In this paper, I will investigate the possible impact of weak artificial intelligence (more specifically, I will concentrate on deep learning) on human capability of action. For this goal, I will first address Arendt’s philosophy of action, which seeks to emphasize the distinguishing elements of action that set it apart from other forms of human activity. According to Arendt, action should be conceived as praxis, an activity that has its goal in its own very performance. The authentic meaning of action (...) includes the “passion” for articulation of one’s own individuality; I can only manifest myself as a distinct personality insofar as I introduce myself as a novel beginning to the web of human interactions demonstrating both my relevance and distinction from others. From this Arendtian standpoint, I will analyse the impact of deep learning in modern AI from two possible angles. First, I will argue that the direct interaction between AI and action is impossible. Since AI operates on the principle of efficiency, it can neither suggest certain goals for action for us nor overtake their implementation because action is not guided by the instrumental need to be efficient but by the existential desire to be someone. Second, I will also analyse the possibility of the indirect impact of AI on action. More specifically, I analyse neural network’s ability to circulate actions among individuals based on mathematical calculation. As I will argue, the efficiency of this circulation that surpasses human cognitive capacities can potentially organize a broader network of interaction among individuals and serve as a catalyst for the ability to act. (shrink)