Can a computer simulation simulate another computer running another simulation

[sleemanj]

The consideration of "universe is a simulation" is, in my opinion, no better (or worse) than the consideration of "religion", it's a philosophical debate, not one of scientific rigour. 

Simply, it's not falsifiable, at least I can't think of a manner in which the answer "no" can be proven, it, as per the god question, always boils down to "it could be that they constructed the universe in such a way as for us find this negative answer".

Certainly it is possible, some might even say it is **likely** from a probabilistic perspective (I guess here at least here it has one-up on religion, a clear logical and reasoned progression to provide a meaningful argument for the affirmative), but it doesn't really make for a useful line of scientific enquiry.

[xrunner]

I found this. It might be a suitable place to start really damaging your mind. If it is in fact really your mind.
http://www.simulation-argument.com/simulation.html

^^^ Thanks for that link.

Simply, it's not falsifiable, at least I can't think of a manner in which the answer "no" can be proven, it, as per the god question, always boils down to "it could be that they constructed the universe in such a way as for us find this negative answer".

sleemanj, yes, indeed it must be falsifiable, I understand. How do you falsify a perfect simulation that limits your ability to investigate. Things like wave particle duality, which have no real explanation - the property simply is and you can't explain it at a more fundamental level. I don't know, this is the challenge. All we can hope for is that it's not really a perfect simulation, and keep looking.

[zapta]

I've always jokingly pondered on the idea that the universe is just a big simulation running on God's computer. 

And since god is omnipotent, he himself can be a simulation running on a 0402 resistor.

[MK14]

There was a Sci-Fi related, British comedy with this very topic.

The episode "Back to reality", implements it. I won't spoil the episode for you (as you may have NEVER seen it). But it is at least partly about simulations of real life.

I CAN'T find any (free) full screen copies of it (probably because it is copy righted). But here is a part sized screen version:

[ebclr]

You can rin on a Linux machine a vmware session running a windows machine, that will run proteus with a 8051 simulation.

3 levels of simulation just as a sample. far more than that are easy

[xrunner]

Well, let me ask this then -

Can a computer simulate anything, or is there a thing, or process, that a computer theoretically can never simulate?

Can anything be reduced to ones and zeros, even human intelligence?

[ebclr]

Not everything, I don't know how to make sex with a computer

[MK14]

Well, let me ask this then -

Can a computer simulate anything, or is there a thing, or process, that a computer theoretically can never simulate?

Can anything be reduced to ones and zeros, even human intelligence?

One big problem area, is simulating "REAL" random numbers. I.e. NOT pseudo random numbers.
These can be created (sort of) via hardware random number generators, using e.g. Thermal noise in electrical circuits.

The problem is simulated ones (such as pseudo random numbers, generated using defined mathematical methods/formulas), usually have mathematical side effects, which can be spotted by clever mathematical analysis software, such as DieHard Randomness Tester.
https://en.wikipedia.org/wiki/Diehard_tests

[steverino]

There is nothing new or exciting or deep about computers simulating computers, etc.  IBM was doing this in the early 70's.  They even have a complete OS designed for such purposes (VM CMS).  A simulation, by definition, involves creating an abstraction of some entity or concept and then using models, created by humans, to implement the abstraction.  Your simulation is only going to be as good as your models.  But it will never be more than a simulation.  A more interesting concept, I believe, is Artificial Intelligence software.  Here the computer actually learns as it solves problems and establishes it's own knowledge (as opposed to pre defined models in the simulation world.)

[Hypernova]

Surprised no one posted this XKCD yet.


For me the only real question is: What can we do if we are?

[Artlav]

Can a computer simulate anything, or is there a thing, or process, that a computer theoretically can never simulate?
Can anything be reduced to ones and zeros, even human intelligence?

Pretty much.
The universe is a mathematical object, and as such can be simulated.
Basically, anything you have a description of can be simulated.

Two problems here, however.
One, we don't have good descriptions for everything so far.
And two, the simulation speed is not going to be particularly fast.

Compare Bochs and VMWare, for example - one runs slowly, while the other runs fast.
The first is a full emulator, computing the results of every instruction of the PC architecture. It would run the same on a completely different CPU.
VMWare, on the other hand, is virtualization. It uses the host CPU to directly execute most instructions, and only computer the privileged ones. So it's much faster, but won't run on a different kind of a CPU.

In the same way, you can compute the protein folding, for example, on a regular computer but it would take you months while in the test tube it would fold in femtoseconds.
However, a not-quite-invented-yet quantum computer would be able to "virtualize" the computation by using the universe's "instructions" directly, and compute the folding in minutes or seconds.

So, the closer the simulated thing is to reality (defined as the host medium), the easier it is to simulate.

The consideration of "universe is a simulation" is, in my opinion, no better (or worse) than the consideration of "religion", it's a philosophical debate, not one of scientific rigour.

Perhaps, but it opens up the idea to look for "bugs" in the simulation, to look in the places you won't ordinarily bother to look, to consider what consequence this might have, and so on.

At the root of science is coming up with the ideas, which are then to be tested, falsified, developed and so on.
But if it never occurs to you to, say, check if  the space is actually perfectly flat everywhere, then you'd never falsify that hypothesis and enable understanding of what we know as general relativity.

Philosophy is quite important in that regard, since it lets you look at things from different angles and can help to notice something interesting.

[eugenenine]

Its not really simulation if you use qemu or vmware or virtualbox since your running the same cpu architecture across them all, its just multitasking at another level.
To truly simulate you need a different architecture, for example run an Amiga (M68000) emulator under your x86 linux box and then run a commodore 64 emulator under the Amiga emulator.   
I used those specific examples because I've done it.

[botcrusher]

I used bochs as an exanple for this reason. Bochs is an x86 emulator.

[douardda]

A french comics writer i really like  wrote this blog post a while ago, enjoy -:)







edit: put links for the whole story

[CatalinaWOW]

This thread shows that the capacity to argue how many angels can dance on the head of a pin is still very much with us.

By the way, I would interpret Gödel's incompleteness theorems as proof there is something that cannot be simulated, but that may be my incomplete understanding at work.

[daqq]

The universe is a mathematical object, and as such can be simulated.

The Universe is applied math.
https://xkcd.com/435/

[xrunner]

In theory ...

Could an Intel 386 PC simulate the human brain, given enough memory and time, or is the brain not subject to being modeled in lines of code? Do you think human level AI can be created by programming at all?

What am I trying to ask, I guess, is can human consciousness be created and reside in a computer? Maybe we really don't know but what do you all think?

[daqq]

Could an Intel 386 PC simulate the human brain, given enough memory and time, or is the brain not subject to being modeled in lines of code? Do you think human level AI can be created by programming at all?

What am I trying to ask, I guess, is can human consciousness be created and reside in a computer? Maybe we really don't know but what do you all think?

I think so. Given an accurate enough model of a neuron, the connections, the physics, enough memory and time I don't see why not.

edit: Though getting around the architecture limitations, particularly the memory limit might get tricky 

[radar_macgyver]

A well-known cellular automaton is Conway's "Game of Life". Some liken it to a subatomic physics simulation (the gliders and space ships are the analog of quarks, for example). One can simulate the Game of Life within the Game of Life:

http://www.conwaylife.com/wiki/OTCA_metapixel

[Kilrah]

Could an Intel 386 PC simulate the human brain, given enough memory and time, or is the brain not subject to being modeled in lines of code?

That following the "simulation" paradigm I'm also really fond of the brain may not even be a thing beyond something we're "told" to look at as being the source of the "human consciousness", and yes by definition that consciousness would be computer-generated but would have nothing to do with a brain...

[Kilrah]

Before we had computers (from the jist of stories I have read about), some people simulated them by hand. To illustrate how useful they would be, and prove they work. I.e. they would write out the program on paper, then execute it line by line, with pencil and paper, or something.

I guess in that, the "computer" running the pencil and paper simulation is our own "wetware" machine. 

The wetware machine is pretty flexible and can actually execute any other machine's instructions - done daily by programmers when debugging code, and comparing the output of their own processing of the instructions to what the machine they're trying to program puts out

[timofonic]

Could an Intel 386 PC simulate the human brain, given enough memory and time, or is the brain not subject to being modeled in lines of code?

That following the "simulation" paradigm I'm also really fond of the brain may not even be a thing beyond something we're "told" to look at as being the source of the "human consciousness", and yes by definition that consciousness would be computer-generated but would have nothing to do with a brain...

No quakery, please. I'm atheist.

[steverino]

I just finished writing my simulation routine for god (God).  Wanna see it?

void god() {return;}

[Artlav]

By the way, I would interpret Gödel's incompleteness theorems as proof there is something that cannot be simulated,

The odd thing here is that a mathematical pattern exists regardless of whether it is computed by something or not.
I made a visual example for that here: http://orbides.org/apps/motion.html

Could an Intel 386 PC simulate the human brain, given enough memory and time, or is the brain not subject to being modeled in lines of code? Do you think human level AI can be created by programming at all?

It will. Probably is going to be absurdly slow, but it's fundamentally possible.

A lot of "brain computing power" estimation are off by a bag of orders of magnitude.

Consider a CPU runs at 1GHz.
If you were to simulate every transistor of it, then the simulation (on another 1GHz CPU) will run an instruction per second if you're lucky.
If you were to simulate the logic gates the transistors implement, you'd get a KHz.
If you simulate the instruction set the logic gates implement, you'd get an MHz.
If you rewrite the algorithms that the processor runs to the target CPU, then you will run at a GHz or more.

Same with the brain - trying to simulate every neuron would take an obscene amount of computing power.
While trying to re-implement the algorithms the neurons run into modern hardware should need much less than that.
I would guess an FPGA for the visual cortex, and a regular Core i7 for the rest of the brain to run it realtime, but it might be more or less - we don't really know yet.

[xrunner]

Same with the brain - trying to simulate every neuron would take an obscene amount of computing power.
While trying to re-implement the algorithms the neurons run into modern hardware should need much less than that.
I would guess an FPGA for the visual cortex, and a regular Core i7 for the rest of the brain to run it realtime, but it might be more or less - we don't really know yet.

I don't know enough about it, but I've thought about how to attempt to create an AI, at least to get the inputs from the environment to "IT". Seems like you would devote an entire processing "unit" to each of the senses: eyesight, hearing, touch, etc. Each unit would then classify what it was receiving into available known items from a very large database.

So the sight processor would classify all the objects it was looking at, such as in a room it might perceive a TV, people, pictures on the wall, animals, and so on. Then the sound processor would classify all the sounds it heard into known categories such as what music was playing, what people were talking and so on. So all the pre-"classification" would already be done for the main AI system.

The database of known objects to the human race, objects, sounds and so on, could reside on the cloud where a large computer database would be maintained.

All the known "objects" would then go to a main processor, where they would be acted upon by a cognitive system, yet to be devised. Is that thinking way out of line?