Stephen Wolfram has a model of the universe, the Ruliad, in which every set of possible rules for the laws of physics is operating in parallel. His motivating hypothesis is the Principle of Computational Equivalence, which states that no matter how a computer is built, whether it be a regular computer like a Macbook or an implicit computer running via cellular automaton, it is capable of computing the same things.

Jorge Luis Borges wrote an excellent short story called The Library of Babel. It describes an infinite library, containing all possible 410-page books. The vast majority are completely worthless nonsense, but somewhere within its corridors, every possible coherent book exists.

Perhaps these ideas can be combined. Consider the Datacenter of Babel. Room after room is full of racks of computers. Every possible codebase is running somewhere on one of its servers. Somewhere there lies a Minecraft server where all the cows are modded to look blue. Somewhere there lies a Bitcoin miner containing Satoshi’s private keys. Somewhere, perhaps, there lies a computer slowly emulating a perfect model of our own universe.

Is the world a simulation? Is the Datacenter of Babel real? In some sense, if the universe can be modeled by a computer at all, then the Datacenter of Babel must be an accurate model of the universe.

Is this a useful model? Imagine strolling through the Datacenter of Babel, connecting your laptop at random to various servers. The vast majority, of course, are running nonsensical code, full of syntax errors, crashing the very moment the machine was turned on. Without a map, you could hike for years without ever encountering a functional server.

Indeed, we have no real map to the space of all possible programs. Wolfram theorizes that it should be possible to search the space of short programs to find interesting ones. To me, it seems like we can create a number of neat-looking pictures, but we have never found anything useful by searching programs. Once you realize that the Datacenter of Babel is not a useful model, you also realize that the Ruliad is not a useful model.

The underlying problem with the Principle of Computational Equivalence is ignoring the efficiency of programs. There is a long history of this - the traditional theoretical definition of “computable” functions ignores how long it takes to compute them - but in practice, something that takes 10^100 time might as well be impossible. The Datacenter of Babel might be a correct model, but it is an inefficient model, in terms of how efficient it is to even locate ourselves within the model, much less make any conclusion from it.

I am entertained by the concept, though, of an infinite datacenter. I wonder how one would set up the routers….