4 ms·
> > But because their simulations will have to run on computers that fit in our universe, presumably the simulated universes will be smaller than ours—in the se
by Karellen 3y ago
> > But because their simulations will have to run on computers that fit in our universe, presumably the simulated universes will be smaller than ours—in the sense of fewer bits and operations needed to describe them.
> Can't it just be the same but slower?
The number of bits needed to simulate another universe can't be larger than the number of bits we can store in this universe - there isn't enough space to store them! It's not a matter of simulation speed.
Theoretically, we could simulate another universe with the same number of bits that exist in this universe, but then there wouldn't be enough bits to store you in this universe. It's the paperclip problem all over again - the simulator doesn't want to destroy you, but your existence uses up bits that the simulator needs for its own purposes.
- WithinReason 3y agomany algorithms allow trading space for time
- Karellen 3y agoEven if you don't run the simulation at all, you can't hold the state of a single moment of the simulation in this universe, if you need more bits to do so than are available. I don't understand how it might even be possible to get around that by "trading space for time". Can you expand on that, at all?
- WithinReason 3y agoe.g. lossless compression of state space and piecewise evaluation. Better compression = less space, more time
- b112 3y agoThat's silly. There are whole regions of space with no actual observer. Those can be swapped to disk, or DB backend until needed. They exist, they are part of the simulation, just not actively rendered. Or they can be as sparse files in linux, null until used. I think the big concern would then be, too many active NPCs would mean too much active universe, and then an OOM issue. But of course people die, so once all the people who lived, and have memories of a thing die, then that global memory can be wiped. There are lots of methods to save space! And to have bits not rendered, or taking active ram. So over-committed sim space is a thing. Heck, maybe the project had plans for adding RAM and DB storage space, before active sim space got too big... but you know, budgets change.
- Karellen 3y ago> There are whole regions of space with no actual observer. Those can be swapped to disk, What disk? Where is this disk? Suppose the universe contains 10⁸⁰ atoms. And we somehow figure out a way to store 10 bits of information per atom, giving our universe a storage capacity of 10⁸¹ How do you propose to simulate a universe that requires 10⁸¹ + 1 bits to describe? What disk are you going to swap the extra information out to? Where will this disk live - inside or outside of our universe?
- jncfhnb 3y agoLazy algorithm and generator functions
- b112 3y agoEven gzip will store lots of zeros in a tiny, tiny space. If a region of space isn't observed, it's a big fat load of null. If it is observed up close, whatever persistent storage you use for that space can be wiped, when the observer ceases to exist. Even now, when we observe close planetary systems, it isn't as if we are looking at all the atoms. We certainly aren't observing all the space on the planets. It's all mega fuzzy, and be therefore a fuzzy property. An approximation. But of the observer population spikes, then too much would be observed and active at once... crash!
- jncfhnb 3y agoEh, you can’t really guarantee that internal actors can observe things in any meaningful fashion that requires data storage. You can’t guarantee that you’re actually experiencing anything you believe yourself to be from the perspective of an outside observer.
- b112 3y agoWell yes I can, because it's part of the theory I concocted above. You're welcome to espouse alternative theories of course.