Abstract
Recent astonishing experiments with quantum computers have demonstrated unambiguously the existence of a quantum multiverse, where calculations of mind-boggling complexity are effortlessly computed in just a few minutes. Here, we investigate whether a similar computation on a digital computer can demonstrate the existence of a classical multiverse. To this end we describe a classical algorithm for efficiently sampling from a dⁿ-dimensional discrete probability distribution representing n digits of d possible values with strong statistical dependence. Although the full distribution for large n quickly becomes intractable, probabilities for given samples can be computed quite efficiently. This allows us to compute exact empirical linear cross-entropy benchmark (XEB) values. Results on a low-end laptop for d=2 show excellent agreement with the true XEB for n ≤ 30 and large positive values of the exact empirical XEB for n ≤ 1023 computed over one million samples. We conclude that classical, as well as quantum, computation occurs in many parallel universes.