This essay examines the infrastructures and temporalities of modern AI technology based on artificial neural networks (ANN) and aims to contribute to a more substantial understanding of its political challenges. In order to unlock the different temporalities of ANN, a theoretical framework for the relationship of media and infrastructures is suggested that also might help to distinguish between the different levels of analysis related to specific steps and aspects of the machine learning process (the collection and production of learning data, the training of AI models etc.). An important reference point for the following considerations is ethnographic research conducted at TwentyBN,1 a Toronto and Berlin based AI company specialized in ANN and computer vision that just recently developed an app for the fitness market.