3 ms·

well, the upper bound for the amount you'd have to simulate would be smaller than the universe, you'd only need to simulate a volume containing all particles th

by sullyj3 10y ago

well, the upper bound for the amount you'd have to simulate would be smaller than the universe, you'd only need to simulate a volume containing all particles that could causally affect you within the duration of the simulation, which would be limited by lightspeed. But I think, even then, you could simulate distant stuff at far lower resolution than near stuff, in a way that dropped off in expense very very quickly. You wouldn't have to fully simulate a star I was looking at for instance, in order to represent it to my eye in a way that was indistinguishable from the real thing. you just need a little twinkling thing.

vidarh 10y ago
You can go further than that. Depending on your goal, you only need to simulate enough that subject being simulated can not notice the difference. Everything else can be generated on demand. E.g. consider the game Elite. It creates an - admittedly very simple - illusion of a persistent universe by using pseudo-random numbers to generate star systems. It then populates those systems with ships whenever you visit one. But those ships only "exists" when you are in the system. Leave it and they disappear. That's obviously too simple, but the point is that most of what you see you don't interact with more than superficially: For most of the ships you have no way of knowing (other than by looking at the code...) that if you turn around and fly back to the same systems, the simulations you run into are new rather than have just moved. In general, you would be able to discontinue and discard most simulated objects after varying amounts of time. E.g. you may need to be able to regenerate a beach the subject has visited so that it looks the same, but only hours afterwards you can discard the sandcastle the subject built. In fact, I'd argue that one of the problems with some open-world simulation games today is that they permanently retain state they shouldn't. E.g. if you chop down a tree in Minecraft and leave a stub, it will remain there forever. This is made worse, because the single player version and most servers will only simulate chunks where there are players - other chunks will be "in stasis", so if you quickly leave an area, states that even in-game should be brief and temporary gets frozen in time and restarted long after they should have changed if the player later comes back. A "perfect" simulation will continue simulating any area "touched" by the "players", but over time gradually erase those state changes: Stubs rot, area overgrown etc.; buildings fall apart. Over time you can let state left behind by a player converge on a generator function that lets the simulation eventually discard the state change entirely and fall back on just generating the area on demand.
- sullyj3 10y ago
  I was kind of vaguely gesturing towards that idea, but you crystallized and articulated it much better than I could :)
  - vidarh 10y ago
    It's been a favourite thought of mine ever since I first realised that was how Elite managed to pack "that much universe" into just a few kilobytes :) There's also an interesting parallel in philosophy. While the most famous dichotomy is that of the idealists vs. materialists, with the idealists basically saying (vastly oversimplified) we can't trust sensory input - it could all be in our mind, in the in the late 1800's there was a philosophy called emirio-criticism or alternatively Machism (after Ernst Mach) that took a middle road: They argued that only that which you could directly observe had a meaningful real existence. Scientific laws for example, according to Ernst Mach, existed to make data comprehensible, but as such according to Mach is just an idealised representation of the data rather than a description of reality. As such, he for example opposed atomic theory, because he saw it as insufficiently "economical" - you can not directly observe atoms, and so a theory that propose them rather than aim for the most economical possible description that matches the fact seemed unwarranted to him. Today his theory is less known than one of the dominant criticisms of it: an almost equally obscure book by Lenin called "Materialism and Empirio-Criticism" written to discredit Bogdanov who believed in Mach's theory. It's not a very good book. But while Machism doesn't seem very useful for interpreting the world(e.g. what makes human sensory organs more reliable than mechanical aids?) it's interesting for its argument of drawing a line between the economical idealised representation of reality and reality based on what is directly observable, which I find fascinating given the issue of simulation. Basically, the fidelity of simulation largely boils down to resolving observation to direct observation of known/knowable detail vs. idealised economical representations that the observer has no way of disproving.

by sullyj3 10y ago

vidarh 10y ago

You can go further than that. Depending on your goal, you only need to simulate enough that subject being simulated can not notice the difference. Everything else can be generated on demand. E.g. consider the game Elite. It creates an - admittedly very simple - illusion of a persistent universe by using pseudo-random numbers to generate star systems. It then populates those systems with ships whenever you visit one. But those ships only "exists" when you are in the system. Leave it and they disappear. That's obviously too simple, but the point is that most of what you see you don't interact with more than superficially: For most of the ships you have no way of knowing (other than by looking at the code...) that if you turn around and fly back to the same systems, the simulations you run into are new rather than have just moved. In general, you would be able to discontinue and discard most simulated objects after varying amounts of time. E.g. you may need to be able to regenerate a beach the subject has visited so that it looks the same, but only hours afterwards you can discard the sandcastle the subject built. In fact, I'd argue that one of the problems with some open-world simulation games today is that they permanently retain state they shouldn't. E.g. if you chop down a tree in Minecraft and leave a stub, it will remain there forever. This is made worse, because the single player version and most servers will only simulate chunks where there are players - other chunks will be "in stasis", so if you quickly leave an area, states that even in-game should be brief and temporary gets frozen in time and restarted long after they should have changed if the player later comes back. A "perfect" simulation will continue simulating any area "touched" by the "players", but over time gradually erase those state changes: Stubs rot, area overgrown etc.; buildings fall apart. Over time you can let state left behind by a player converge on a generator function that lets the simulation eventually discard the state change entirely and fall back on just generating the area on demand.

sullyj3 10y ago
I was kind of vaguely gesturing towards that idea, but you crystallized and articulated it much better than I could :)
- vidarh 10y ago
  It's been a favourite thought of mine ever since I first realised that was how Elite managed to pack "that much universe" into just a few kilobytes :) There's also an interesting parallel in philosophy. While the most famous dichotomy is that of the idealists vs. materialists, with the idealists basically saying (vastly oversimplified) we can't trust sensory input - it could all be in our mind, in the in the late 1800's there was a philosophy called emirio-criticism or alternatively Machism (after Ernst Mach) that took a middle road: They argued that only that which you could directly observe had a meaningful real existence. Scientific laws for example, according to Ernst Mach, existed to make data comprehensible, but as such according to Mach is just an idealised representation of the data rather than a description of reality. As such, he for example opposed atomic theory, because he saw it as insufficiently "economical" - you can not directly observe atoms, and so a theory that propose them rather than aim for the most economical possible description that matches the fact seemed unwarranted to him. Today his theory is less known than one of the dominant criticisms of it: an almost equally obscure book by Lenin called "Materialism and Empirio-Criticism" written to discredit Bogdanov who believed in Mach's theory. It's not a very good book. But while Machism doesn't seem very useful for interpreting the world(e.g. what makes human sensory organs more reliable than mechanical aids?) it's interesting for its argument of drawing a line between the economical idealised representation of reality and reality based on what is directly observable, which I find fascinating given the issue of simulation. Basically, the fidelity of simulation largely boils down to resolving observation to direct observation of known/knowable detail vs. idealised economical representations that the observer has no way of disproving.