THE FOUNTAIN (2024)
The Fountain is an interactive installation designed to provoke contemplation on the interplay between two essential resources: water and data. At a time when water scarcity is a growing concern, and the demand for data and processing power continues to surge with the rise of the internet and AI, our relationship with these resources becomes increasingly complex. This installation challenges participants to explore the connection between them.
While water may seem unrelated to the production of AI, large server farms, essential for storing and processing vast amounts of data, often rely on water evaporation cooling mechanisms to function. AI server cooling consumes on average of 9 liters of water per kWh of energy used. With an average of 0.003 kWh usage to generate an AI image, we use roughly 27 millimeters of water per AI image generated.
As participants approach The Fountain they can choose to either drink water from the fountain or pour the water into the server which will trigger the generation of AI images of nature conveying a distopian connection to the natural world. Through the experience AI generated sounds of trees crackling and AI modeled sounds of nature can also be heard.
Through this interactive experience, participants are prompted to reconsider their perceptions of resource consumption, sustainability, and the intricate balance between human needs and technological advancements. As they witness the direct correlation between their actions and the functioning of the AI generation, they are encouraged to reflect on the value of AI in the context of limited resources, particularly in the face of water scarcity.
With ComfyUImodel we created over 80 distinct tree designs and more than 100 varieties of shrubs and flowers. With the use of different prompts describing plants, we produced unique plants, resulting in an extensive catalog of digital flora. Once the images of the plants were generated, we utilized TouchDesigner to integrate them into a dynamic particle system. This system animates the plants, creating the illusion of an evolving, thriving digital ecosystem. By pouring water in the sink the system arranges the AI-generated plants in a random sequence each time, constructing a new natural landscape with every iteration. This ensures that no two presentations are the same, allowing for a unique visual experience in every instance.
To complement the evolving digital flora, we developed an AI-generated soundscape rooted in the natural acoustic signatures of trees. Researching how trees emit creaking sounds when under environmental stress, we trained a variational autoencoder on recordings of tree creaks, wind-swept branches, and other organic resonances. Additionally, we incorporated a second model trained on a diverse range of natural soundscapes, from rustling leaves to distant bird calls, enriching the sonic environment. These AI-generated sounds were then spatialized and dynamically mapped to each plant, ensuring that every tree and shrub had a distinct auditory presence.
Concept + Creative Direction + Sound Design + Industrial Design: Sandy Davis
Visual Design + Technical Design: Martijn van Nunen
