer memory is physically, although not technologically, more complex than the computer processor. Any reasonable amount of memory is physically bigger and contains more physical processes than a processor.)
I guess if one compares a non-solipsistic probability based world and a solipsistic probability based world, the logically possible non-solipsistic probability based worlds are more harder to calculate on average. But even then most of all processes happen in non-simulated worlds like the original argument demonstrates. You will still have to demonstrate that solipsistic worlds are hugely more represented in simulations in order to show that our current processes are more likely in a simulation than in the "real" world.
I already did! Weight is irrelevant when I'm not holding it. So it doesn't need to render that quality at that time. Just like a computer that doesn't render anything beyond a horizon in an fps. It's the same concept.
Quoting Qmeri
If there's no such thing that doesn't affect anything else then you're proposing hardcore causality. It comes with it by definition.
Quoting Qmeri
Classical mechanics is good enough most of the time so a simulation would use that and relativity to describe the macro world most of the time. QM is but one of the examples I gave, that while more complex, it doesn't require us to introduce every instance that went before as input as long as outcomes stay within the probability. So when there's epistemic probability we can forego a lot of history.
But you seem to missing the point. The point is, a simulation does not require to simulate every instance fully and certainly not the entire natural history. A processor doesn't need to simulate itself, it's simulated by whatever is outside and it only needs to do that when we're directly observing the processor. The outcome of the processor would be caused by the simulator not the processor. Only a direct observation would have to simulate the world as if the simulated processor caused the simulated outcomes.
If I close my eyes the simulation can stop rendering colour, light and dimensions etc. The point is that not all instances in the world need to be simulated and certainly not every quality at all times.
EDIT: another example, the movement of planets can be described accurately without knowing how they were formed, what they consist of, how many meteors struck it and whether it supports life or not. I can simulate the movement of planets, without having simulated the planet's history.
So, a simulation is a limited activity of some consciousness/awareness which ultimately must belong to a greater consciousness/awareness because nothing can exist outside of consciousness/awareness. Therefore, consciousness is perspective (e.g. a simulation, a concept); it is also existence, as well as reality, and is ultimate.
You seem to be thinking of experience as a computer game, where after simulation one has to render that simulation. I disagree. I think experience comes naturally out of our physical processes and needs no extra rendering step and if we simulated a human mind experiencing something it would not need to be rendered in order for that simulated mind to experience the simulation.
Although, if you think of a matrix-like simulation where after simulating, that simulation has to be transferred to a human brain, then yes, that is rendering the simulation to a brain. It's actually a process reality doesn't have to make and increases the amount of processes that kind of simulation requires compared to reality.
Quoting Benkei
Not true. While we may not have the technical expertise to differentiate between a classical and quantum system in macro scale, it can be fundamentally done and is therefore philosophically relevant, since there would be a difference between a "true" and a simulated experience no matter how minute. Even if a single electron changes its place because of quantum mechanics, it changes the mass distribution of a planet changing its course.
Quoting Benkei
I agree with that one does not have to simulate times before and after our experiences therefore reducing the required processes. I already acknowledged this in my last post where I said that a solipsistic system needs less processes than a non-solipsistic system. You still have to demonstrate that A: it reduces the amount of required processes enough to counterweight the enormous advantage reality has in the number of processes. B: solipsistic simulations are so frequent that they outnumber the amount of singular observers in the real world which still has more processes in it than simulations have and can therefore fundamentally contain more observers.
EDIT: A solipsistic simulation of an observer is a more complex system with more processes than a non simulated observer for the same reasons a simulation of a world requires more processes than just the real world.
I'm only superficially familiar with Simulated World theories, and mostly with those of Tegmark's Mathematical Universe and the Matrix movie. But I note that most of those theories seem to assume that the processor is a space-time machine limited by the laws of physics. But what if the processor actually creates Space-Time? I'm referring to Donald Hoffman's concept of space-time as the interface of our reality. In that case, physical objects are symbols (icons on a screen) referring to metaphysical Forms. If so, the hypothetical simulated worlds would be simulations within a simulation.
I don't understand. Why does a simulation processor need to simulate itself?
Quoting TheMadFool
It doesn't. That was just an example to demonstrate that a processor that could simulate a more complex thing than itself in real time would cause absurdity because then it could simulate a more powerful processor than itself which could then simulate a more powerful processor than itself etc... producing an infinitely powerful processor.
So much so, I'm confused by the original post. Truthfully, it's nonsensical.
Yes, the world is a type of simulation.
Where you might be confused is whatever image you apply to to term 'simulation.'
Rather unparsimonious, possibly going the way of regress. (Simulations all the way down?)
Otherwise, such a simulation would be targeted specifically at experiencers, animals, us (kind of self-elevating).
So, the Moon would literally not be there when no one was looking, as it were.
This starts to reek of deus deceptor, dream thought experiments, Zhuangzi's butterfly, perhaps solipsism, the usual.
Moving towards such unverifiable-unfalsifiable hypotheses seems like a difference that makes no difference.
It would renders our's a "toy universe" at the whims of the simulators.