The Post-Simulation Argument
With this year’s Loebner prize awarded to Elbot, artificial intelligence is one step closer to mimicking human behavior. While this achievement is quite astonishing on it’s own, it also makes certain people worry about our future. It’s not only the Terminator enthusiasts who are out shopping for shotguns and canned food. A large, and still growing, number of highly renowned scientists and philosophers agree that it’s very, very likely for us to be a mere manifestation of supreme artificial intelligence. In other words, that we’re all part of a humongous simulation.
Since Nick Bostrom’s publishing of the simulation argument in Philosophical Quarterly I’ve supported the theory. It’s an intriguing way of looking at reality, albeit a bit on the nihilistic side. The most important proof reality could give me was the astonishing resemblance between sub-atomic physics and 3D render engines. Just think about it for a second. A well written render engine will only render the 2D result of your current view port. All the other environmental data won’t even be processed. In the Copenhagen model of quantum mechanics the state of a particle is completely unknown until measured. Basically Schrödinger’s zombie cat paradox. Bostrom provided the mathematical evidence to support the possibility to simulate the human brain. With enough of these simulated brains in parallel you’ve got yourself a simulated society. He also pointed out, using basic probability calculations, that the probability of us being the unsimulated society is alarmingly low.
As a low-level software engineer I’m not comfortable with this idea. There’s a huge problem Bostrom has overlooked.
We’re moving towards a technological level on which we can successfully start to simulate the human brain. We’re also breeding like rabbits, adding more brains to the system. So in essence we’ve got two processes adding brains to the overall system; the natural process and the synthetic process. The natural process is rather slow. It’s a low exponential increase with certain environmental limitations, further diminishing the growth. The synthetic growth is truly exponential and will rise rapidly.
When I look at the simulation argument from a low-level hacker’s point of view, a capacity problem arises. And it’s a big one. The natural growth will behave like a parallel recursive function. It’ll slowly, but at an increasing rate, consume all available resources. The simulation program will have to spawn new brain simulation processes in order to keep up with the growth. From the moment these simulated brains start producing their own simulation of themselves the recursion turns into recursive parallel recursion. The exponent of the natural exponential growth will grow exponentially. These programming errors will bring any finite system to it’s knees within moments. I usually refer to this event as the critical state of a recursive system. Logic suggests that the critical state of our universe has to be lower than that of the host on which we are simulated. The critical state isn’t so much a computational boundary, as it’s an effect of the finite amount of storage for the state itself.
Either of two things will happen if we successfully start to simulate our own brains. The first possibility is that nothing happens, we are simply not simulated. The second possibility is a little more complicated. Since the capacity of the machine is not limited by the capacity of it’s creator’s brain, the machine is likely to have a higher capacity than it’s creator. It will be able to simulate a better brain which will reach the point of building a simulation of an even better brain sooner. This recursive evolution is only limited by the critical state of the original machine. The simulation, on any level of the recursion, may hit it’s critical state at any point in time. The inhabitant of that level can not be aware of the critical state of his level, or any of the recursive levels prior to him. See, with our exponential growth of capacity, both natural and synthetic, we will hit our critical state. If we are simulated, that is.
So in short we’re pretty much doomed as a simulation. Let’s just hope we’re not.
In lack of a better place to put this: You’ve been tagged