The making of a digital twin:

Over the past 6+ months, after the initial idea of "Conflicts" came to be, my team and I have developed a custom pipeline for the creation of life-like AI NPC characters, based off of real-life personalities. Here's how it works:

"Conflicts" utilizes a bespoke conversational pipeline that allows two large language models to interact with one another inside of Unreal Engine. The “backend” of this system pulls information from a Google Form which either character in the simulation has previously filled in.

This form data indicates various character traits and information about the person as well as their opinions and views on the other person. Then, data is connected to a custom-made Metahuman avatar in Unreal Engine. Everything about the character is then generated based on the language output of the large language model (similar to how ChatGPT generates responses).

By having two NPCs communicate with each other, they enter a feedback loop of neverending conversation where they are faced with their conflicts in a perpetual dialog of attempting resolution.

Parag K. MITAL, Ph.D. is an internationally exhibited artist and interdisciplinary researcher working between computational arts and machine learning practices for nearly 20 years. His work examines the nature of perception in both humans and computers, often exploring the disjunct between the two, through mediums such as generative film, augmented reality, and large-scale archival interpretations. The balance between his scientific and arts practice allows both to reflect on each other: the science driving the theories, and the artwork re-defining the questions asked within the research.