When did we start labelling selectively surprising observations as paradoxes?

It's definitely not at all a paradox that certain things are easier to code than other things...