Objective Functions: (In)Humanity And Inequity In Artificial Intelligence

It is hard to say which is more satisfying: the discovery of order in the material world (science), or the imposition of order on that world (technology). When human language takes material form, through digital computing and information storage devices, it seems even more satisfying to treat knowledge itself as the object of either science or craft. In Cambridge, we hedge our bets between these options, with departments of both Computer Science and Information Engineering. In this essay, I explore new methods of "machine learning" that have swept both departments, promising an information-theoretic revolution in artificial intelligence. I focus in particular on the subjectivities embedded in these mechanical systems, and the human satisfactions and ambitions in constructing them. Key questions include the relationship between order and information, common sense and classification, and the necessity of mathematical objective functions.