Wired: Gosiwon is a 3D animated VR film that portrays three characters living in Gosiwon, which are low-cost, single-room accommodations in South Korea that are often susceptible to fires. Wired figures are used to symbolize the entangled and fragile nature of individuals in contemporary life and to question the complex realities and substances emerging from emptiness.
The film integrates generative AI for textures and machine learning systems for sound and background music as assets that belong in a virtual world. By leveraging AI’s imperfections, the film reveals the contradictions and complexities of data within the physical and digital realities that we inhabit. The current version is a 360-degree cinematic experience, and there will be different versions incorporating real-time navigation and interaction.
Co-creation with Artificial Intelligence (AI)
In this work, machine learning (ML) for sound was implemented by Mike Lukaszuk (Sound & Music Director). This approach relied on classification, in which sound recordings were continuously fed into a system that returned enhanced or degraded versions of the audio files based on how accurately sound sources such as musical instruments could be detected. ML was also applied to generate a collection of sounds using regression algorithms on the software parameters of a computer-generated piano, allowing for a large number of variations from a small amount of training data. This process results in a series of sounds that each relate to the training data in different and unexpected ways, ranging from characteristically pianistic moments to abstract electronic drones. With both classification and regression, we emphasize that the impression of failure allows the audience to discover hidden aspects of the sound and experience new variations on familiar sounds.
In addition, generative AI was utilized during pre-production to design characters made of wires and to generate 2D textures for the virtual environment. Text-to-image and image-to-image AI techniques were employed to create uncanny and imperfect images that evoke a tactile yet strange sensation for familiar materials and objects. By blending these AI-generated textures with real-world captured textures, the work intentionally blurs the boundary between AI-generated imagery and actual physical textures, provoking a sense of confusion about the distinction between virtually simulated and embodied physical realities.
Wired: Gosiwon won Second Prize in the VR Section at Digital Arts Zurich 2024, the largest festival for digital culture in Switzerland. The work is also officially selected for the Awareness Festival 2024 in Los Angeles and the Festival of International Virtual & Augmented Reality Stories (FIVARS) 2024 in Toronto.
This work is part of the research project called Meta-metaverse: digital art-based research on reflective approaches to the metaverse, funded by the Social Sciences and Humanities Research Council (SSHRC) and supported by the Department of Film and Media at Queen’s University.