I'm seeing a lot of big-shot scientist talk about the probability that we live inside a simulation. I'm just a guy on the Internet who writes too much code, but... that's actually very relevant. So hear me out.

With all due respect... just as any computer can emulate any other computer, you can simulate a universe in a universe. But the performance hit is huge!

Yes, you can simulate certain physical processes in better than real time, but that's because you're taking liberties with stuff that doesn't matter to the highly specific question you're asking. When you simulate the stability of a bridge, you don't need to know if over a million years, a new species of slug would evolve to feed on the particular variety of bird poop that lands on that bridge.

But for the kind of open-ended simulation we imagine the aliens running, one in which we might emerge to wonder if we live inside it... shortcuts won't do.

Emulating, say, the Atari 2600 video game console is a much better analogy than simulating a bridge. For all of the games to work in your emulator, you must simulate every last wacky quirk of the original hardware, right down to the precise timing, because the people who wrote those games exploited every inch of that. And so does natural selection. If physics has a quirk that makes something just a little easier, evolution has probably exploited it, somewhere, sometime.

So if you want to simulate a universe that contains human brains, or something equally complex, in real time then you'll need hardware that beats the daylights out of the original hardware. That works for emulating video game consoles because modern computers are vastly faster. But all the aliens have to work with is a universe like the one they want to simulate.

So the beings running the simulation would have to wait, oh, let's say 100 years to see the results of one year, and that's optimistic. And the simulation would consume far more resources than the real thing.

You can trade off the latter for the former somewhat by running the simulation in parallel, because things can't interact with each other instantaneously at a vast distance. But you're definitely going to pay a huge price in terms of time, resources or both. 

Yes, they could be immortal and have the ability to slow their own perceptions of time. But it would still be much less of a hassle to set up the conditions that interest them on a real planet and just watch.

One problem with my "just use a control Earth and an experimental Earth" option is that it wouldn't allow you to experiment with different physical laws. I suppose it's conceivable that our hella-powerful aliens really, really want to know what evolves when the speed of light is, for instance, 299,792,459 meters per second.

But do they want to know that badly enough to tie up the resources of an entire galactic spiral arm in order to simulate one solar system? I doubt it. I think they want to ask more interesting questions.

And if they do ask more interesting questions... and if we're living in their simulation... that means the aliens live in a universe with significantly different physical laws than ours. And yes, that means it could be a universe where simulating things is easy — for some reason that seems absurd to me but makes sense given their totally different physics.

But that effectively puts us in the realm of metaphysics. How is running on a simulation inside a computer in an unrecognizable, unknowable universe different, from our standpoint, from the universe being the product of an unfathomable supreme being or natural process?

Fear is one answer. Science fiction authors speculate about the aliens getting bored and pulling the plug on the simulation, the aliens running out of funding and pulling the plug, or the simulation being inaccurate and one day calling an undefined function and shutting itself down. 

But if their fundamental physical laws are unknowably different than ours, then the usual scary speculations are not relevant because we don't know if people and programs in their universe actually have such tendencies, or completely different tendencies, such as just getting more awesome over time. Loosely speaking, we don't know if the laws of thermodynamics apply, or apply backwards, or emit golf-club particles on alternate Thursdays. Maybe the computer and the program both arose spontaneously because their native universe is like that. Who knows?

This throws a monkey wrench in the already sketchy machinery of trying to estimate the odds we live in a simulation in a way that has meaningful implications for us. Either the "host universe" simulating us is similar to ours, which makes it extremely difficult and expensive to simulate us, to the point where they would almost certainly just set aside some planets to watch instead... or it is dissimilar from ours, in which case there is no way of guessing at the future of the simulation because none of our assumptions hold. 

So I think Neil deGrasse Tyson is wrong. Sort of.

But hey, I'm prepared to be corrected.

4/24 '16