# RE: Tough questions

O'Regan, Emlyn (Emlyn.ORegan@actew.com.au)
Tue, 7 Sep 1999 19:04:52 +1000

> No reason to think the physics would be anything like ours and no reason
> to
> think it's in anything close to real time and no reason to think that any
> part
> of our universe exists (is being simulated) when nobody is looking at it.
>
> John K Clark jonkc@att.net
>
Fair enough to points one & two. On the third...

If bits of our universe are not simulated when we are not looking at them, then how is consistency maintained? Is the simulation so coherently understood by it's owners that future states can be derived from past states without calculating all the intermediate states? I'm impressed. If that is the case (ie: that state(n+x) = f(state(n)) for x > 0), then how is our universe anything different from a lookup table? And in that case, why run it at all?

Or is it being fudged (this is I think the theory behind breaking reality - spike talked about this I think)? So that if a tree falls in a forest, and no one is there to see, then it doesn't make a sound, or have any coherent effect (it does not in fact even fall), until someone *does* come to look. Then, the sim quickly(ie: slowly, with execution suspended) calculates a probable & consistent state for the local region (eg: a forest with a tree which has fallen since it was last observed).

If the universe is in fact non-deterministic, then perhaps this is after all useful. Instead of calculating the one true state of the universe at all times (state(n)), localised regions are left unsimulated until a observed. Then, the sim determines a state through some heuristic (this means a big guess), and validates it against the current state of the rest of the universe, to ensure that it is in fact a consistent local state. Non-determinism gives you this, because there are multiple (infinitely many?) consistent local states, and using a heuristic plus validation could conceivably save work over simulating everything all the time, just in case.

If the universe is deterministic, then the heuristic must come up with the one and only possible state, or with a state which is plausible but not entirely correct. If it comes up with the one and only state, then again we have state(n+x) = f(state(n) for x > 0, ie: a lookup table. If the heuristic picks a state which is plausible but not correct (ie: looks good, but wouldn't stand up to detailed analysis at the subatomic level) then we have a chance of detecting (and possibly breaking) the sim.

On a different tack: What is the motivation of this simulation? There appear to be three classes of motivation.
1) The universe is a simulation, the outcome of which is not known, being run as an experiment to reveal some kind of knowledge to the experimenters. 2) The universe is an environment for simulated beings (us!), so the outcomes could be known by the experimenters, but there is a large motivation to fool us.
3) The universe simulation is a state table, everything known in advance to the designers, run for inexplicable reasons (religious? art?)

To expand on these:

1. The universe is a simulation, the outcome of which is not known, being run as an experiment to reveal some kind of knowledge to the experimenters.

If a tree falls in the forest, and there is no one there to hear it, then it makes a sound.

In this case, accuracy is everything. All of the universe is important, it is deterministic (or non-deterministic in our universe, but deterministic over a multiverse). We are inside a system which doesn't have any logical flaws. Our only chance to break out is if the basic physical laws of the universe themselves are not closed. This could be true in an experimental setting; if the point is to explore the consequences of iterating a rule set, then the experimenter may not know if the system is open or closed. But then, the system might crash (and we'd be stuffed big time!)

2) The universe is an environment for simulated beings (us!), so the outcomes could be known by the experimenters, but there is a large motivation to fool us.

If there's no one there to see it, then there's no forest.

The matrix? (I haven't seen it). The heuristic approximation theory (localised bullshitting) comes into play in this scenario. This does give us a chance to break the system, but only if the locality detecting systems are fairly course; if we try to set detection of a particular area, we might be detected as observers. The heuristics for picking a possible alternative local state in a non-deterministic universe could be skewed, producing quantum "decisions" which turn out to be non-random under scrutiny. In a deterministic universe, we could just find anomalies, which don't fit physical laws, and which can't be incorporated into those laws without creating special cases for each anomaly.

But then, how do we get out? Does our detecting the flaws signal to our captors that we've reached the required level of ability, and they conjure us out into the next level universe? Or do errors turn up in the system log, which prompts a clean re-boot of the system? Maybe the universe cracks open before our very eyes and we are hurled headlong into a paradise the likes of which we cannot dream? There's probably no way to know this without testing the theory.

Then we don't even have an illusion of free will left. We cannot surprise our captors (they have skipped to the end of the book, as it were). We can't do anything that they don't know about. All we can hope for is to be given autonomy after the fact for some inexplicable reason. Or just enjoy the universe as it is.

The conclusion? Well, there are good arguments to say that the universe is fully simulated, for experimental purposes, or just because they can, and there are good arguments to say that the universe is a big smoke & mirrors trick, with localised fudges, for zoological purposes, or if we are the experiment.

Say that in the universe outside ours, in which we are executing, the physical laws are the same as here. This does not imply 1:1 correspondence between time there and time here. So it also does not imply a physical architecture size. The universe sim might come down to a few simple inductive laws and an initial state.

We could be running on a programmable calculator (with a lot of memory used for state information). We'd just be running R E A L L Y S L O W L Y . . .

E M L Y N