A foundational theme in contemporary critical theory is the close relation between critique and crisis. This talk describes the introduction of computational technologies, and more specifically machine learning algorithms, into artistic practice as a historical crisis, which fundamentally puts into question the relation between making and understanding. From the standpoint of the artist, the question concerns the extent to which she understands the technologies that she is using. The artist often employs technical means whose internal mechanisms are obscure to her. But in art means and ends are essentially intertwined. Any opacity in the means extends to the ends for which they are used, and so potentially threatens the integrity of artistic agency. Similar concerns about the opacity of technology have been raised, outside the artworld, in the scientific community itself, giving rise to discussions about how to render machine learning interpretable or explainable. The prevalence of these discourses suggests that the obscurity in question is not a matter of individual ignorance. It pertains to the historical constitution of the technologies themselves in their essential character as formal systems, and to their formative role in what has been described as a society of hyper-control.