2 ms·

As computer engineers we understand it's doable. Give it a couple of centuries and our civilization could reach a computing capacity to simulate the entire obse

As computer engineers we understand it's doable. Give it a couple of centuries and our civilization could reach a computing capacity to simulate the entire observable universe This seems at the very least questionable. Why do you think you can simulate the entire universe just using a tiny fraction of the resources of the universe? It seems to me that could only work if the universe were highly compressible and I don't see any obvious evidence for that. And even then I think it could not work because if the universe is in some sense redundant and could be compressed then it actually provides less resources to work with than if it would not have those redundancies. Take as an analogy for the later two processors executing the same code in lockstep as it is sometimes done for redundancy in safety critical systems. You can mostly get away with only simulating one of the processors but you essentially also only have one processor to work with. And while you can probably simulate the processors on the processors the simulation will probably be at least ten times slower than the actual processor if you only want to use one tenth of the available resources, i.e. clock cycles. And finally memory will be a real problem because you only have as much memory available as you want to simulate.

chr1 9y ago
You don't need to simulate the entire universe, just the tiny bit where interesting things are happening (read surface of earth), places farther away can be simulated with much lower precision as long as they look plausible. In that sense the universe is indeed highly compressible as it is mostly empty/uninteresting. The question is how large computing device is needed to simulate a brain, and whether that device can be digital or analog. If the size of such device can be comparable to the size of human body, then we can create a useful simulation.
- danbruc 9y ago
  The part of the universe that is empty also provides nothing you can use for a computer to run a simulation on, i.e. you don't have to simulate it but it doesn't help running the simulation either. And given that we probe the entire visible universe looking at gravitational waves from across the galaxy, spectral lines in the light of galaxies at the other end of the universe, and the decay products of Higgs particles, you still need a quite good simulation or good heuristic approximations for a lot of stuff in order to not produce inconsistencies. Or you have to limit what your simulated humans can do and keep them actively away from discovering your speed optimizations. Faking parts of something as complex as our universe without it becoming obvious would probably not be a small feat.
elorant 9y ago
For all we know, 99,999% of the universe could be just a projection. You don't need to simulate all of it down to the proton level, just the things we can see. And the reason our cosmological theories fail miserably in things like dark matter or dark energy could be because there's really nothing out there.
- danbruc 9y ago
  If 99.999 % of the universe is fake, than you only have to simulate 0.001 % of the universe but you also only have 0.001 % of a universe to turn into a computer and run a simulation on.
  - elorant 9y ago
    Does a computer that runs a simulation of the universe has to be as big as the universe? In which case the universe is the simulation.
    - danbruc 9y ago
      I don't know the answer but I guess one could at least put limits on that with information theoretical or thermodynamical methods. My intuition is that you can trade size for execution speed, i.e. you can simulate the entire universe with a small computer if you are willing to wait long enough. On the other hand storing the configuration of the simulated universe would probably take up the entire universe, or at least to me it is not obvious how you could compress the state of the universe in the general case. There are certainly low entropy states that would require less memory than the system provides but they are extremely rare. The only way out I can imagine is storing a low entropy state and spending even more time to calculate parts of higher entropy states on demand but never trying to store those high entropy states. But it is again not obvious to me that this would actually work because it would require rather strong bounds on possible intermediate states. And I don't think it would be meaningful to say the universe is the simulation. A processor executing some code is not a simulation of a processor executing some code, an electron flying across the room is not a simulation of an electron flying across the room, a dog walking down the sidewalk is not a simulation of a dog walking down the sidewalk. A simulation requires, at least that would be my ad hoc definition, one system recreating the states and behaviors of a different system, they can not be the same system.

chr1 9y ago

You don't need to simulate the entire universe, just the tiny bit where interesting things are happening (read surface of earth), places farther away can be simulated with much lower precision as long as they look plausible. In that sense the universe is indeed highly compressible as it is mostly empty/uninteresting. The question is how large computing device is needed to simulate a brain, and whether that device can be digital or analog. If the size of such device can be comparable to the size of human body, then we can create a useful simulation.

danbruc 9y ago
The part of the universe that is empty also provides nothing you can use for a computer to run a simulation on, i.e. you don't have to simulate it but it doesn't help running the simulation either. And given that we probe the entire visible universe looking at gravitational waves from across the galaxy, spectral lines in the light of galaxies at the other end of the universe, and the decay products of Higgs particles, you still need a quite good simulation or good heuristic approximations for a lot of stuff in order to not produce inconsistencies. Or you have to limit what your simulated humans can do and keep them actively away from discovering your speed optimizations. Faking parts of something as complex as our universe without it becoming obvious would probably not be a small feat.

elorant 9y ago

For all we know, 99,999% of the universe could be just a projection. You don't need to simulate all of it down to the proton level, just the things we can see. And the reason our cosmological theories fail miserably in things like dark matter or dark energy could be because there's really nothing out there.

danbruc 9y ago
If 99.999 % of the universe is fake, than you only have to simulate 0.001 % of the universe but you also only have 0.001 % of a universe to turn into a computer and run a simulation on.
- elorant 9y ago
  Does a computer that runs a simulation of the universe has to be as big as the universe? In which case the universe is the simulation.
  - danbruc 9y ago
    I don't know the answer but I guess one could at least put limits on that with information theoretical or thermodynamical methods. My intuition is that you can trade size for execution speed, i.e. you can simulate the entire universe with a small computer if you are willing to wait long enough. On the other hand storing the configuration of the simulated universe would probably take up the entire universe, or at least to me it is not obvious how you could compress the state of the universe in the general case. There are certainly low entropy states that would require less memory than the system provides but they are extremely rare. The only way out I can imagine is storing a low entropy state and spending even more time to calculate parts of higher entropy states on demand but never trying to store those high entropy states. But it is again not obvious to me that this would actually work because it would require rather strong bounds on possible intermediate states. And I don't think it would be meaningful to say the universe is the simulation. A processor executing some code is not a simulation of a processor executing some code, an electron flying across the room is not a simulation of an electron flying across the room, a dog walking down the sidewalk is not a simulation of a dog walking down the sidewalk. A simulation requires, at least that would be my ad hoc definition, one system recreating the states and behaviors of a different system, they can not be the same system.