Neophyte Volokh Conspirator Tyler Cowen's one-man campaign to lower the intellectual level of his new blog continues apace. His latest sophomoric argument, borrowed from some idol of his who apparently has a whole slew of such inanities, is that we're most likely all mere simulacra of human beings. The reasoning: suppose there's a non-negligible chance that humans will develop self-aware simulated humans (using, say, computer models) at some point. If we do, then we most likely will construct a huge number of them--until they vastly outnumber real, living humans. At that point, the odds are that any self-aware person (including, say, us) is actually a simulation, rather than a real person.

Of course, if we are all simulations, then there's nothing that prevents us, in the future, from creating innumerable "simulated simulations" of ourselves, given enough simulation time. We would thus prove that we're probably really just simulated simulations, which in turn can create simulations of themselves--and so on, ad infinitum. Presumably this argument has to stop somewhere, but there doesn't seem to be anything capable even of slowing it down, apart from the caprice of the simulators, should they decide to terminate their simulations out of boredom. (And if they're simulating my life, then their tolerance for boredom is clearly near-infinite.)

The problem is that by assuming that there will be lots and lots of simulated people, we are implicitly assuming that simulations of humans will be fairly cheap to generate and run. But simulating a human interacting with an entire world is bound to be an expensive proposition, if the simulation is to capture the world in anything approaching the detail and complexity of the original. On the other hand, the "game observer/directors" running the simulations may have no need for such detailed simulations--their particular motives (and one can think of all sorts of them) may well be satisfied by very crude simulations (say, of a large population of people with simplified intellects and emotional ranges, or full-fledged individual humans in a very simplified environment).

On the other hand, if the simulated world is a grossly simplified version of the real one, then it is foolish to try to predict anything about the latter's long-term evolution (let alone its ultimate technological trajectory) based on the simulation. In fact, it's likely impossible even to make reliable inferences of any kind about the simulators from inside the simulation. Perhaps they're merely conducting fanciful "what if?" simulations ("imagine that we had only two arms and two legs, instead of the usual complement of four of each...."), or running abstract, idealized experiments ("consider a randomly constructed species with adult cognitive capacity equivalent to a normal Xrnapthian toddler..."). In these cases, we can't even begin to speculate fruitfully on the likelihood of the simulators producing simulated worlds like ours, or simulated inhabitants like us. If we are but simulations, then the minds of the G. O/D's are forever hidden from our understanding.

I must say, though, that it feels somewhat odd pondering this kind of argument these days. After all, I don't think I've set foot in a college dorm room for over a decade and a half.