To some extent this is feasible, but we the exponential time issue is more of a problem than one immediately thinks. Suppose you want to simulate some region for one second with a level of detail down to 1/N the size of the region. So the amount of time it takes to do this is going to be some function of N. The problem is that the time isn’t going to vary like, say, 100 * N, or like N^3. It’s going to vary exponentially, that is, like 10^N.

And exponential computation times are kind of hideous. Suppose N is 100. Then you’re looking at the mighty googol, that is, 10^100.

And the computational work might also be exponential in the length of the time that you want to simulate as well.

So it seems kind of unfeasible, even for SF. This said, I can vaguely imagine a somewhat comic or satiric story involving the problem. A galaxy-spanning civilization works flat out for a trillion years just to simulate some guy eating a hard-boiled egg. “The Egg.”

There’s no hurry when you’re a simulation … you don’t experience the time slicing.

“The rate at which a device can simulate an equivalent volume of reality”

A cube of sea-water would have a value of exactly one – it can simulate itself in real-time, but no more.

A modern intel CPU can barely simulate a handful of atoms. If the CPU is made up of 10^20 atoms and it can simulate 100 atoms (10^2), then it would have a value of 1 / 10^18. An alternative way of expressing this would be say that it has a value of “18″, and the scale is inverse exponential.

So then a hypothetical future quantum computer that can see into the future would simulate reality 100x faster than real-time – it would have a value of 1 / 0.01 or 1 / 10^-2 or just “-2″.

Here’s a rather simplistic hypothetical: what if we discovered a way to simulate atomic structure using subatomic particles? Material substance, when reconfigured, could then express an exponentially larger amount of pseudo-atomic information than its atomic predecessor. It would then seem an imperative (amorally speaking) to reconstruct the universe to this more optimal state.

And couldn’t we call this reconfigured substance, say… computronium?

And maybe we should add beautiful as well

I think that summarises your point. Get out more people. And great pictures Rudy, they really make the point.

You said “The notion of leaving the details up to randomness is an interesting move. But maybe they aren’t random”

But the entire point is that for a simulation _it doesn’t matter_. You won’t know the difference. It only matters from a philosophical goal of a “perfect simulation”. But what is the value in a perfect simulation vs. one that is “good enough” that you can’t tell it apart from the real thing?

If I present you with 1000 pictures of snowflakes, you will be completely unable to tell me how many of them I’ve reused multiple times unless you were allowed to zoom in to a great level of detail and study them for a long time. _Unless you zoom_ (use a microscope, or whatever) there’s no reason to take the computational cost of making a vast number of different versions.

I’ve seen human game players be blatantly fooled by completely trivial tricks in that respect into thinking they are dealing with a far richer environment than they are. Human brains have evolved into highly sophisticated pattern matching machines where it is safer to find a pattern where there is none (prey moving about rather than the wind rustling the trees) than miss a pattern where there is one (the prey gets away because you don’t see it, and you go hungry). As a result we are VERY easy to fool.

We frequently “see” things that are not there because pattern matching kicks in. Our memory is woefully incomplete (Try a simple experiment: draw your shoes without looking at one, then put them in front of you and draw them again. Most people, regardless of skill level, will draw an “idealized” version of their shoes the first time, lacking a lot of details not because they can’t draw them, but because their memory of it only contains the important features) Our reading comprehension is terrible – just try to have a group of five people read a paragraph and then summarize it 30 seconds later and see how different they are (and how none of them are likely to match it completely, unless they’re told in advance they’ll have to memorize or summarize it)

In short, we’re lossy machines to start with, designed to fill in details based on similarity and experience rather than reality. Present us with a lossy world, and, as long as it takes the “right” shortcuts, we won’t be able to tell the difference.

We’ll assume that the leaves on the ground have been blown around by the wind if we even notice they’ve shifted somewhat. If we raked them together in a nice pile we’ll notice if the pile is gone completely, but not if it’s subtly changed because the location of individual leaves was picked randomly when you returned.

We can fool a human in this way _without_ access to thought processes, but a simulation of the world including our brains would be able to tell exactly what we remember of what we’ve seen and experienced, and can selectively cull details our brains deemed unimportant rather than make guesses. Fooling us when you can check the simulation of our brain and adjust the randomized model until it is consistent with expectations would be trivial compared to doing the simulation in the first place.

In other words, a simulation can make changes we _can’t_ catch out unless we actively try to. Even then it could make it extremely hard by limiting changes possible changes that had a high enough probability of happening naturally.

We might even be able to make it impossible to catch. Worst case, if we found it acceptable, the simulation could alter our neural response to match the changes, but even to me that’s one step too far.

Yes, I hold a fairly mechanistic view of nature. I do believe we’re all a bunch of machines. Of course, we don’t _know_. It’s possible that consciousness can’t be captured by a simulation. But until we have something that indicates otherwise, that’s my basic assumption, and that boils it down to making the inputs match at the right stage.

I get the feeling from what you write that your starting point is very different, and that that is the basis for bringing up a “complete” simulation.

To me, a complete simulation is wasteful, and waste is terribly inelegant when I could instead use those cycles to “add features” – I’d rather have a virtual reality that augments what we have today than one that aims to perfectly duplicate the real thing. If it’s just a duplicate, then what’s the point?

For a virtual reality to make sense, it would have to offer vastly more than “real” reality.

(While we’re on this subject, btw., whether you agree with the starting assumptions or not, everyone should read this: http://www.simulation-argument.com/)

N. G. Chernyshevsky (Н. Г. Черныше́вский) wrote about art and Art and Reality back in the 1800′s, and argued that a painting of an apple isn’t an apple. Magritte’s artwork emphasized this theme in his famous “This is not a pipe” painting and others. Replacing Art with VR can get very similar discussions.

While the currently impossible process of recreating an exact duplicate of a living thing (Star Trek transporters–still impossible today, right?) seems exciting, I also think that there are more possibilities in exploring a VR that isn’t an exact replica. After all, we already know the real world. Where’s the fun in that?

I was pondering duplicated environments and Lifeboxes this week when I got my new computer. My initial thought was to exactly duplicate my old environment on the new machine, but then I had a change of heart. I brought over the files I currently use, left the old stuff on the old computer, and now I’ve got two similar environment that are changing in their own directions. It would be nice to have them both on one machine (as Virtual Machines running on a VMM) so that I could watch them both over time, but I’ll need a bigger machine still for that to happen. How interesting it would be to have two Lifeboxes with slightly different threads that I could maintain…or three, or more.

Anyway, thanks for the VR posts, Rudy–it made me exercise my brain the way it should be exercised.

Scott