Their mathematics is a statistical model which outputs the statistically most likely result, in effect leaving only the most significant independent variables. Which decreases information complexity, by definition of "information complexity".
And more importantly, will it give the same response every single time?