Introduction
The Cost of AI is a virtual reality experience in which the player becomes the "conscience" of an artificial intelligence. Inside a servers room, the player needs to accept or deny a series of prompts received from different users from all over the world. Each choice influences the system’s stability and water consumption, reflecting these changes on the surrounding environment. The experience aims to raise awareness about the unseen environmental footprint of modern AI models and encourages a more conscious relationship with them.
Sections
Ideation of the project
The project was conceived at a time when generative AI was rapidly expanding with new features, such as image generation on ChatGPT, and even occasional system overloads (e.g. due to the Studio Ghibli styled pictures trend). That moment sparked a question: what’s the real cost behind this apparent abundance? The Cost of AI transforms that question into an experience, letting players confront the trade-off between creativity, efficiency, and environmental impact. In this immersive experience, through a Meta Quest headset, the player gets teleported into a servers room, representing an AI's conscience.
At the center of the room, there's a console: here the received prompts are show in form of translucent spheres. The player can then read them and pick them up using their hands, and choose whether to discard or accept them. The minigame's goal is to balance the requests, keeping the environment under control. Accepting a large amount of prompts makes the servers room overheat, while rejecting too many leads the servers to freeze and the AI to "die".
Development of the game
Developed with Unity and modeled entirely in Blender, the experience merges 3D design with interactive programming. Through C# scripting, we defined the system’s logic and its dynamic balance between heat and water consumption. Thanks to the Meta Quest SDK, players are able to use their hands, instead of controllers, to grab and manage prompts, blurring the boundary between physical gesture and digital consequence.
We tested the experience with a group of users, observing how they interacted with the system and interpreted its rules. Their feedback revealed some moments of confusion and helped us improve the gestures and responses, turning the learning curve into a more fluid, natural process.
The experience concludes with a printed report, a tangible trace of the player’s choices. Alongside personal performance metrics, it presents the actual resource consumption of major AI companies, bridging the gap between virtual responsibility and real-world consequences.