An obvious choice would be ridiculously large Neural Networks. However, any problem where you start with a random seed and converge on a solution benefits from starting at several different random seeds to avoid local minima. Also search algorithms (chess programs etc) often paralyze vary well.