Algorithmic Cinema: On the Production of Oracular Artificial Intelligence

Abstract

This dissertation interrogates how generative models are transforming components of contemporary film aesthetics, labor, and spectatorship. I argue that algorithmic cinema is not adequately understood through the “good” or “bad” moral binary structuring creative AI ethics debates, but instead through the oracular. I define the oracular as an epistemic awareness of what cannot be known through computing. In the context of cinema, it orients AI as both an instrument of control and aesthetic medium for encountering epistemic limits. The oracular reframes what AI is, what it does, and how it might be otherwise in film and media industries. The dissertation begins with the “death of cinema” trope that paradoxically emerges alongside the medium’s birth in 1895. In Chapter 1, I trace this trope throughout film history up to so-called generative AI as the latest existential crisis and closely read an AI-infused film, annotating how aesthetic form registers anxiety. In Chapter 2, I situate this cultural anxiety within Hollywood labor histories and investigate the 2023 SAG-AFTRA and WGA strikes. I conduct participant observation on the picket line and interview workers, finding that they perceive managerial control of AI as an existential threat to the “soul” of cinema. In Chapter 3, I ground the oracular in these perceptions of mystique and genealogies of oracles as sites of computational unknowability. I then chart the AI image on a continuum with photographic and computer-generated images, identifying its spellbinding effects as a continuity and its predictive-of-the-past structure as a rupture. In Chapter 4, I show how an underlying epistemic fantasy of superhuman vision is what makes the oracular necessary. I place Soviet filmmaker Dziga Vertov’s kino-eye project as expressed in Man with a Movie Camera (1929) into dialogue with Cold War histories of computer vision. I conduct a humanistic experiment, in which I use a vision language model to interpret a sequence from the film by pausing at each interval between frames. The experiment transforms epistemic fantasy into awareness with the interval as a means of awakening to the limits of computational spectatorship, thereby enacting the oracular. The dissertation contributes the oracular as a portable concept and method that extends across creative disciplines. Cinema offers a privileged site for this analysis because technological transformation is a defining feature of the medium; early film theory anticipates contemporary questions of machine intelligence, mechanical reproduction, and synthetic reconstruction of material reality. By integrating historical and theoretical inquiry with discursive analysis, empirical fieldwork, and material experimentation, the methodology addresses AI as a transnational cultural project. Materially, it elucidates AI as a sociotechnical assemblage of data-intensive computational systems and semiotic processes trafficked through global networked labor infrastructures that automate prediction, classification, and content generation tasks. I situate these insights within the AI hype cycle: the rise and impending fall of generative models that compute image, text, and audio files by statistically predicting patterns from co-opted training data. Where the predictive-of-the-past structure exposes the uncomputability of media that exceed historical datasets, the spellbinding dimension remediates cinematic histories of magical illusion that cast an unknowable spell of affective ineffability. The oracular is thus generated through spellbinding media that is predictive of the past. It invites an awareness that algorithmic cinema is best understood not only through what it can compute, but what it cannot.