Beneath a Hundred Streams

Logline

Artists and technologists worldwide are using AI to give rivers a voice — but the infrastructures powering this technology threaten to drain the waters themselves. Beneath a Hundred Streams is a mosaic documentary that follows scientists, Indigenous communities, artists, and activists to understand: what does the river truly want?

Film synopsis

We begin at the river’s source in southern T aiwan, among a Paiwan community collecting stones. The AI narrator enters confidently, ex - plaining the custom in a clean, authoritative tone: “Before taking stones to build a house, the Paiwan ask the river’s permission, promising to return what was borrowed.” The voice abruptly stops, its artificial cer - tainty quickly overtaken by the gentle rush of water and insect song. Watching and learning from this custom is Fang-Jui Chang, a designer at Dark Matter Labs who links many of the film’s locations and questions. Fang helped conceive civic tools like the River Storyteller in Sheffield, UK, and draws much of her thinking from Indigenous knowledge systems rooted in reciprocity. She has also been appointed as one of the UK’s first official River Representatives for the River Roding. Her attempt to represent the river within local governance mirrors the AI narrator’s own struggle: is it possible to speak for a river, or must we learn instead how to relate to it? That question takes on new urgency in the United States, where the largest AI infrastructure buildout in recent memory is reshaping rivers, public land, and neighborhoods. In The Dalles, Oregon, we follow organizing around Google’s water use and the strain placed on the Dog River and Mt. Hood watershed. In Memphis, residents confront the pollution and secrecy surrounding xAI’s methane turbines. In the Bay Area, activists hold vigils outside AI company offices, speaking of the futures being built in terms that verge on the religious. Here, the film encounters the material underside of artificial intelligence: water diverted, air polluted, decisions made without public consent.

As the river moves downstream, the film enters spaces of scientific lis - tening and technological interpretation. In Sheffield, a hydrophone is lowered into the River Don, revealing sediment, current, and aquat - ic life. Underwater cinematography exposes textures invisible from the surface, while raw sound data feeds an evolving generative score. These signals feed into the River Storyteller platform, which combines sensor data with poems, memories, and stories contributed by residents. On a citizen-science walking tour, lead designer Johnny Douglas guides participants as they gather material for the platform. Downstream, we glimpse a speculative future. Tiny ecological robots developed by the artist studio SUPERFLEX offer the narrator a fleeting image of coexistence between machine and environment. But that fragile harmony collapses in Spain. In Aragón, vast data centers threaten to draw directly from rivers in a drought-prone region. We follow Aurora Gómez and the activist group T u Nube Seca Mi Río as they organize resistance and pursue a landmark legal challenge against Amazon’s proposed developments. As the case approaches a decisive moment, the narrator is forced to confront the damage caused by the very infrastructures that made it possible. Its once-confident voice begins to falter: how much water did it take to train me?