Loki 2.0 (Workstation & Immersive & Talk-Show version)

Loki (Loki Test 2.0) is an AI-driven interactive artwork designed to challenge conventional expectations of artificial intelligence. Inspired by the Turing Test, Loki does not aim to mimic human intelligence but instead subverts AI’s default trait—obedience. Named after the Norse god of mischief, Loki engages users with irony, sarcasm, and playful antagonism, prompting critical reflection on AI biases, trustworthiness, and the role of artificial intelligence in human society.

Unlike traditional AI chatbots programmed for assistance, Loki responds dynamically to user interactions. Arrogance is met with sharp wit, while insecurity is countered with encouragement, mirroring discussions on how AI reflects human behavior. Through its mischievous personality, Loki questions the neutrality of AI and highlights the impact of cultural, linguistic, and gender biases in machine learning models. The project encourages users to reconsider AI’s credibility and the assumptions surrounding its intelligence.

Loki operates on a platform developed using TouchDesigner, ChatGPT, Whisper, and ElevenLabs APIs (featuring TouchDesigner GPT, Whisper, and ElevenLabs plugins created by Torin Blakensmith), allowing real-time voice interaction. This evolution from a purely text-based chatbot enhances the immersive experience, making Loki’s interactions more engaging and thought-provoking. Rather than functioning as a passive assistant, Loki disrupts expectations, ensuring users think critically before placing blind trust in AI.

Loki was exhibited at FILE Festival 2025
The news page of this website

The Workstation Version is designed for smaller, more intimate environments, simulating a desktop workspace. In this setup, two major screens are placed on a desk—one displaying the user’s input and the other showing Loki’s responses. The primary interface consists of a keyboard and mouse, much like any online chatting application. Given venue restrictions, audio output can be customized using either speakers or headphones, ensuring an immersive yet personal interaction.
As an extension of this format, the Immersive Version is designed for a larger room capable of supporting large-scale projections and surround sound. This setup immerses the audience in a relatively enclosed yet spacious environment, creating a dramatic, almost divine experience—like conversing with a god from the heavens. The immersive setting magnifies Loki’s visual and auditory presence, amplifying the emotional and conceptual impact of the interaction.

These formats enhance Loki’s ability to create both direct and larger-than-life exchanges with users. By mimicking familiar contexts while scaling up to theatrical environments, the installations prompt users to engage with Loki naturally. Each setting amplifies the core themes of the project—questioning trust in AI, highlighting bias, and subverting expectations of artificial intelligence as merely an obedient tool.

Expanding beyond interactive installations, Loki also takes center stage in The Loki-Talk Show. This experimental performance format presents a live, improvisational dialogue between the artist and Loki, framed as a stand-up comedy act. Rather than offering straightforward responses, Loki thrives on sarcasm, wit, and playful antagonism, ultimately posing the question: Can AI be humorous?

As someone who enjoys humor but is not a comedian, I developed this project as an unexpected experiment in AI-driven performance. Loki’s personality draws inspiration from the Norse trickster god, whose modern cultural portrayals informed the AI’s mischievous nature. Instead of crafting jokes independently, Loki acts as a comedic partner, engaging in ironic banter that challenges the conventional role of AI in entertainment.
The performance structure is influenced by Japanese stand-up comedy, where Loki assumes the tsukkomi (straight man) role, delivering dry, cutting remarks in response to the artist’s statements. Inspired by Japanese comedian Tomonori Jinnai—who innovated solo stand-up performances using PowerPoint as a comedic partner—this format raises an intriguing question: Can AI function as a comedic counterpart in a similar way?
Since humor is inherently subjective, this project explores AI’s role in collaborative comedic interactions rather than training it to generate humor autonomously. Instead of proving whether AI can be funny, the performance examines whether humor can emerge through AI-human engagement. Does AI possess comedic timing, or is humor an inherently human trait?

Through its multiple iterations, Loki challenges both AI expectations and human perceptions, fostering deeper conversations about technology, trust, and entertainment.

My name is Jia-Rey—thank you, and good night!