Our perceptions are like an adversarial network -- we are always getting better at spotting the simulation, so simulations have to keep improving to try to fool our perceptions.