When Runway's CEO pitched AI-enabled volume filmmaking at a recent industry event, the tech press framed it as disruption. But the logic is older than silicon: flood the zone, let the hits surface. The question is what gets lost when abundance replaces scarcity as the dominant creative mode.
Volume, Creativity, and What AI Agents Actually Do
A 2026 paper on arXiv by Franceschelli and Musolesi, "On the Creativity of AI Agents," argues that LLM-based systems demonstrate human-comparable creative behaviors in generative tasks but systematically underperform in evaluative ones. Meaning: AI can make 50 films. It cannot reliably tell you which one is good. That gap is not a bug. It is the entire aesthetic challenge of the volume-production model. Studios betting on AI abundance are essentially outsourcing the filter to the audience, which is either democratic or chaotic depending on your taste politics.
From Hollywood to the Gallery: The Bennett Prize Problem
Contrast this with the Bennett Prize, which awards a single $75,000 grant and a traveling solo exhibition to one women figurative realist painter. One. The scarcity model, maximally applied. The cultural logic running underneath both structures is identical: someone has to decide what counts. In the AI film economy, that someone is the algorithm and the opening weekend. In the prize economy, it is a jury. Neither is neutral. TurboFund's breakdown of investor research mistakes applies here too: backing volume over signal is a classic error whether you are funding films or startups.
What Runway is actually proposing is not democratization. It is the industrialization of the long tail. The arXiv paper on exploration versus exploitation in LM agents maps the same tension onto AI decision-making: systems that explore too broadly fail to exploit genuine quality signals. Hollywood's blockbuster monoculture was the over-exploitation end of that curve. Fifty AI films per budget cycle might be the over-exploration end. The productive zone, as usual, is somewhere between.