Quantum Computers Speed Up Classical with Probability Zero

Abstract

Let $f$ denote length preserving function on words. A classical algorithm can be considered as $T$ iterated applications of black box representing $f$, beginning with input word $x$ of length $n$. It is proved that if $T=O(2^{n/(7+e)}), e >0$, and $f$ is chosen randomly then with probability 1 every quantum computer requires not less than $T$ evaluations of $f$ to obtain the result of classical computation. It means that the set of classical algorithms admitting quantum speeding up has probability measure zero. The second result is that for arbitrary classical time complexity $T$ and $f$ chosen randomly with probability 1 every quantum simulation of classical computation requires at least $Ω(\sqrt {T})$ evaluations of $f$.