This study examines the Machine Visions works of visual artist Trevor Paglen (1974-), who is based in the United States and Germany. Since the beginning of his artistic practice in 2006, Paglen has focused on the invisibility of regimes of secrecy and...
This study examines the Machine Visions works of visual artist Trevor Paglen (1974-), who is based in the United States and Germany. Since the beginning of his artistic practice in 2006, Paglen has focused on the invisibility of regimes of secrecy and surveillance, as well as of the contemporary technologies that reinforce these systems. Developed since 2016, the Machine Vision works consist of photographs, videos, and installations that dissect the invisible systems of computer vision—the visual field of artificial intelligence—and critically visualize the images produced by these systems. This study critically analyzes the relationship between computer vision and images, and investigates how Paglen visualizes this relationship through his Machine Visions works.
Prior to analyzing Machine Visions works, this study examines Paglen's early and mid-career works that visualize the physical and technological objects that constitute regimes of secrecy and surveillance. Particular attention is paid to the technological critiques that emerge in these works. First, through an analysis of his landscape photography series, the paper examines the complex characteristics of the camera, which appears as a subject while simultaneously visualizing the invisibility of secrecy and surveillance systems. Next, the study examines the undersea cable series, focusing on how these works reveal the illusion that network technologies are immaterial. Finally, the study analyzes the Impossible Objects series, focusing on the possibility of alternative technologies distinct from those developed for state and corporate secrecy and surveillance systems. Through these analyses, the study demonstrates that Paglen's critical perspective on contemporary technology, articulated in relation to social, political, and historical contexts, was established from his early practice. In addition, invisibility is identified as a key concept that cuts across both his early works and the Machine Visions works, manifesting not only in regimes of secrecy and surveillance but also in the technologies that constitute their core.
The study then examines the mechanisms of computer vision and the resulting changes in the characteristics of images from three perspectives. First, drawing on the concept of the operational image articulated by filmmaker Harun Farocki (1944–2014), it analyzes the characteristics of images produced by automated machines and considers how Paglen reinterprets operational images through computer vision. This analysis reveals that as computer vision operates in an invisible domain, images themselves also become invisible. Second, the study analyzes the problems of image datasets used to train computer vision systems, focusing on the relationship between platform capitalism, a digital platform-centered economic system, and ImageNet, an early large-scale image database for datasets. It further examines the research and projects that Paglen conducted in collaboration with AI researcher Kate Crawford (1974-) to critique ImageNet. Third, this study examines the image generation models of generative AI, which have developed in earnest around 2014, and the problems with the images they generate. Specifically, the study examines the mechanisms of generative AI, which differ from those of previous technological media, and critically analyzes the characteristics of images transformed by these processes.
On the basis of this theoretical framework, the study examines how Paglen visualizes the invisible mechanisms of computer vision and the problems with images transformed by these processes in his Machine Visions works. The study closely analyzes the Machine Visions works by dividing them into three categories: works that visualize the mechanisms of facial recognition technologies; works that dissect the system of datasets and the images that constitute their components; and works that critically visualize the mechanisms and images of generative AI. In conclusion, the study demonstrates that the Machine Visions works critically visualize not only the invisible mechanisms of computer vision, but also the social and political contexts embedded in the images that the system learns from and generates.
This study’s primary contribution lies in providing a comprehensive analysis of Paglen’s Machine Visions works, which have gained attention amid the growing discourse on artificial intelligence but have so far been addressed only fragmentarily. In addition, the study makes a secondary contribution by offering a timely discussion of computer vision as an emerging technological medium and the changing characteristics of images it produces.