This brief essay distinguishes between learning algorithms and 'dumb' algorithms. It suggests that each can have added value as well as drawbacks, depending on how they are used for what purpose. The question of when to use what algorithms, however, is neither a purely technical question (if there is such a thing) nor a purely political one (if there were such a thing). Moving to law, the essay will draw the line between, on the one hand, legal certainty and, on the other hand, both arbitrary decision-making and rigid application of inflexible rules.