The Simulation Argument

[Bri]

I saw a watered-down reference to this on television, and decided to look it up. This is from a paper that was published a few years ago.

From Are You Living in a Computer Simulation? (http://www.simulation-argument.com/):


ABSTRACT. 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.


According to the FAQ, the probabilities assigned to each of the three possible propositions is personal opinion given the lack of evidence for any of them. The author of the paper apparently gives slightly more probability to proposition (2) than the others, and therefore believes there is about a 20% probability that we are living in a computer simulation.

I was wondering what others make of this argument and what probabilities you would place on each of the propositions being true?

-Bri

[kellyb]

I think it's...silly.

[Henners]

The bit that I like is this...

If you thought that you were living in a simulation, but could neither prove not disprove it, one of the things that you might do is to develop a computer model of a simulation that contained intelligent agents whose role is to identify whether they are inhabiting a simulation. Of course the simulation would need to be fairly complex, or it would be too easy for the agents to tell that they were in a simulation.

So we're all intelligent agents in a computer simulation.

Here's the rub:

If we find a way to identify that we are a simulation, that means that the "people" who are running that simulation can then identify that they are a simulation, and so on, all the way up.

We will therefore have outlived our usefulness, and the plug will be pulled.

[arthwollipot]

Wow. I've got the most amazing sense of deja vu reading that. I'm certain that I've seen those words before. It was in a thread on (I think) the JREF. It could have been one of my other forums though - I'll have to check that, but I'm pretty sure it was here. That poster was posting as though it was all his (her, its) idea.

You've got me going now. You've got me so I can't sleep at night.

[Henners]

Wow. I've got the most amazing sense of deja vu reading that. I'm certain that I've seen those words before. It was in a thread on (I think) the JREF. It could have been one of my other forums though - I'll have to check that, but I'm pretty sure it was here. That poster was posting as though it was all his (her, its) idea.

You've got me going now. You've got me so I can't sleep at night.

If you ever come up with an original idea, wollipot, get it patented so we can all see how clever you are.

http://www.newscientist.com/article/mg19325901.400.html

[arthwollipot]

No, no! I'm sure I was arguing with someone about this very idea. I can't find it on JREF or CreationTalk, so I'm starting to think it was all just deja vu on my part.

Damn, that's weird.

[Henners]

Either way, unless it happened about ten to fifteen years ago, it's unlikely to have been original.

[JoeEllison]

I had the original thought.... dammit, why didn't I have a pen handy!

[Stimpson J. Cat]

I remember seeing this before as well. In fact, I distinctly remember being involved in the discussion. At that time the claim being made was a bit stronger. The poster argued that there would be an infinite number of such simulations (due to each simulation which progressed far enough containing simulations of its own). The conclusion being that we are almost certainly in a simulation.

I also have seen presented the idea that the universe is a simulation being run by god in an attempt to try to figure out where he came from. The idea being that eventually physical beings would evolve into gods capable of creating such simulations themselves, and so on. Essentially the same idea, but with more theistic overtones.

Dr. Stupid

[arthwollipot]

I remember seeing this before as well. In fact, I distinctly remember being involved in the discussion. At that time the claim being made was a bit stronger. The poster argued that there would be an infinite number of such simulations (due to each simulation which progressed far enough containing simulations of its own). The conclusion being that we are almost certainly in a simulation.

Yes, that's the one. And no, it wasn't that long ago, so of course it wasn't an original idea. The tone of the thread led me to believe that the person making the claim was the originator of the idea.

I can't find the thread in my subscriptions list though.

[Henners]

Don't you realise how dangerous this is?

Listen chaps, you are going to have to stop this discussion ASAP.

The world will end if you come up with the answer.

[PixyMisa]

It's turtles all the way down.

See also Greg Egan's Permutation City.

And yeah, the idea came up here previously.

[Thabiguy]

I was wondering what others make of this argument and what probabilities you would place on each of the propositions being true?

I find the argument logically flawed. Points 1 and 2 deal with our universe only, while point 3, "we are almost certainly living in a computer simulation," makes implications about existence of something outside our universe. In their argument, they try to mask this by silently assuming the simulating universe to be just like ours, but this is just an illusion: even if the simulating universe is just like ours, it is not part of our universe, by definition, and point 3 thus still makes implications about something outside our universe. The word "simulation" has no meaning within the boundaries of the universe in question.

There is also what I consider a grave logical error in their FAQ:
9. Isn’t the simulation-hypothesis untestable?

There are clearly possible observations that would show that we are in a simulation. For example, the simulators could make a “window” pop up in front of you with the text “YOU ARE LIVING IN A COMPUTER SIMULATION. CLICK HERE FOR MORE INFORMATION.” Or they could uplift you into their level of reality.
I disagree with this. Even if such observation is made, it would not show that our universe is simulated. Specifically, you could not tell (inside the universe!) the difference between a simulated universe and a non-simulated universe, in which the appearance of such a window is a part of the laws of nature.

Some might argue, "but it is obvious that if a window like that appears, our universe is simulated". But why would you consider that obvious? Because you would find such observation unusual. Because you have always lived in a universe where random things like that don't happen. Because you silently assume that "non-simulated" automatically implies "governed by a fixed set of rules with no ludicrous exceptions". But this does not logically follow. Why would you think that this needs to be so? Because you have observed our universe to behave like that, and because you subconsciously associate our universe with being non-simulated. It is therefore an understandable fallacy, but a logical fallacy nonetheless.

If you lived in a universe where pop-up windows with taunting inscriptions appear every now and then, you would certainly not consider the appearance of the aforementioned window to be any evidence that the universe is simulated. Even in our kind of universe, the appearance of such a window would have countless possible explanations other than our universe being simulated: our universe could actually contain god(s), demon(s) or other supernatural being(s) playing jokes on you; magic actually works and someone has cast a mischievous spell; the universe, in spite of all we have observed so far, actually behaves unpredictably every now and then; etc. - all implying that the universe is real and non-simulated.

The fact remains: no observation within the universe, no matter how exotic and unusual, can lead to any conclusion whether the universe is simulated or not. The very definition of "simulated universe" implies the existence of something outside that universe, and when constrained to the universe only, loses all meaning.

... And as for simulations of universe in general, I have recently come across this great essay (http://www.disenchanted.com/dis/technology/strange-attractor.html) on the topic. You may want to check it out.

[INRM]

If this was hypothetically true,

How many humans would have to realize they were in a simulation before the plug was pulled... one, two, a thousand, a million, all?


INRM


The bit that I like is this...

If you thought that you were living in a simulation, but could neither prove not disprove it, one of the things that you might do is to develop a computer model of a simulation that contained intelligent agents whose role is to identify whether they are inhabiting a simulation. Of course the simulation would need to be fairly complex, or it would be too easy for the agents to tell that they were in a simulation.

So we're all intelligent agents in a computer simulation.

Here's the rub:

If we find a way to identify that we are a simulation, that means that the "people" who are running that simulation can then identify that they are a simulation, and so on, all the way up.

We will therefore have outlived our usefulness, and the plug will be pulled.

[Henners]

If this was hypothetically true,

How many humans would have to realize they were in a simulation before the plug was pulled... one, two, a thousand, a million, all?


INRM


Well, since we're only here to deliver up a way of spotting that we're in a simulation, it's just the one that would need to solve it. However, it is entirely possible to deliver up a clue that would allow the "people" running the simulation to finish solving it on their own, it could easily be less than one.

Of course, I have part of the answer.

I can tell that we are either not in a simulation, or we are in a top-level one.

We are definitely not in a simulation of a simulation.

[PixyMisa]

I can tell that we are either not in a simulation, or we are in a top-level one.

We are definitely not in a simulation of a simulation.
How so?

[Paul C. Anagnostopoulos]

If we are a simulation, and have never experienced being "real," how are we going to tell we're in a simulation? Is there some little piece of physics that will be the smoking gun?

Hey Joe! I figured out we're in a simulation because the speed of light is off by .00001. No wait, that was just something I dreamed. We're in a simulation because the rules of logic are wrong.

~~ Paul

[Henners]

I already posted a link to a New Scientist piece that explained the speed of light, and a few other things, based upon reality being a simulation. You should have a look at it, Paul.

My reasoning for the level of simulation that we are in is this. An intelligence in the real reality would not make their simulation any more complex than it needed to be. It would be minimally complex.

That top-level simulation could not then itself make a convincing simulation that is any less complex than the already minimally complex model that was their own reality.

So a second-level simulation could not be convincing enough to its inhabiting agents, and would be unfit for purpose because it would be throwing up spurious solutions on a regular basis.

[Myriad]

Assuming that a simulation would have a characteristic clock rate (this is a big assumption but it's true of most simulations we create), an object or particle that's spinning at a sufficient angular velocity would be indistinguishable from a stationary one.

Also re the infinite nested hieararchy of simulations: each simulation necessarily has to be smaller (contain fewer elements) and run more slowly than the parent simulation. At any given moment, there is likely to be an arbitrarily large number of layers of simulations, of which only one (the "bottom" layer) can be in a state where it has not yet developed a sub-simulation of its own. All the layers except the bottom must contain beings capable of developing a sub-simulation.

So, the chance of any given sentient beings like ourselves being in the bottom layer is vanishingly small. And yet, if there were an infinite (or merely very large) hierarchy of simulations, that's where we must be.

Therefore, the existence of a hieararchy of simulations is extremely unlikely.

Respectfully,
Myriad

[Cuddles]

My reasoning for the level of simulation that we are in is this. An intelligence in the real reality would not make their simulation any more complex than it needed to be. It would be minimally complex.

That top-level simulation could not then itself make a convincing simulation that is any less complex than the already minimally complex model that was their own reality.

So a second-level simulation could not be convincing enough to its inhabiting agents, and would be unfit for purpose because it would be throwing up spurious solutions on a regular basis.

That's pure assumption. If you're going to assume that someone can make a simulation of an entire universe, why assume they wouldn't do it properly? It might be simplified, but unless you can prove that that is the only possible way it could be done, your conclusion is not valid.

[Henners]

That's pure assumption. If you're going to assume that someone can make a simulation of an entire universe, why assume they wouldn't do it properly? It might be simplified, but unless you can prove that that is the only possible way it could be done, your conclusion is not valid.

It might.

Always assuming that you were able to simplify something that was already designed to be as simple as possible.

However, if you found such a method of simplifying the simulation, you can be sure that the individual who set up this universe would then simplify it as well.

There's no point in making it too complex, for it will run too slowly.

[Jimbo07]

Don't we have to at least create a working demonstration of something indistinguishable from our own intelligence before realistically discussing the probability of simulating intelligence?

So far, the only argument in this thread that I really agree with is this


All the layers except the bottom must contain beings capable of developing a sub-simulation.

So, the chance of any given sentient beings like ourselves being in the bottom layer is vanishingly small. And yet, if there were an infinite (or merely very large) hierarchy of simulations, that's where we must be.

Therefore, the existence of a hieararchy of simulations is extremely unlikely.


With the caveat that I'm not sure if the word extremely is valid, I do think it is unlikely. Considering my above statement, I'm not sure on a realistic evaluation of probability.

[Thabiguy]

My reasoning for the level of simulation that we are in is this. An intelligence in the real reality would not make their simulation any more complex than it needed to be. It would be minimally complex.

That top-level simulation could not then itself make a convincing simulation that is any less complex than the already minimally complex model that was their own reality.

So a second-level simulation could not be convincing enough to its inhabiting agents, and would be unfit for purpose because it would be throwing up spurious solutions on a regular basis.

Following your assumptions, by realizing that they are unable to construct a second-level simulation that would be convincing enough to its inhabitants, the inhabitants of the top-level could also discover that they are in a simulation. Therefore, the top-level simulation is also not convincing enough to its inhabitants. This means that the top-level simulation is not as complex as it needs to be, which contradicts your first assumption.

[Paul C. Anagnostopoulos]

Assuming that a simulation would have a characteristic clock rate (this is a big assumption but it's true of most simulations we create), an object or particle that's spinning at a sufficient angular velocity would be indistinguishable from a stationary one.
But wouldn't this also be the case in a real universe with a Planck time?

~~ Paul

[Cuddles]

It might.

Always assuming that you were able to simplify something that was already designed to be as simple as possible.

However, if you found such a method of simplifying the simulation, you can be sure that the individual who set up this universe would then simplify it as well.

There's no point in making it too complex, for it will run too slowly.

Having thought about this, you actually have it exactly backwards. You don't make a simulation as simple as possible, you make it as complex as possible within the limits of available computing power. The only way to simulate something exactly is to have an exact copy. Any simplification at all will cause inaccuracies.

If you could find a method of simplifying a model that would give identical results to a complete model, then yes, that would be done. However, your assumption that the simplest model possible would be used is just wrong. The simplest possible model that gives the most accurate possibe result is what would be used, which is identical to the most complex method possible.

[Henners]

Assuming that a simulation would have a characteristic clock rate (this is a big assumption but it's true of most simulations we create), an object or particle that's spinning at a sufficient angular velocity would be indistinguishable from a stationary one.

Also re the infinite nested hieararchy of simulations: each simulation necessarily has to be smaller (contain fewer elements) and run more slowly than the parent simulation. At any given moment, there is likely to be an arbitrarily large number of layers of simulations, of which only one (the "bottom" layer) can be in a state where it has not yet developed a sub-simulation of its own. All the layers except the bottom must contain beings capable of developing a sub-simulation.

So, the chance of any given sentient beings like ourselves being in the bottom layer is vanishingly small. And yet, if there were an infinite (or merely very large) hierarchy of simulations, that's where we must be.

Therefore, the existence of a hieararchy of simulations is extremely unlikely.

Respectfully,
Myriad


I see what you are saying, Myriad.

However, it is possible for a simulation to create a sub-simulation with fewer components than itself. The clock tick issue is not really what matters, what matters is the recalculation time for the system - and as this happens in between the simulation-level clock ticks, the simulation cannot identify how fast the clock is ticking in its parent(s).

...and in its descendants, the clock ticks are as long as is required to recalculate the system. In the simulation, the clock might even appear to be smooth.

I think that it is reasonable for the top level reality to be able to look into any simulation in the hierarchy, and to expect it to incorporate all discovered improvements directly into the top-level simulation.

[Henners]

The simplest possible model that gives the most accurate possibe result is what would be used

Oddly, that's what I thought I had said.

I also said that any simpler a model would continuously generate false positives and would therefore be useless.

I also said that the reality-level might allow the top level to generate a simulation of its own in order to gain access to any new technical simulation advances that could be incorporated into the top-level simulation.

However the second-level simulation would not be a viable simulation for the minimum viable simulation already exists as the top-level simulation.

I believe that we inhabit a viable simulation, so it is either a top-level simulation or it is real reality.

[Thabiguy]

But wouldn't this also be the case in a real universe with a Planck time?

Of course not; in any universe with the same laws of physics as our universe (simulated or not), increasing an object's spin rate means increasing its energy and thus also its gravitational influence on other objects. Not to mention that rotation of a solid object only works because of tensile strength of the material that does not allow its constituents to continue travelling in the directions of their (opposite) velocities and curves their paths into circles. When this fails to work properly (either because the velocities overcome the tensile strength or because of simulation inaccuracies), the constituent parts of the object just fly apart.

[Henners]

Following your assumptions, by realizing that they are unable to construct a second-level simulation that would be convincing enough to its inhabitants, the inhabitants of the top-level could also discover that they are in a simulation. Therefore, the top-level simulation is also not convincing enough to its inhabitants. This means that the top-level simulation is not as complex as it needs to be, which contradicts your first assumption.


This always assumes that the individuals at the real reality level are correct in their conjecture that their simulation is capable of giving them a way to figure out whether they are simulation.

But, as the people at real reality are real, and not a simulation, it is impossible for the simulation to ever give them a way of knowing that they are real.

[Cuddles]

Oddly, that's what I thought I had said.

I also said that any simpler a model would continuously generate false positives and would therefore be useless.

I also said that the reality-level might allow the top level to generate a simulation of its own in order to gain access to any new technical simulation advances that could be incorporated into the top-level simulation.

However the second-level simulation would not be a viable simulation for the minimum viable simulation already exists as the top-level simulation.

I believe that we inhabit a viable simulation, so it is either a top-level simulation or it is real reality.

But apparently you missed this part:
which is identical to the most complex method possible.

You don't appear to understand what simplification means. If you simplify the model, it is wrong. Depending on the simplification, the results might still be close enough to reality for your purpose, but it will not actually be the same. What you are saying is that someone would make a model of the universe that is so simplified that it would not be possible for anything in the simulation to simplify it any further. If that was the case, the simulation would be a very long way from being accurate, becaues there would be things possible in the real world that are not possible in the simulation. That means that if there is any way to improve the simulation, it will be done, which means that the model will become more complex.

The simplest model that gives the most accurate results possible is exactly the same as the most complex model that is possible with the available resources. It is not the simplest model that it is possible to ever make, which is what you are assuming. The simplest model that it is possible to make is horribly inaccurate and would never be used if there was an option to improve it.

[Bri]

Also re the infinite nested hieararchy of simulations: each simulation necessarily has to be smaller (contain fewer elements) and run more slowly than the parent simulation.

That's pure assumption. If you're going to assume that someone can make a simulation of an entire universe, why assume they wouldn't do it properly? It might be simplified, but unless you can prove that that is the only possible way it could be done, your conclusion is not valid.

My initial thought was that it would be impossible for us to simulate our universe inside of a computer because any simulation must be simpler than the universe it resides within (that it would take at least as much data to simulate an atom as the atom itself). But I don't know if that is an accurate assumption. Is it an accurate assumption? If not, might we be able to simulate an even more complex universe than our own?

If it is an accurate assumption then proposition (2) is almost certainly the case -- unless we are living in a simplified simulation of a more complex universe.

That top-level simulation could not then itself make a convincing simulation that is any less complex than the already minimally complex model that was their own reality.

So a second-level simulation could not be convincing enough to its inhabiting agents, and would be unfit for purpose because it would be throwing up spurious solutions on a regular basis.

That is based on an assumption that the people who created the simulation have the technology and the desire to accurately simulate their own world rather than a simplified version of it. It is also based on an assumption about the reason they would want to create the simulation to begin with (to try to find out if they are in a simulation) which seems like an unlikely reason to create a simulation given that they themselves might have been created for the same purpose and being more complex than their simulation should be better equipped to do it than their simulated people would be.

One of the reasons we make simulations is to create models that eliminate factors we're not interested in looking at, so we purposely create simulations that are unrealistic -- specifically simulations that are simpler than reality. If our programmers were only interested in simulating intelligence or consciousness, for example, we might very well be in a simulation that is simpler than the universe our simulation exists in.

In that case (if we are a simulation of a more complex universe) then there is probably a limit to the amount of nesting that could take place and still be complex enough to simulate consciousness.

There is also what I consider a grave logical error in their FAQ:

I disagree with this. Even if such observation is made, it would not show that our universe is simulated. Specifically, you could not tell (inside the universe!) the difference between a simulated universe and a non-simulated universe, in which the appearance of such a window is a part of the laws of nature.


While I agree that it would be impossible to tell for certain that we are in a simulation based on this evidence, it seems that it would be more likely than other explanations given that a god or other supernatural being would have little incentive (other than to screw with us) to pop up a window telling us we're in a simulation. It is also possible that the window is the result of natural forces, but again that seems unlikely given those events don't happen all the time (and if they did, the programmers would presumably use a different method of letting us know we are in a simulation that would be an event that would not occur regularly in our universe).

Additionally, since the programmers of the simulation have the technology to create the simulation, it is possible that they also possess the ability to re-program our brains in such a way that we know/believe that we are in a simulation if they wanted us to know we were in a simulation. I suppose that also doesn't preclude the god idea since a god could potentially do the same thing to screw with us -- but then you're getting into the possibility that nothing we know at all is true since a god could simply be manipulating our brains so that we "know" what it wants us to know.

-Bri

[Soapy Sam]

What if the plug was pulled centuries ago, but it turned out it wasn't actually wired to anything?

Maybe we have it backwards.

It's turtles all the way up.

[Henners]

The simplest model that gives the most accurate results possible is exactly the same as the most complex model that is possible with the available resources.

I see now what you are saying, Cuddles.

But models aren't made to return the most accurate results possible. Models are made to return sufficiently accurate results.

The only times and places where models are made to be as accurate as possible is in cases where it is known from the outset that the model cannot be made complex enough using existing technologies.

If that is the case here, I think that we can be reasonably certain that we are inhabiting the real reality, otherwise we would not be in a very good simulation and would be coming up with valid reasons why we are inhabiting a simulation on a regular basis.

(I am, possibly foolishly, assuming that these events are not being edited out.)

[Paul C. Anagnostopoulos]

Of course not; in any universe with the same laws of physics as our universe (simulated or not), increasing an object's spin rate means increasing its energy and thus also its gravitational influence on other objects. Not to mention that rotation of a solid object only works because of tensile strength of the material that does not allow its constituents to continue travelling in the directions of their (opposite) velocities and curves their paths into circles. When this fails to work properly (either because the velocities overcome the tensile strength or because of simulation inaccuracies), the constituent parts of the object just fly apart.
Well, it may be an "of course not" to you, but you'll need to explain it a bit more to me. Doesn't the Planck length discretize space so that one cannot discern movements shorter than that length?

~~ Paul

[Paul C. Anagnostopoulos]

The only times and places where models are made to be as accurate as possible is in cases where it is known from the outset that the model cannot be made complex enough using existing technologies.
But what if the folks running the simulation want to keep us from figuring it out for as long as possible?

~~ Paul

[Henners]

Hi Bri,
In that case (if we are a simulation of a more complex universe) then there is probably a limit to the amount of nesting that could take place and still be complex enough to simulate consciousness.

Nesting is not relevant at all.

If you are in a simulation, your smallest clock tick (Planck time?) defines the intervals between each recalculation of your reality.

In your parent universe, generations could live and die between your clock ticks. Time would still appear to tick by in the simulation.

If real reality is a finite universe with a certain number of particles in it, it is obviously impossible to replicate that entire universe within a simulation.

However, take a look around the room that you are in, and then look back at the screen. Do you really think that the room is now fixed in your mind, or do you just think that you think that?

[Henners]

But what if the folks running the simulation want to keep us from figuring it out for as long as possible?

~~ Paul

Sorry, Paul. I thought I had made it clear that the purpose of running the simulation is to find a way of answering the question, "How can I find out if I'm inhabiting a simulation."

If you want to speculate that it's a simulation competition and that we are just one of many simulations waiting to get knocked out as soon as we know that we are simulated, that's fine, too.

[Jimbo07]

If you want to speculate that it's a simulation competition and that we are just one of many simulations...

"Go Our Simulation!"

"Yay Team!"

:D

[Paul C. Anagnostopoulos]

Sorry, Paul. I thought I had made it clear that the purpose of running the simulation is to find a way of answering the question, "How can I find out if I'm inhabiting a simulation."
Oops, yes, indeed you did. But the folks running the simulation would want to make the simulation as accurate as possible in order to the give the virtual inhabitants a real job. Otherwise the inhabitants might come up with evidence of simulation that had nothing to do with the simulators' situation, due to shortcuts in the simulation.

Kind of like when CGI goes bad and there is a hole in the cow away from the camera.

~~ Paul

[PixyMisa]

Henners - ah, that's why your posts didn't seem to make sense. It wasn't clear that they were all tied back to your first post. While it's a plausible reason for running the simulation, it's only one of many, and not one that most of us are focused on.

[Naughtyhippo]

Won't this whole line of thinking boil down to 'Am I real? It's all very well saying I think therefore I am but I do I prove that I'm thinking? that these are my thoughts?' and other things guareented make me have a sudden craving for chocolate icecream and an escapist movie.

ETA: Any wandering grammar police, should I have a question mark at the end of my convoluted sentence above? As for the spelling police, well, you know that I know that I should really be doing hard labour somewhere....

[PixyMisa]

If real reality is a finite universe with a certain number of particles in it, it is obviously impossible to replicate that entire universe within a simulation.
Of course, it's perfectly possible that the real universe is infinite or transfinite. For suitable definitions of "possible" and "real"...

[PixyMisa]

Won't this whole line of thinking boil down to 'Am I real? It's all very well saying I think therefore I am but I do I prove that I'm thinking? that these are my thoughts?' and other things guareented make me have a sudden craving for chocolate icecream and an escapist movie.
Descartes answered that one pretty well, I think.

[Henners]

Descartes answered that one pretty well, I think.

However, you could easily be programmed to think that.

[Naughtyhippo]

However, you could easily be programmed to think that.

that's what i was trying to say. how can i tell that it is me thinking? apologies for lack of caps, only have one hand free.

[Thabiguy]

While I agree that it would be impossible to tell for certain that we are in a simulation based on this evidence, it seems that it would be more likely than other explanations given that a god or other supernatural being would have little incentive (other than to screw with us) to pop up a window telling us we're in a simulation.
Are you speaking from actual experience with supernatural beings? :)
It is also possible that the window is the result of natural forces, but again that seems unlikely given those events don't happen all the time (and if they did, the programmers would presumably use a different method of letting us know we are in a simulation that would be an event that would not occur regularly in our universe).
This illustrates the essence of what I was saying. Assuming that an unusual event (where "unusual" can be tailored for the particular universe) would be evidence that the universe is simulated, is logically equivalent to assuming that non-simulated universes would not contain unusual events. This assumption, however, does not logically follow. There is no basis for it other than saying, "it seems likely, because I've seen our universe behave that way, and it seems (based on my gut feeling) that if a universe is not simulated, it should be something like our universe". Do you see the problem with that reasoning? Think about it.

Additionally, since the programmers of the simulation have the technology to create the simulation, it is possible that they also possess the ability to re-program our brains in such a way that we know/believe that we are in a simulation if they wanted us to know we were in a simulation. I suppose that also doesn't preclude the god idea since a god could potentially do the same thing to screw with us -- but then you're getting into the possibility that nothing we know at all is true since a god could simply be manipulating our brains so that we "know" what it wants us to know.
A god - or a programmer of the simulation? How do you distinguish a god manipulating a real universe from a programmer manipulating a simulation?

I'm not trying to imply that nothing can be known for sure, and so we should abandon our futile endeavors. What I'm trying to say is that it cannot be determined whether a universe is simulated or not from within the universe. Not because there would be a law of nature against that, but because this is a logical consequence of the definition of the term.

You may look at it this way: "Universe A is simulated" is not a statement about universe A. It doesn't make any claims or predictions about universe A; it says nothing about it. It is in fact a statement about an implied universe B: "someone (in universe B) is performing a simulation of universe A". Universe A is just a universe, described by its contents and rules. Whether it is simulated or not, is a question about universe B, therefore meaningless within universe A (which doesn't contain B).

[Cuddles]

Sorry, Paul. I thought I had made it clear that the purpose of running the simulation is to find a way of answering the question, "How can I find out if I'm inhabiting a simulation."

If this is the reason for doing the simulation, your whole argument doesn't make sense. If the simulation is so simple that people inside it can't make their own simulation, then the simulation doesn't describe reality and is useless. If it does simulate reality well enough for them to make their own simulation then we can't tell that we are in a top level simulation because it is possible to make more simulations within the simulation. Either your logic is valid, which means that your assumption about the simplicity of the simulation is wrong, or your assumption is correct which means the logic is invalid.

Is there a prize for most uses of the word "simulation" in one paragraph?

[PixyMisa]

However, you could easily be programmed to think that.
Descartes answered that one pretty well, I think.

[Thabiguy]

Well, it may be an "of course not" to you, but you'll need to explain it a bit more to me. Doesn't the Planck length discretize space so that one cannot discern movements shorter than that length?
Okay, let's see...

Imagine a simplified solid object: two balls connected together with a rigid rod. If this object is rotating (along the obvious axis), it means that in a chosen reference frame (any frame really, but let's take the frame of the object's mass center for simplicity) the velocities of the balls differ. One of them tries to go right, the other tries to go left. But because they are connected, the balls can't fly apart. Instead, the binding forces within the rod pull on them, curving their trajectory so that they inscribe a circle and return to their "original positions".

Now as you increase the spin rate of the object, you increase the kinetic energy of the balls. It's really two sides of the same coin: rotational kinetic energy of a spinning object and the sum of kinetic energies of its constituents are just two ways of talking about the same thing. When you keep increasing the kinetic energy of the balls, of course eventually it will overcome the (negative) binding energy of the connecting rod and the system will become unbound - the rod will snap and the balls will fly apart.

The argument was, in essence - if we take a superrigid rod and make the balls spin fast enough, won't it be indistinguishable from an object that doesn't spin at all? As in, the balls move so quickly that the time in which they "go around" is under the "resolution of the universe"? And the answer is, if the balls' velocity is so high that the rotation period would somehow be under the resolution of the universe, then the even finer-graded 360deg change in their velocities, which needs to happen along their track to allow them to return to their original position, is by the same logic also under the resolution of the universe, the universe will be unable to perform it, and the balls will just fly apart with their large velocities.

[Bri]

Are you speaking from actual experience with supernatural beings? :)

Apparently there would be no way for me to know even if I did, right?

This illustrates the essence of what I was saying. Assuming that an unusual event (where "unusual" can be tailored for the particular universe) would be evidence that the universe is simulated, is logically equivalent to assuming that non-simulated universes would not contain unusual events. This assumption, however, does not logically follow. There is no basis for it other than saying, "it seems likely, because I've seen our universe behave that way, and it seems (based on my gut feeling) that if a universe is not simulated, it should be something like our universe". Do you see the problem with that reasoning? Think about it.

However, it seems as though if our programmers wanted to let us know that we were simulated, windows popping up with messages (possibly messages that answered our questions) would be evidence that we are in a simulation. Sure, there would be other possible explanations (including the possibility that the windows are a completely natural occurrence in our universe) but that doesn't seem to be the most likely explanation.

A god - or a programmer of the simulation? How do you distinguish a god manipulating a real universe from a programmer manipulating a simulation?

Only one would likely provide us an unambiguous message that we are in a simulation. It is possible that a god would provide the same message if they wanted to screw with us, but that doesn't seem the most likely explanation.

I'm not trying to imply that nothing can be known for sure, and so we should abandon our futile endeavors. What I'm trying to say is that it cannot be determined whether a universe is simulated or not from within the universe. Not because there would be a law of nature against that, but because this is a logical consequence of the definition of the term.

Sure, we couldn't determine it for certain, but wouldn't that be the most reasonable explanation if windows were to pop up telling us so and answering all of our questions?

You may look at it this way: "Universe A is simulated" is not a statement about universe A. It doesn't make any claims or predictions about universe A; it says nothing about it. It is in fact a statement about an implied universe B: "someone (in universe B) is performing a simulation of universe A". Universe A is just a universe, described by its contents and rules. Whether it is simulated or not, is a question about universe B, therefore meaningless within universe A (which doesn't contain B).

Universe B contains universe A, right? So are you saying that it would be impossible for inhabitants of universe A to be provided any evidence whatsoever that they are also members of a larger universe B? I would think a violation of the usual rules of universe A (particularly a coherent conversation with an inhabitant of universe B) would indeed be evidence, although not definitive, that universe A is contained within another universe.

-Bri

[Paul C. Anagnostopoulos]

The argument was, in essence - if we take a superrigid rod and make the balls spin fast enough, won't it be indistinguishable from an object that doesn't spin at all? As in, the balls move so quickly that the time in which they "go around" is under the "resolution of the universe"? And the answer is, if the balls' velocity is so high that the rotation period would somehow be under the resolution of the universe, then the even finer-graded 360deg change in their velocities, which needs to happen along their track to allow them to return to their original position, is by the same logic also under the resolution of the universe, the universe will be unable to perform it, and the balls will just fly apart with their large velocities.
Got it. Thanks!

~~ Paul

[malaka]

I haven't read the entire thread, but has anyone yet mentioned The Thirteenth Floor (http://imdb.com/title/tt0139809/)?

[andyandy]

Wow. I've got the most amazing sense of deja vu reading that. I'm certain that I've seen those words before. It was in a thread on (I think) the JREF. It could have been one of my other forums though - I'll have to check that, but I'm pretty sure it was here. That poster was posting as though it was all his (her, its) idea.

You've got me going now. You've got me so I can't sleep at night.

i presume you're referring to this thread

http://forums.randi.org/showthread.php?t=87370

it was discussed in depth - i introduced it (post 14).....but i think you'll find references to Bostrom on every page :)

this is a form of Nick Bostrom's computer simulation argument - the paper is http://www.simulation-argument.com/simulation.html

[Wowbagger]

I think that what we know of about thermodynamics, and quantum physics, would place a severe limit on the number of simulations-within-simulations there could possibly be, in the Universe.

At least that is my understanding. I will have to try to dig up more about this.

[Thabiguy]

However, it seems as though if our programmers wanted to let us know that we were simulated, windows popping up with messages (possibly messages that answered our questions) would be evidence that we are in a simulation. Sure, there would be other possible explanations (including the possibility that the windows are a completely natural occurrence in our universe) but that doesn't seem to be the most likely explanation.
This is equivalent to saying that it seems likely that a typical non-simulated universe will not contain such windows. How many non-simulated universes do you know to draw such conclusions? I'll propose a counter-hypothesis: our universe is simulated and all non-simulated universes in the metaverse, as a rule, do contain unusual events, such as pop-up windows. If this is the case, a window popping up would actually be evidence that the universe is not simulated! What makes this proposed counter-hypothesis unlikely? Your hypothesis and my hypothesis have exactly the same logical grounds (none) and exactly the same empirical grounds (none, because we have never experienced anything but our own universe and we have no justification for extrapolating our experiences outside our universe). The only difference is in your feeling.

Only one would likely provide us an unambiguous message that we are in a simulation. It is possible that a god would provide the same message if they wanted to screw with us, but that doesn't seem the most likely explanation.
Based on what? What makes the idea of a god screwing with us not likely, but the existence of entities outside our universe who happen to be much like our programmers and decide to make changes to physical rules of the universe to inform us about stuff, likely? There are no logical grounds or empirical reasons to prefer either variant.

Sure, we couldn't determine it for certain, but wouldn't that be the most reasonable explanation if windows were to pop up telling us so and answering all of our questions?
Of course it would be reasonable. But there is no way of knowing that this would be correct. You would have no way of knowing whether you're dealing with programmers of a simulation, or mischievous gods in a real universe - or perhaps deluded gods, dreaming that they are directing a simulation. Assuming the simplest variant is reasonable, because it wouldn't matter to you, as the difference could not ever be told in this universe alone, by definition.

Universe B contains universe A, right? So are you saying that it would be impossible for inhabitants of universe A to be provided any evidence whatsoever that they are also members of a larger universe B?
Correct.
I would think a violation of the usual rules of universe A (particularly a coherent conversation with an inhabitant of universe B) would indeed be evidence, although not definitive, that universe A is contained within another universe.
Rules of universe A cannot be violated within universe A. If the programmers simulate universe A for some time, and then decide to start conversating with the inhabitants of the unverse, they are no longer simulating universe A. They are now simulating some other universe, C, which includes all input from the programmers, because it is necessary to fully describe its contents and rules. Universe C is a universe with "gods" that change physical laws and tell the inhabitants various things.

The "gods", however, and all their manifestations, are still part of the universe C! Inhabitants of universe C have no way of finding out whether C is 1) simulated in universe B, 2) simulated in universe D (where entirely different programmers simulate universe C including all its complex rules of conversation with gods), or 3) C is a non-simulated universe, containing gods who dream they are programmers in another universe.

I understand why you may think that scenario 1 would be "most likely", but the differences between the scenarios are unobservable for the inhabitants of C. Statement "C is a simulated universe", again, says nothing about universe C, because universe C would be universe C whether it were simulated or real; the conversation with gods is an essential part of the definition of universe C. It is, again, a statement about an implied outside universe, such as B or D.

[Beth]

Interesting thread Bri. Personally, I don't think it would be possible to tell a sufficiently rich simulation from reality, but then again, I'm not sure when you are talking about an entire universe what the difference is between 'real' and 'simulated' to inhabitants of that universe.

I have a few comments on some statements:

Also re the infinite nested hieararchy of simulations: each simulation necessarily has to be smaller (contain fewer elements) and run more slowly than the parent simulation.

I don't think it need run more slowly. Simulations are often created that speed things up considerably. Consider simulations of galaxy, planet and star formation. Clearly, we can create simulations that mimic the effect of eons in minutes.

If you simplify the model, it is wrong. Depending on the simplification, the results might still be close enough to reality for your purpose, but it will not actually be the same.

The simplest model that gives the most accurate results possible is exactly the same as the most complex model that is possible with the available resources. It is not the simplest model that it is possible to ever make, which is what you are assuming. The simplest model that it is possible to make is horribly inaccurate and would never be used if there was an option to improve it.

I am reminded of a quote from Dr. George Box "All models are wrong. Some models are useful". At any rate, complexity doesn't necessarily align with accuracy. It's often the case that a simpler model provides more accurate answers. The more complex ones are much harder to program, more prone to errors, harder to verify, etc.

My initial thought was that it would be impossible for us to simulate our universe inside of a computer because any simulation must be simpler than the universe it resides within (that it would take at least as much data to simulate an atom as the atom itself). But I don't know if that is an accurate assumption. Is it an accurate assumption? If not, might we be able to simulate an even more complex universe than our own?


I think it depends on the details of the model and what it is being used for. It can be tremendously complex to model every possible stage and position of even simple things, while simple but carefully chosen rules of behavior can reap incredibly complex results when applied iteratively. Does anyone here remember the computer game of life (I think by Jon Von Neumans?) with cells that lived, died, or were born in each successive interation of the game. Simple rules. You need only decide on the initial placement of live cells in a matrix and it could produce a complex result.

Another example: a simulation of snowfall that specified the position, velocity and unique exact structure of every snowflake individually might be incredibly complex or it might be an iterative application of some relatively simple rules applied to some randomly generated seed values. The first type could be run over and over with exactly the same results, while the second sort would never turn out the same twice. Depending on why you were running the simulation, either one might suit your needs better.

[INRM]

Soapy Sam,

I don't get it... if the plug was pulled wouldn't it shut off? How could it not be connected to something


INRM

[Bri]

This is equivalent to saying that it seems likely that a typical non-simulated universe will not contain such windows. How many non-simulated universes do you know to draw such conclusions? I'll propose a counter-hypothesis: our universe is simulated and all non-simulated universes in the metaverse, as a rule, do contain unusual events, such as pop-up windows. If this is the case, a window popping up would actually be evidence that the universe is not simulated! What makes this proposed counter-hypothesis unlikely?

One of the premises in the paper was that we will develop the ability and desire to implement "ancestor simulations" (that when we develop the technology, we would have reason to simulate our evolutionary history). Premise (3) in the paper was that there is a good chance that we're in an ancestor simulation -- a simulation developed not by some inhabitants of some unknown universe, but developed by inhabitants of a universe that our universe is a simulation of. That means that our universe would be an approximation of the meta-universe, and I think it would be fairly safe to say that it would be highly unlikely that windows pop up randomly in the meta-universe but not in ours.

-Bri

[Bri]

Interesting thread Bri. Personally, I don't think it would be possible to tell a sufficiently rich simulation from reality, but then again, I'm not sure when you are talking about an entire universe what the difference is between 'real' and 'simulated' to inhabitants of that universe.

Yeah, but the thing that is probably most disturbing about the paper is that it's not just a remote possibility that we're in a simulation -- the probability might actually be quite high that we're in a simulation.

-Bri

[Beth]

Yeah, but the thing that is probably most disturbing about the paper is that it's not just a remote possibility that we're in a simulation -- the probability might actually be quite high that we're in a simulation.

-Bri


I'm not sure I agree with that assessment of the probability - it is, after all, based on very subjectively assigned probabilities. But I don't understand why you would consider it disturbing. Even if it were possible to make the determination and it was conclusively proved true, what difference would it make?

[Bri]

Well, for one thing, the programmers might decide to pull the plug at any moment. Not that the universe might not be destroyed by perfectly natural means even if we aren't a simulation, but the probability of our sudden demise seems a lot higher if we are a simulation.

-Bri

[blobru]

...
I was wondering what others make of this argument and what probabilities you would place on each of the propositions being true?

Well, right away we're asked to accept: Substrate-Independence; current projections for increased computing power. That's asking a lot. Technological extrapolation is kinda iffy. Check out any sci-fi from fifty years ago and see how well it matches up.
With the assumptions in place, however, I think the rest follows. There were a couple of passages struck me funny though:

In addition to ancestor-simulations, one may also consider the possibility of more selective simulations that include only a small group of humans or a single individual. The rest of humanity would then be zombies or “shadow-people” – humans simulated only at a level sufficient for the fully simulated people not to notice anything suspicious. It is not clear how much cheaper shadow-people would be to simulate than real people. It is not even obvious that it is possible for an entity to behave indistinguishably from a real human and yet lack conscious experience. Even if there are such selective simulations, you should not think that you are in one of them unless you think they are much more numerous than complete simulations. There would have to be about 100 billion times as many “me-simulations” (simulations of the life of only a single mind) as there are ancestor-simulations in order for most simulated persons to be in me-simulations.

Why if the simulators can create "shadow-people" whose behavior can mimic and fool conscious beings would they bother to include conscious beings in their simulations? What information would they gain from attaching qualia to one's behaviors? These qualia would only affect the simulacrum's 'inner experience', not its behavior, and thus have no impact on the course of the simulation.

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”.

How can we know anything about the simulators' ethos? The author states accepting his sim-arg gives us "a reason to behave ethically", but does it? Are the simulators Presbyterian, Episcopal, Shinto, Hedonist, what? Since for some reason they've supplied us with sensation and then refuse to intervene to curtail pain I'd assume they must be either stoic voyeurs, bored sadists, or both. So maybe only peeping-toms and serial killers get into virtual heaven.

Anyway, all in all some very interesting spec philosophy; thanks for posting. :) (I didn't tackle the probabilities because without knowing the likelihood of those two initial assumptions they don't mean much.)

[Thabiguy]

I apologize beforehand for a very long post.

Well, for one thing, the programmers might decide to pull the plug at any moment. Not that the universe might not be destroyed by perfectly natural means even if we aren't a simulation, but the probability of our sudden demise seems a lot higher if we are a simulation.
Your "demise" would be undetectable for you.

One of the premises in the paper was that we will develop the ability and desire to implement "ancestor simulations" (that when we develop the technology, we would have reason to simulate our evolutionary history). Premise (3) in the paper was that there is a good chance that we're in an ancestor simulation -- a simulation developed not by some inhabitants of some unknown universe, but developed by inhabitants of a universe that our universe is a simulation of. That means that our universe would be an approximation of the meta-universe, and I think it would be fairly safe to say that it would be highly unlikely that windows pop up randomly in the meta-universe but not in ours.
First of all, the paper, even if we accept it as valid, argues that at least one of (1), (2) and (3) is true. It doesn't say that (3) is true; the author says he believes that (2) is true. (3) is not a premise, it is a proposition that is a part of an argument about whether or not it is true depending on the truth value of the other propositions. Specifically, if (1) or (2) is true, the argument makes no claim about whether (3) is true or not.

This has nothing to do with the question 9 in FAQ. This question does not begin, "Assuming that variant (3) is true, ..." The question is quite independent of the argument presented in the paper, and the given answer should therefore hold regardless of whether (3) is true or not.

And, I still find the presented argument itself logically flawed. Specifically, I can't agree with the sleight-of-hand switch of universes during the argument. The argument says that if (1) and (2) are false - if we have a good chance of becoming posthuman and if posthuman civilizations can be expected to run many ancestor-simulations - then we are almost certainly living in a computer simulation. The reasoning is simple and goes like this: Consider your neighbor Joe. If we do indeed evolve to run many ancestor-simulations, then in the future, we will likely run many simulations of Joe. There would be many more simulated instances of Joe than the single real one, so a randomly chosen instance of Joe would almost certainly be a simulated one. Therefore, if we do evolve to run many ancestor-simulations, we should assume that we ourselves are a simulation.

The flaw is this: let's say that we run the simulation of him a million times. How many simulated Joes are there, to choose from? A million, right? Imagine now that the first time the simulation is run, the computer that simulates Joe dumps its memory and state after each step that it takes and stores this in some internal storage. And on subsequent runs, at each step that the computer should take, instead of computing the next state from the previous state, the computer takes advantage of the fact that the result has already been computed and loads the next state from the internal storage, to improve performance. How many Joes are there now - still a million? It should be so, because this is an internal optimization of the computation process, and caching results instead of recomputing them is a common programming technique even today. As long as the results are correct (and they will be), there can be no objections. But then, how can you tell whether Joe has been simulated again or not? Should loading data from internal storage to memory and discarding them again to load next data be considered an independent instance of Joe? Our current DRAMs work by periodically refreshing (re-reading) data even when they're not accessed by the processor. If the internal storage is anything like that, does it mean that there will be not a million, but trillions instances of Joe? Or would just the very first computation of Joe count? How many Joes would there be?

The answer is of course that the number is arbitrary. Simulating Joe means equating an abstract mathematical process - a particular computation of a Turing machine - with Joe. But an abstract mathematical process has no inherent physical existence. One can arbitrarily define what counts as an "instance" of Joe in the simulating universe, therefore arbitrarily define the number of Joes to choose from. But within Joe's universe itself, no instances exist. The Joe that you run today is the same Joe that you have run yesterday. In the simulated universe, there is no difference; any information about "running" Joe, pausing him, restarting him, skipping parts of him, or running him multiple times in parallel, is not only imperceptible, but outright undefined in Joe's universe; it only exists in the simulating universe.

So, after you have arbitrarily picked the number of Joes you want to use, the argument goes: if Joe thinks, "I am simulated", for so many instances of Joe this will be right, and only for one it will be wrong. Therefore, if Joe assumes that he's simulated, he's most likely correct, right? - Well I don't think so. This statement is meaningful and can be verified only in the simulating universe, where we can consider each instance of Joe independently. In Joe's universe, the statement is meaningless. There are not N other Joes; whether Joe is run N times or M times is a concept that lies entirely outside of Joe's universe. But point (3) does make such a switch, from considering the argument from the viewpoint of the "parent" universe, to considering it from the viewpoint of the simulated universe. This, I believe, is a logical flaw, and invalidates the argument.

To illustrate how this could lead to a paradox: let's assume that (1) and (2) are false, i.e. we are likely to evolve to run many ancestor-simulations. Let's also assume that simulating a posthuman civilization is prohibitively expensive; as the paper puts it: "Simulating even a single posthuman civilization might be prohibitively expensive. If so, then we should expect our simulation to be terminated when we are about to become posthuman." According to the argument, we will run many ancestor-simulations, therefore the vast majority of instances of "us" will "be simulated", and so we can expect that this is our case; as the paper puts it: "If we do go on to create our own ancestor-simulations, this would be strong evidence against (1) and (2), and we would therefore have to conclude that we live in a simulation." But, by the same logic, given the aforementioned assumptions, the vast majority of instances of "us" will be terminated when they are about to become posthuman. Therefore, we would also have to conclude that this is most likely our case as well, and we are therefore not likely to get to run many ancestor-simulations. But this leads to contradiction, because we assumed that (1) and (2) are false. - I believe this paradox is resolved by recognizing that logical statements that are true in the simulating universe needn't be true or even valid when considered from the simulated universe. But this, while resolving the paradox, also means that the conclusion that if (1) and (2) are false, (3) must be true, cannot be made.

If you have read it all the way here, congratulations.

[INRM]

What is the definition of "Post Human"

INRM

[Bri]

Your "demise" would be undetectable for you.

As would (likely) getting my throat slit with a sharp knife in my sleep, but that doesn't mean that I would want it to happen.

First of all, the paper, even if we accept it as valid, argues that at least one of (1), (2) and (3) is true. It doesn't say that (3) is true; the author says he believes that (2) is true.

Actually, in the FAQ the author states that he believes (2) to be slightly more likely than the other two propositions, giving (3) approximately a 20% probability. That means he believes that there is a 20% (1 in 5) chance we are currently living in a simulation.

(3) is not a premise, it is a proposition that is a part of an argument about whether or not it is true depending on the truth value of the other propositions. Specifically, if (1) or (2) is true, the argument makes no claim about whether (3) is true or not.

I mixed up my terms -- I apologize. When I referred to "premise (3)" I meant "proposition (3)." The premise is that the simulations that it's talking about are ancestor simulations, not some other type of simulation, so the assumption is that it is a simulation of the same universe that the programmers (post-humans) live in.

I believe that the paper says that at least one of the three propositions is true. However, it would seem that if (1) or (2) is true, then (3) must be false (if it is very likely that humans go extinct before reaching a posthuman state, or if posthumans are unlikely to run ancestor-simulations, it seems very unlikely that posthumans would have developed a simulation in which we are currently living).

This has nothing to do with the question 9 in FAQ. This question does not begin, "Assuming that variant (3) is true, ..." The question is quite independent of the argument presented in the paper, and the given answer should therefore hold regardless of whether (3) is true or not.

I think the case is made in the paper that the 20% likelihood (or whatever number you place on it) is for ancestor-simulations, not for some other type of simulation. Therefore, I assume #9 in the FAQ also refers to ancestor-simulations.

Even so, I think it can be reasonably said that inserting a window into our world as described would provide evidence that we are in a simulation, particularly if we were able to communicate with the programmers. Given a 20% (or so) likelihood that we are in a simulation and a 100% likelihood that simulations exist, compared to a completely unknown likelihood that there are lying gods, and contrary to observed evidence that such a window is probably not a naturally occurring event (particularly if we can hold a coherent conversation with it), it seems far more likely that such a window would be telling the truth. So I still think I am in agreement with the point that we could be provided some fairly substantial evidence that we are in a simulation if the programmers wanted.

And, I still find the presented argument itself logically flawed.

...

The flaw is this:

...



Let's assume that a simulation that runs exactly the same each time is a single simulation. We can still assume that there would be many different simulations, and that we have a good chance of living in one of them. So even if you assume that a cached simulation playing back like a video tape is a single one, the argument still holds.

Let's also assume that simulating a posthuman civilization is prohibitively expensive...

Here I think I agree with you. It sounds to me like an artificial argument that is being made to avoid a paradox of infinite recursion. Regardless, if we accept the assumption that a post-human universe is "too expensive" to simulate, it would seem that for the simulated universes (1) is most likely true, which means that (3) is probably false.

But is the probability that we are likely a simulation (assuming 1 and 2 are false for the "real" universe) based on the possibly infinite recursion or is it entirely based on simulations running in the "real" universe?

If you have read it all the way here, congratulations.

I did! Congratulations for writing all the way here.

-Bri

[arthwollipot]

Won't this whole line of thinking boil down to 'Am I real? It's all very well saying I think therefore I am but I do I prove that I'm thinking? that these are my thoughts?' and other things guareented make me have a sudden craving for chocolate icecream and an escapist movie.

ETA: Any wandering grammar police, should I have a question mark at the end of my convoluted sentence above? As for the spelling police, well, you know that I know that I should really be doing hard labour somewhere....

Yes. :)

i presume you're referring to this thread

http://forums.randi.org/showthread.php?t=87370

it was discussed in depth - i introduced it (post 14).....but i think you'll find references to Bostrom on every page :)

That was the one. Yes. No wonder it was familiar.

Great thread by the way folks. Back to lurk mode...

[Thabiguy]

Even so, I think it can be reasonably said that inserting a window into our world as described would provide evidence that we are in a simulation, particularly if we were able to communicate with the programmers. Given a 20% (or so) likelihood that we are in a simulation and a 100% likelihood that simulations exist, compared to a completely unknown likelihood that there are lying gods, and contrary to observed evidence that such a window is probably not a naturally occurring event (particularly if we can hold a coherent conversation with it), it seems far more likely that such a window would be telling the truth. So I still think I am in agreement with the point that we could be provided some fairly substantial evidence that we are in a simulation if the programmers wanted.
Oh, okay. I have stated my position and see no point in reiterating it. Just a few points regarding the quoted paragraph:
- The "20%" could be 100% just as well as one in a billion. It's pure speculation and the value has no meaning (except possibly as subjective Bayesian probability). The probability is unknown.
- 100% likelihood that simulations exist in our universe.
- Consider what you would think if a window appeared, saying "This is NOT a simulation. You live in a real, non-simulated universe."

Let's assume that a simulation that runs exactly the same each time is a single simulation. We can still assume that there would be many different simulations, and that we have a good chance of living in one of them. So even if you assume that a cached simulation playing back like a video tape is a single one, the argument still holds.
Even the number of different simulations is arbitrary. Consider that the machine stores the state dumps in an encoded state; this is rather self-obvious, as the simulated universe does not consist of binary data and must be encoded somehow. Because there appears to be no preferred encoding, it can be argued that the simulation of a universe remains a simulation, no matter how the data are encoded. - However, regardless of the chosen encoding, it can be easily shown that for any given universe, and another universe of the same "length" (time steps computed), there exists a unique time-step-dependent encoding function that encodes the dumped states of the latter universe in such a way that their binary representation is equal to the binary representation of the dumped states of the former universe. It can therefore be argued that by simulating one universe in a particular encoding, the computer has also simulated another universe in another encoding, and an arbitrary number of other universes with the same number of time steps. Because there appears to be no reason why any particular encoding should be "more real" than another particular encoding, you would again have to arbitrarily decide just how many different universes you have simulated by running the program just once.

None of this, again, has any significance for inhabitants of any of the simulated universes. None of them would have any contact with any other universe. In the simulating universe, you could decide to say, "I have simulated one million different universes that all contain Joe, so the chance is a million to one that a randomly chosen Joe is real", but again, this statement is meaningful and can be verified only in the simulating universe, where we can consider each instance (however defined) of Joe independently. In none of the universes that you have decided you have simulated, such probability would have any meaning. It could be the case that for any universe that you decide you have simulated, the same universe also exists as a real, non-simulated universe. If a Joe in a particular universe decided that he's simulated, he could be correct when this statement is examined in the simulating universe, but wrong when considering that the same Joe, in the same universe, exists for real. His decision would be true or false depending on where is it examined; when restricted to his own universe, the truth value of such a decision is undefined.

Here I think I agree with you. It sounds to me like an artificial argument that is being made to avoid a paradox of infinite recursion. Regardless, if we accept the assumption that a post-human universe is "too expensive" to simulate, it would seem that for the simulated universes (1) is most likely true, which means that (3) is probably false.
Try to think about this a little more. How could (3) be false for simulated universes? If (3) were somehow false for the simulated universes, then the core argument falls apart: if the simulated universes are not likely to be simulated, then by the logic of the argument, you are likely to live in one of them, which means that (3) is also false for you. (All while assuming that (1) and (2) are false.) This is a major contradiction. Think about why this paradox appears and what part of the reasoning that leads to it might possibly be flawed.

But is the probability that we are likely a simulation (assuming 1 and 2 are false for the "real" universe) based on the possibly infinite recursion or is it entirely based on simulations running in the "real" universe?
I'm sorry, I don't understand the question.

[Dan O.]

A simulation doesn't need to be in the form of discrete elements and time steps. A more appropriate simulation of a universe would be in the form of a thought experiment. Here there is no time, space, energy or matter except as created through recursive refinement of the logic to answer higher level postulates.

Such a simulation of the real universe would inevitably include simulation of sub universes since this action is part of the real universe. Each sub universe would inevitably include a simulation of it's own sub universe ad-infinitum. This infinite recursion would result in a stack overrun in most computers, but a simulation run on logic could avoid this fate. If the results of the infinite recursive universe simulation cannot be found through logic then the simulation will lock up and that will be the result. This does not require the infinite recursion to actually take place so the higher level simulation would not fail due to resource exaustion.

[OnlyTellsTruths]

C.S. Lewis used this idea of reaching more and more real universes for the ending of one of the early Narnia books and expanded upon it in the ending of one of the later Narnia books, some ~70 years ago.

Though I know it was discussed already, you can't tell what level a simulation is by how complex you determine it to be, all you can know is that it gets more complex one direction, and less so the other. That's if it's even a simulation in the first place, which you can't know. There's no use wasting time wondering what's beyond a singularity like the big bang or black holes.

The real moral here for all the kids to learn is:

Always wait ten years for your computer to emulate the current generation of console videogames and save yourself a boatload of money.

[Bri]

- The "20%" could be 100% just as well as one in a billion. It's pure speculation and the value has no meaning (except possibly as subjective Bayesian probability). The probability is unknown.

Yes, but I'm assuming that the author is correct about the probabilities. He admits that the probabilities that you assign to the three propositions are entirely subjective.


- 100% likelihood that simulations exist in our universe.

And, if we are simulated, in the meta-universe as well. And, if that one is simulated, in the meta-meta-universe. So, in all universes that matter, there is a 100% likelihood that simulations exist.

- Consider what you would think if a window appeared, saying "This is NOT a simulation. You live in a real, non-simulated universe."

I would at first be skeptical that it was a hoax, and after I was satisfied that it wasn't a hoax, if I could carry on a conversation with the window my belief would depend on the windows' explanation of its presence. I would certainly consider the possibility that the window was lying, but without a reason to believe it was lying, I would probably be inclined to believe whatever it told me it was (a god, an alien with advanced communications, whatever) until evidence presented itself to the contrary.

Even the number of different simulations is arbitrary.

Wouldn't quantum randomness indicate that the exact same scenario wouldn't necessarily occur even under the same conditions? If that is the case, then caching a simulation wouldn't make any sense to begin with except to review the results of a simulation. I think we can safely count playing back a recording of a simulation as the same simulation.

In addition, I don't recall the paper being based on a particular number of simulations being run. I think it was based on the assumption that a "significant number" of simulations are run, however you want to define that. The higher the number of simulations, presumably the higher the chances we are in one of them. So define a simulation however you want, and then decide whether (2) still holds (i.e. the posthuman civilization will run a significant number of them).

In the most conservative case, assuming that proposition (1) is false, if a "real" posthuman society will run only a single simulation, we would have a 50% chance of being in it rather than being in the real thing leading up to the posthuman civilization, right? And that's assuming that the ancestor-simulation is terminated before it reaches the stage where it can spawn its own ancestor-simulation.

Try to think about this a little more. How could (3) be false for simulated universes? ... Think about why this paradox appears and what part of the reasoning that leads to it might possibly be flawed.

Maybe I misunderstood your point, but I just meant that the probabilities assigned to each proposition might be different for the "real" universe than for the simulated ones. For example, there is a major difference between the "real" universe and a simulated one: simulations may always be terminated before reaching the posthuman stage (as the paper says, it might be too "expensive" to allow them to proceed to the posthuman stage). In that case, simulated universes would not have their own ancestor-simulations whereas the "real" universe would still proceed to the posthuman stage (proposition 1 would be true for simulated universes, but false for the real one).

Even if that is the case (simulated universes cannot have their own simulated universes) and even if the "real" post-human universe creates just one ancestor-simulation, we might still very well be in it. If it creates more than one, the chances increase.

I'm sorry, I don't understand the question.

There are a (possibly infinite) number of levels of simulation possible IF simulations are allowed to progress to the posthuman stage where they could be expected to spawn their own simulations. This might increase our chances of being in a simulation considerably.

But I think the point of the paragraph you quoted is that it still seems like we have a 50% chance of being in a simulated universe even if simulated universes don't spawn their own simulations, and even if the "real" posthuman civilization creates only a single ancestor-simulation. And we have a "significantly" greater chance of being in a simulation if the posthuman civilization creates a "significant" number of ancestor-simulations.

-Bri

[Bri]

The real moral here for all the kids to learn is:

Always wait ten years for your computer to emulate the current generation of console videogames and save yourself a boatload of money.

And, of course, if you successfully simulate the universe on your Commodore-64 and then fast-forward the simulation until the next generation of console video games are invented, the graphics will still suck.

-Bri

[Paul C. Anagnostopoulos]

Indeed. For example, the back side of the cow will still be unrendered.

~~ Paul

[Cuddles]

The argument was, in essence - if we take a superrigid rod and make the balls spin fast enough, won't it be indistinguishable from an object that doesn't spin at all? As in, the balls move so quickly that the time in which they "go around" is under the "resolution of the universe"? And the answer is, if the balls' velocity is so high that the rotation period would somehow be under the resolution of the universe, then the even finer-graded 360deg change in their velocities, which needs to happen along their track to allow them to return to their original position, is by the same logic also under the resolution of the universe, the universe will be unable to perform it, and the balls will just fly apart with their large velocities.

While I agree with your post, I think there is an easier way to look at it, similar to the questions about spinning rigid rods having ends moving faster than light. The answer is that it is physically impossible for a rod to exist which could do this. If you make unphysical assumptions, your conclusion can't tell you anything about the real world.

I am reminded of a quote from Dr. George Box "All models are wrong. Some models are useful". At any rate, complexity doesn't necessarily align with accuracy. It's often the case that a simpler model provides more accurate answers. The more complex ones are much harder to program, more prone to errors, harder to verify, etc.

I'm assuming that anyone capable of simulating an entire universe will also be able to check that the programming is correct before running it. If a complex model is actually correct, it will always be more accurate than a simple one because it is closer to the reality. Questions about errors in models could be interesting, but aren't really relevant to this discussion. As for being harder to verify, it's really very simple. Is the answer from the model anywhere near actual reality? It is just as easy to check this for any model, no matter how simple or complex.

[Beth]

I'm assuming that anyone capable of simulating an entire universe will also be able to check that the programming is correct before running it. Okay, but that's not an assumption I'm willing to make. Why do you think this is a reasonable assumption? Often, the easiest and quickest way to assess the validity of a simulation is just to run it and see how well it works.

If a complex model is actually correct, it will always be more accurate than a simple one because it is closer to the reality. I'm not convinced that complex is necessarily better or more accurate even if you assume it's correctly done. I think in some cases, simple rules iteratively applied can generate better results than a more complicated one, particularly when modeling chaotic and/or dynamic systems. Additional complexity doesn't necessary add anything of significant value.
Questions about errors in models could be interesting, but aren't really relevant to this discussion. As for being harder to verify, it's really very simple. Is the answer from the model anywhere near actual reality? It is just as easy to check this for any model, no matter how simple or complex. What 'reality' the model is supposed to be closer to if the entire universe is hypothesized to be a simulation? What aspects of that 'reality' are most important to the designer? How do measure how closely model A fits with reality versus model B if A is closer in some ways and B is closer in others? These are not trivial questions to answer for the simulations and models we can build already. Further, another question that occurs to me is whether or not the simulation is supposed to reflect some sort of 'reality' outside the simulation. Not all human designed simulations have that goal in mind, why should we presume it is a goal of a being that would create our universe.

[INRM]

A simulation doesn't need to be in the form of discrete elements and time steps. A more appropriate simulation of a universe would be in the form of a thought experiment. Here there is no time, space, energy or matter except as created through recursive refinement of the logic to answer higher level postulates.

Such a simulation of the real universe would inevitably include simulation of sub universes since this action is part of the real universe. Each sub universe would inevitably include a simulation of it's own sub universe ad-infinitum. This infinite recursion would result in a stack overrun in most computers, but a simulation run on logic could avoid this fate. If the results of the infinite recursive universe simulation cannot be found through logic then the simulation will lock up and that will be the result. This does not require the infinite recursion to actually take place so the higher level simulation would not fail due to resource exaustion.

I'm not sure I understand that... wouldn't any simulation or computer have limits as to storage and computational capacity?

INRM
BTW... you know you all are basically covering existentialism albeit far more advanced than Descartes (I think)

[blobru]

I'm not sure I understand that... wouldn't any simulation or computer have limits as to storage and computational capacity?
...

Yeah, I'm a little lost too (as usual). :)

Wouldn't any simulation "crap out" well in advance of the point where the sims are capable of running a simulation as advanced as the one they're in (for they could never construct a computer as complex as the simulators')?

If that's the case, and there is a premium on memory savings, why would the simulators waste it by giving their sims consciousness, tracking all the physiological data that [supposedly] add up to embodied "consciousness", when they'd only need zombies -- behaviorally identical and cheaper to code -- to run a useful simulation?

Btw, your earlier question: "post human" I think refers to future humans who can "play God" so to speak -- create human beings, society, reality from scratch.

[Thabiguy]

I'll be slowly wrapping this up, if you don't mind, Bri. This takes a lot of time, and even though I think it has been worth it so far, I need to devote some to other issues, too.

And, if we are simulated, in the meta-universe as well. And, if that one is simulated, in the meta-meta-universe. So, in all universes that matter, there is a 100% likelihood that simulations exist.
This does not logically follow. But if I haven't made you see the error in that reasoning by now, I guess I'm not meant to.

I would at first be skeptical that it was a hoax, and after I was satisfied that it wasn't a hoax, if I could carry on a conversation with the window my belief would depend on the windows' explanation of its presence. I would certainly consider the possibility that the window was lying, but without a reason to believe it was lying, I would probably be inclined to believe whatever it told me it was (a god, an alien with advanced communications, whatever) until evidence presented itself to the contrary.
All right, good answer.

In addition, I don't recall the paper being based on a particular number of simulations being run. I think it was based on the assumption that a "significant number" of simulations are run, however you want to define that. The higher the number of simulations, presumably the higher the chances we are in one of them. So define a simulation however you want, and then decide whether (2) still holds (i.e. the posthuman civilization will run a significant number of them).
I tried to show that the number of simulated universe is subject to arbitrary definition to illustrate that the attempt at formal calculation in the paper has some significant flaws that are not addressed at all, and appears to lack the suggested rigor. I find the main flaw more fundamental, though - the lack of introduction of proper framework that would be necessary when dealing with issues such as interpretation of probability across universes (indeed, considering likelihood of events that have no physical reality in the universe in question), which is something that the classical probability theory doesn't address, but no extension is proposed. I believe that the reckless shifts of logical statements between different universes causes serious logical issues that flaw the entire argument, and I have illustrated this by showing how the failure to account for the difference in logical viewpoints across universes makes a seemingly reasonal additional assumption considered in the paper quickly lead to a contradiction. Because of these problems, I don't find the logic of the argument sound, nor its conclusions convincing. I might change my mind if these issues are properly addressed.

I think this sums up my position, and have no more to add.

In the most conservative case, assuming that proposition (1) is false, if a "real" posthuman society will run only a single simulation, we would have a 50% chance of being in it rather than being in the real thing leading up to the posthuman civilization, right? And that's assuming that the ancestor-simulation is terminated before it reaches the stage where it can spawn its own ancestor-simulation.
No, I find the entire notion of deriving probabilities (Bayesian, apparently) outside a universe and claiming that they apply when restricted to that universe, nonsensical. - Also, I have shown why it is impossible to objectively determine the number of universes that has been simulated by a civilization, or rather, why it is subject to arbitrary definition. A probability value derived from an arbitrarily chosen number has no objective meaning.

Maybe I misunderstood your point, but I just meant that the probabilities assigned to each proposition might be different for the "real" universe than for the simulated ones. For example, there is a major difference between the "real" universe and a simulated one: simulations may always be terminated before reaching the posthuman stage (as the paper says, it might be too "expensive" to allow them to proceed to the posthuman stage). In that case, simulated universes would not have their own ancestor-simulations whereas the "real" universe would still proceed to the posthuman stage (proposition 1 would be true for simulated universes, but false for the real one).
Yes. Then it follows that the argument's main point is flawed when it concludes that under certain assumptions we "almost certainly live in a simulation", but fails to adjust for the consequences of this on the part of the argument already presented before making such conclusion. In certain cases, this can be shown to lead to a paradox.

Even if that is the case (simulated universes cannot have their own simulated universes) and even if the "real" post-human universe creates just one ancestor-simulation, we might still very well be in it. If it creates more than one, the chances increase.
I will not argue with this; I will assume you are correct, and just ask you to ponder about the interesting prospect of dramatically increasing our chances of being simulated by running many ancestor-simulations. (As the paper puts it, "If we do go on to create our own ancestor-simulations ... we would therefore have to conclude that we live in a simulation.") And then about another interesting prospect: dramatically increasing the chances of pop-up windows appearing in our universe, by running many ancestor-simulations, letting them evolve to posthuman stage and inserting pop-up windows into their universes just after they create their first simulations. Because there would be so many of these simulations, it would be likely that we ourselves live in one such simulation, and could therefore expect pop-up windows to appear any minute after creating those simulations. Try to find differences between the reasoning behind this prospect and the reasoning behind the simulation argument.

All in all, thanks for a very interesting thread. :)

[Thabiguy]

While I agree with your post, I think there is an easier way to look at it, similar to the questions about spinning rigid rods having ends moving faster than light. The answer is that it is physically impossible for a rod to exist which could do this. If you make unphysical assumptions, your conclusion can't tell you anything about the real world.
Yes, you are right. I had pondered writing a few more sentences on how there would be other issues to consider in our universe, and that relativistic effects would hit before quantum mechanical effects, but then I decided not to, for the post was long enough, and also because the way it is, it can readily be applied to many toy universes with discrete time, such as various straightforward simulations of Newtonian mechanics. Virtually every person trying to make one will sooner or later witness a tightly bound system fly apart when it hits the resolution limit of the simulation.

[Dan O.]

I'm not sure I understand that... wouldn't any simulation or computer have limits as to storage and computational capacity?

When you get to the point of simulating a simulation, the recursion doesn't need to take place because the results have already been worked out. Just plug in the answer from the current top level simulation as the result of the simulated simulation and move on.

What I described as logic based simulation could also be interpreted as an observer driven system. If there isn't an observer of an event that will make a decision based on that event then the event will never exist. The event itself could have been an observer of other events in which case those other events will not exist either because there would be no need to simulate them. Whole systems of events could be treated as a bulk statistical property if there isn't an observer focusing on the individual events within the bulk.

If you are part of such a simulation then that you observe that you exist in this moment means that your existence in this moment is important in the grand scheme of things as is every detail of your environment that you observe. But your memory of past moments (including a moment ago when you thought you were important) may only be a bulk statistical simulation that was only necessary because your memory of that moment is all that is important.

[INRM]

Then what about the "first simulation" then Dan O?

INRM

[Dan O.]

Then what about the "first simulation" then Dan O?

INRM

There would probably need to be some optimizations to prevent the system from being boundless. Things like establishing a zero point where the time scale begins and local equivalence so rules that are worked out for one situation would apply to any similar situation regardless the absolute motion or orientation.

[blobru]

When you get to the point of simulating a simulation, the recursion doesn't need to take place because the results have already been worked out. Just plug in the answer from the current top level simulation as the result of the simulated simulation and move on.
...

But won't the relevant values and variables almost certainly change each time a simulation is run within another?
For example, at the lower level the simulators plug in initial values for whatever start point they like. Then in the simulation the sims get to a point where they can run simulations. If the simulation history and science isn't exactly like the lower level history and science, and hard determinism notwithstanding it shouldn't be, the bulk statistical values will be no good, as they'll be using different variables with different values.

[Beth]

Well, for one thing, the programmers might decide to pull the plug at any moment. Not that the universe might not be destroyed by perfectly natural means even if we aren't a simulation, but the probability of our sudden demise seems a lot higher if we are a simulation.

-Bri
It's true, if a creator exists, it's possible it could decide to destroy it's creation at any time, In fact, that's a favorite theme of certain fundamentalist preachers. But I don't see much difference between assuming the probability is high that we are a simulation and assuming the probability is high that an intelligent creator exists. In fact, given the nature of the 'simulation', it seems like pretty much the same idea to me.

Are there any other reasons you find the idea that we might be a simulation disturbing? I suspect there is a certain inherent revulsion to the idea. At least I initially felt that way, but the more I thought about what the word 'simulation' means when applied to our entire universe, the less it bothered me.

[Bri]

Well, I think most fundamentalists tend to believe that if their God chose to destroy the world, they would not be destroyed even if others are. I'm also not a fundamentalist.

So, it may be the same idea, but in this case it would seem that the programmers could easily pull the plug on all of us rather suddenly for any reason whatsoever or even by accident (perhaps a power failure).

Also, the idea is still pretty creepy to me.

-Bri

[Correa Neto]

Well, an omniscient creator will know that things in Universe V.googolplex.googol will not work... So, why bother creating?

[Bri]

I'll be slowly wrapping this up, if you don't mind, Bri. This takes a lot of time, and even though I think it has been worth it so far, I need to devote some to other issues, too.

Not a problem. I appreciate the discussion, but understand the time concern.

Because of these problems, I don't find the logic of the argument sound, nor its conclusions convincing. I might change my mind if these issues are properly addressed.

Fair enough. There are a lot of links to related articles on that website, so perhaps these issues are discussed and addressed.

Because there would be so many of these simulations, it would be likely that we ourselves live in one such simulation, and could therefore expect pop-up windows to appear any minute after creating those simulations. Try to find differences between the reasoning behind this prospect and the reasoning behind the simulation argument.

I think it would be unlikely that the programmers would want to pop up windows in every simulation even if they chose to do so in some of them. I also don't think the paper assumes that multiple simulations would unfold in precisely the same way, otherwise there wouldn't be any reason to create more than one simulation. Therefore, it may be that some universes decide to pop up windows in more simulations than others -- some maybe not at all. So I'm not sure that we could expect a window to pop up in our own post-human universe as soon as we created a simulation, even if we chose to do so in some or all of our own simulations.

-Bri

[INRM]

Using the simulation argument... would one suggest that the computer plays god?

lol

INRM

[Cuddles]

Okay, but that's not an assumption I'm willing to make. Why do you think this is a reasonable assumption? Often, the easiest and quickest way to assess the validity of a simulation is just to run it and see how well it works.

It's a reasonable assumption because that's what the discussion is about. We're not discussing possible troubleshooting and debugging processes of hypothetical programs, we're discussing the simulations themselves.

I'm not convinced that complex is necessarily better or more accurate even if you assume it's correctly done. I think in some cases, simple rules iteratively applied can generate better results than a more complicated one, particularly when modeling chaotic and/or dynamic systems. Additional complexity doesn't necessary add anything of significant value.

In a word, no. Look at it like this, imagine a simulation that replicates a phenomenon perfectly. Now simpify it. Which version is more complex and which is more accurate? No matter how you look at it, if you take information out of the simulation, it is impossible to make it more accurate. Of course, it is possible for a complex simulation to be less accurate if the more complex bits are wrong, but that's not the point. Yyou are correct that additional complexity doesn't necessarily add anything of value, it depends on how accurate you want your answers to be. However, while 21 decimal places might not be any more useful than 20, it's still closer to reality.

What 'reality' the model is supposed to be closer to if the entire universe is hypothesized to be a simulation? What aspects of that 'reality' are most important to the designer? How do measure how closely model A fits with reality versus model B if A is closer in some ways and B is closer in others?

Which is exactly why more complex simulations are more accurate. The reason A and B are different for different aspects is because they've both been simplified to the point where they no longer reflect all of reality accurately. If you're only interested in the aspects that A gets right, by all means only use A. If you're interested in them both, you need to either run them both or make a more complex simulation that gets both parts right.

Further, another question that occurs to me is whether or not the simulation is supposed to reflect some sort of 'reality' outside the simulation. Not all human designed simulations have that goal in mind, why should we presume it is a goal of a being that would create our universe.

I can't think of a single simulation that wasn't written with the purpose of understanding a particular part of this universe better. That's why they#'re called simulations - because they simulate things. In any case, the whole point of this thread was about someone creating a simulation of their own universe. We assume this to be the case because otherwise the thread would be about something else entirely.

[Unalienable]

A great nuance on this subject the question of "pulling the plug."

Presumably, if we were in a computer simulation and our parent universe pulled the plug on their supercomputer, we would perish, right?

I argue that we would not. The computer program is simply grinding away according to precise rules, and it is reaching some final conclusion to the mathematical dance that it is performing. Therefore the final state, and all states between now and then, already exist in a Platonic sense.

What do I mean a "Platonic" sense? Well, do you believe that pi had a one trillionth digit? Of course you do, it has to. But you don't know what it is, nobody does, because we've never computed it. Now suppose I build a supercomputer designed to find the trillionth digit of pi, and after a few years, we pull the plug on it. Pi still has a trillionth digit--we just don't know what it is. But those digits exist in a Platonic sense; we are not creating those digits with the computer, we are discovering them.

Therefore, if our universe was being simulated, the computer would not be creating us, it would only be discovering us.

Therefore, even if they pulled the plug, we would continue to exist.

Which makes you think: if the computer is not necessary for us to continue existing, why is it even necessary for us to exist in the first place? Crazy, eh?

[Dan O.]

Well, do you believe that pi had a one trillionth digit? Of course you do, it has to. But you don't know what it is, nobody does, because we've never computed it.

The trillionth digit if pi is 8 in hex. It was computed about 9 years ago. Please to try to keep up.

[blobru]

A great nuance on this subject the question of "pulling the plug."

Presumably, if we were in a computer simulation and our parent universe pulled the plug on their supercomputer, we would perish, right?

I argue that we would not. The computer program is simply grinding away according to precise rules, and it is reaching some final conclusion to the mathematical dance that it is performing. Therefore the final state, and all states between now and then, already exist in a Platonic sense.

What do I mean a "Platonic" sense? Well, do you believe that pi had a one trillionth digit? Of course you do, it has to. But you don't know what it is, nobody does, because we've never computed it. Now suppose I build a supercomputer designed to find the trillionth digit of pi, and after a few years, we pull the plug on it. Pi still has a trillionth digit--we just don't know what it is. But those digits exist in a Platonic sense; we are not creating those digits with the computer, we are discovering them.

Therefore, if our universe was being simulated, the computer would not be creating us, it would only be discovering us.

Therefore, even if they pulled the plug, we would continue to exist.

Which makes you think: if the computer is not necessary for us to continue existing, why is it even necessary for us to exist in the first place? Crazy, eh?


Sure pi has a trilliontrillionth digit -- as yet unknown [I think]. To evaluate it, to create digits as data, we have to convert pi to an algorithm, and then do the calculation (run the program).

Likewise, the values of future states of the sim-universe are unknown. By pulling the plug on the program, the values, whatever they are, will never have been created: they won't exist anywhere as data.

Better to th