ABSTRACT
With the advent of text-to-video generators such as Sora, Luma, and Runway, we are witnessing the growing popularity of tech-demo videos that showcase the potential, successes, and ongoing developments of generative AI. These demos, widely circulated on social media, highlight not only the distinctive aesthetics of the respective AI generators but also common flaws and errors, such as inconsistencies in objects or movement, bodily distortions, or uncanny transformations. However, it would be a mistake to interpret tech-demos that exhibit such flaws and inconsistencies merely as documentation of the current technical state, as lists of problems to be solved, or as purely informative components of technical reports. Instead, this chapter proposes viewing AI errors and failures as “promise machine[s]” (Appadurai and Alexander 2020 , 20) that generate promises and aesthetics of futurity: the promise of updates, of the next improved generation of AI, of photorealism, and of technical fixes for social problems. By examining tech-demo videos as “future media” (Ernst and Schröter 2021, 7), I focus particularly on the ambivalences of such an aesthetics of failure, which help reinforce AI's status as a “promising technology” (Hirsch-Kreinsen 2024, 1641) rather than undermining it.
