Abstract

The paper focuses on the questions of whether, to what extent, and in what ways the implications of the rapid development of artificial intelligence are changing the nature of one of the fundamental philosophical questions, “What does it (even) mean to understand?” It draws on two sources in particular: Hinton’s explanation of the technological development and functioning of deep neural networks and Nietzsche’s deconstruction of human understanding based on his key concept of “embodied errors.” In doing so, it reveals a series of unexpected parallels, relating in particular to the notion of micro- evolution and the function of error in the processes underlying “thinking” and “intelligence.” The paper therefore draws certain parallels and demarcation lines between human understanding and the “learning” procedures of digital neural networks. At the same time, it addresses the question of what it means for the interpretation of human understanding that, for the first time in history, understanding is faced with a real, existing antithesis, represented by intelligent systems which, although they do not understand, are capable of performing the tasks of understanding, and capable of replacing understanding.