The Innocence of Unknowing is an experimental AI-driven animated documentary that interrogates how mass shootings are mediated, remembered, and normalized through news imagery. Developed by Ryat Yezbick and Milo Talwani, the project uses computer vision and Gaussian splat animation to reprocess archival footage of U.S. mass shootings, transforming broadcast news into immersive, computationally re-rendered environments that expose the aesthetics of power embedded within them.
At its core, the work asks whether artificial intelligence, a flawed tool devised by techno-capitalists, can be retrained to see through the lens of the humanities. A custom AI agent analyzes news archives using computer vision, generating metadata about police weapons, uniforms, and formations, as well as fleeing victims’ gestures and postures such as arms raised overhead. Rather than identifying perpetrators, the system focuses on the spatial and political relationships between civilians, police, and media infrastructures. The resulting analysis becomes the foundation for a series of animated reconstructions built with Gaussian splatting techniques, which convert 2D news footage into volumetric environments.
Gaussian splat animation enables the project to suspend moments of crisis in unstable three-dimensional space. Victims fleeing with hands raised, officers corralling bodies, aerial “god trick” vantage points—these scenes are rendered as shimmering, partial forms, composed of millions of luminous data points. The instability of the splats, created through motion, mirrors the fluidity of truth and fact finding in real-time media coverage while reflecting the emotional, interpretive reality of news consumers. Viewers are moved through these environments, inhabiting perspectives that reveal how camera angles, police choreographies, and broadcast framing shape public perception.
By animating these archives through computational processes, The Innocence of Unknowing foregrounds the constructed nature of both media spectacle and algorithmic interpretation.

Screenwriter: Ryat Yezbick, Milo Talwani, AI Humanities Scholar
Project creator: Ryat Yezbick, Milo Talwani
Director: Ryat Yezbick
Technical director: Milo Talwani
Original concept and archive: Ryat Yezbick
Producer: Ryat Yezbick 
Editor: Ryat Yezbick, Milo Talwani, Cary Cronenwett, Linda Franke
Composer: James Rushford 
Sound mixing: Joe Talia 
Cast: Milo Talwani, Ryat Yezbick, Aurora (AI Humanities Scholar)

With support from: Massachusetts Institute of Technology (MIT) Open Documentary Lab; Onassis ONX Studios; Headlands Center for the Arts; Arizona State University

Ryat Yezbick (US) is a visual artist who explores the body, memory, and the politics of witnessing in the era of digital surveillance and decentralized global conflict. They are a published author, multi-time grant recipient, Onassis ONX Studio Fellow, former MIT Open Documentary Lab Fellow, and Faculty member at Arizona State University’s Narrative and Emerging Media Program in Los Angeles. 
Milo Talwani (US) is a filmmaker, composer, and creative technologist based in Los Angeles. Recent projects include Dog Movie, directed by Henry Hanson, and orchestration work for the French premiere of Ted Hearne's oratorio, Place. Milo previously served as associate programmer for the New Frontier section of the Sundance Film Festival.

What is an AI capable of when it has been thoroughly trained in critical theory? Ryat Yezbick and Milo Talwani tested this capacity by engaging in a dialogue with AI system “Aurora” about US mass shootings and their media portrayal. This live discussion analyzing and critically assessing footage of victims, perpetrators, and law enforcement resulted in a highly engaging feature-length animated documentary. The film addresses, not least, the social biases and clichés manifested within the AI and its training data. The artists diligently confront Aurora with its rash conclusions and errors, prompting it to revise its assessments. Yet, one thing eludes Aurora: the admission, “I don’t know.”