I continue to be fascinated and disturbed by Nick Bostrom’s paper from a decade ago: ARE YOU LIVING IN A COMPUTER SIMULATION?

The paper comes down to the following;

This paper argues that at least one of the following propositions is true:

(1) the human species is very likely to go extinct before reaching a “posthuman” stage;

We have to get to the point of vast technological improvement. I use the term technology broadly. Today, you might think of technology as an iPhone. A hundred years ago aircraft were technology. A hundred before that steam trains were the hot shit. So we really don’t have much of a clue what we will mean by technology in another hundred years.

Maybe we’ll get hit by an asteroid or something before reaching this amazing state.

(2) any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof);

This future human race must want to simulate human beings. As far as we know right now, it looks very possible. Around 2040 the computational power to run a brain will be available, as will (destructive) scans of the human brain. In fact, the work is being done right now. If we marry the scanning of a brain with the power to simulate it, you have simulated humans.

Of course, this assumes a bunch of things. It’s not fashionable to be positive about the future despite all the evidence, however, it’s reasonable to conclude computers will continue to get faster and scanning resolution will get better. Both have been doing so exponentially since the dawn of time. Then in 2042 we should be able to simulate two brains. 2044 will mean 4 brains…

Around 2110 we’ll be able to simulate more people than have existed. And, this is just a conservative estimate imagining rough doubling of computational power with Moores Law. The actual computational scope may pull up that curve much more quickly. Additionally, we can posit scanning your brain in the next few years and then simulating it in a few hundred years once the computational power is there.

(3) we are almost certainly living in a computer simulation.

Given that we didn’t die out and we do want to run simulations of the past, then any Tom, Dick or Harry will have the available resources to simulate entire planets of people. Actually, much more resource than that, but you get the idea. So, now there are maybe a hundred billion “real” humans but there are endless trillions of simulated humans simulated every few seconds. As a random human, statistically speaking you’re probably in a simulation.

The paper actually has some simple mathematics you can follow along to get the point, and expands on the ethics and a number of other angles on this hypothesis.

So now we have three propositions: Either we all die out (which would suck), we don’t want to run simulations (appears unlikely) or we’re all just in a simulation. Every day we don’t die out and want to run simulations makes it more likely we’re already in a simulation.

Now, I want to make a leap of my own.

Computer games gain efficiency by not simulating small things and big things. You run around shooting aliens, but you can’t take your gun apart or investigate arbitrary worlds in adjacent galaxies. It’s much easier to simulate the medium scale and forget the very small and very big.

Where does physics break down? The very small and the very large. Quantum Mechanics and Dark Matter. We should have found aliens by now (or very soon); we’re already at the point of imaging extra-solar planets and figuring out their atmospheres. Where is everybody? It’s an unfortunately empty universe we’ve seen thus far.

It seems likely that that the distant stars and galaxies are just poor simulations of what they were really like in 2013, given that we’re really in a computer in the year 3000, or whatever. Simulating all the aliens is just too problematic for the class of simulation we’re running in.

Similarly computer games simulate far away things coarsely but fill in detail as you approach them. A building a mile away is a crappy block representation. As you get closer, details pop in as your Xbox loads textures and makes it look more realistic. In Quantum Mechanics there’s a big mess of stuff which doesn’t resolve unless you measure it, plus a lot of random noise. It feels like this is all just simulated. It’s much cheaper to approximate sub-atomic reality as a random noisy soup, for the purposes of making humans think they’re in a real world.

Bostrom also points out it’s rare and hard to make sure physics at the small scale matches up with the large scale. That is, we can go look at arbitrary electrons in a ham sandwich and see what we expect. But doing that across macroscopic objects to make sure they’re all really made out of the physics we think they are is hard and painful.

The wikipedia page has some further historical perspectives.