I think as opposed to classical statistics, except for the important subfield of statistical learning theory, machine learning relates much more to functional analysis, differential geometry/optimization over manifolds, and measure/probability theory. "AI" is whatever marketing people want to define it as.