ABSTRACT
In this article I investigate a decentralized infrastructure meant to assist autonomous vehicles in making decisions. More akin to a call centre than a centralized control room, Nissan’s ‘Seamless Autonomous Mobility’ (SAM) project imagines that remote ‘mobility managers’ will intervene in the decision-making of autonomous vehicles, with the assistance of live video streams and other sensor data. Different from other kinds of AI microwork in which human workers prepare, imitate, or verify AI, mobility managers are envisioned instead as ‘interveners’, meant to directly and actively intervene in the movements of ‘autonomous’ vehicles when unable to negotiate an obstacle. Firstly, through a comparison between SAM and a traffic management system in Los Angeles, I argue that the former ‘normalizes’ intervention, in which decision-making delays become ordinary, if not altogether desirable. Secondly, through an analysis of a video in which such normalized interventions are imagined, I consider how SAM offers a kind of speculative mundanity in which remote workers, enabled by a technological infrastructure, embody a novel logic that modifies the social settings of driving.
Acknowledgments
Thanks to members of Locating Media for feedback on early versions of this article. Thanks also to Florian Sprenger for feedback on later versions, and to both reviewers for their insightful comments. Any mistakes herein are my own.
Disclosure statement
No financial interest or benefit has arisen from the direct implications of this research.
Notes
1. Acting also as legal buffers, as in the case of Rafael Vasquez, a ‘Vehicle Operator’ (VO) for Uber ATG who was charged with negligent homicide following a fatal crash, demonstrates (Porter Citation2020).
2. The full video has since been removed by Nissan, but an identical version can still be found online (Nissan Citation2017b). B-roll footage of the interviews and mobility manager visuals are also still available, from which some of the figures in the article derive (Nissan Citation2019).
3. This is often referred to as an ‘occupancy set’ within robotics research, see Pek et al. (Citation2020).