The people of Quilicura, a community on the outer edge of Santiago, spent a Saturday running a 12-hour, human-operated chatbot to highlight the environmental cost of automated AI services. About 50 residents rotated through a community center to answer questions and create hand-drawn images on request, and organizers say the project handled more than 25,000 incoming prompts from around the world. The exercise aimed to contrast instant, energy- and water-intensive AI responses with a slower, local alternative and to draw attention to water stress tied to data center cooling in the region. Organizers framed the experiment as both protest and public education about resource use.
Key Takeaways
- Roughly 50 local volunteers staffed Quili.AI for about 12 hours on a Saturday, answering questions and producing on-demand drawings.
- Organizers reported the site received more than 25,000 prompts during the event, an average exceeding 2,000 requests per hour.
- One illustrative interaction: a request for a “sloth playing in the snow” returned a penciled sketch about 10 minutes after a volunteer acknowledged the user in Spanish.
- The project was organized by Corporación NGEN and led publicly by Lorena Antiman to underscore AI’s hidden water footprint in Quilicura and the Santiago region.
- Quilicura has become a hub for data centers; major cloud providers including Amazon, Google and Microsoft have built or planned facilities in the Santiago area.
- Google has said its Quilicura data center—switched on in 2015—is highly energy efficient and pointed to local wetland and irrigation investments while facing legal challenges over water use on other projects.
- Chile has endured roughly a decade of severe drought, a context organizers say intensifies concerns about water used for data center cooling.
Background
Quilicura sits at Santiago’s urban fringe and, in recent years, has attracted cloud-computing infrastructure because of connectivity and land availability. Those same qualities have made the municipality a focal point in debates over the local environmental footprint of data centers, especially where cooling systems can consume electricity and, in some designs, substantial quantities of water. Corporación NGEN and community activists have used public projects and demonstrations to question how new computing capacity affects water-scarce basins such as the nearby Maipo River basin.
Data centers vary in their energy and water intensity: some rely heavily on air-cooling or recycled water, while others use different technologies that shift resource burdens. Multinational cloud providers—Amazon, Google and Microsoft among them—have invested in facilities across Latin America, citing demand and regional expansion strategies. Google has described its Quilicura facility, first activated in 2015, as one of the region’s most energy-efficient sites and has pointed to investments in local restoration and irrigation efforts. Still, some projects have faced court challenges and community scrutiny focused on groundwater and river usage.
Main Event
On the day of the experiment, volunteers worked on laptops inside a municipal community center, rotating through shifts to keep the service online for a full 12 hours. When a user asked Quili.AI for an image—such as a “sloth playing in the snow”—a volunteer first replied in Spanish that a human would respond and to please wait. About 10 minutes later a penciled, cartoonish sloth sketch arrived, demonstrating the trade-off between immediacy and a lower-technology, low-carbon interaction.
Not all interactions produced creative artifacts; some prompts sought local cultural knowledge such as recipes. In those cases volunteers either answered from memory or walked across the room to ask someone else with specific knowledge, reflecting the community-centered approach organizers wanted to showcase. Organizers emphasized that the project was not an argument to ban useful AI applications but a moment to question habitual, high-volume prompting that may carry environmental costs in water-stressed areas.
Project coordinators tracked incoming requests and posted aggregate counts; they reported handling more than 25,000 worldwide prompts during the 12-hour run. Volunteers included younger residents who contributed sketches and older participants who offered recipes or local context. The experiment blended civic education, performance, and protest to visualize how everyday digital habits can have local environmental consequences.
Analysis & Implications
The Quili.AI project reframes a global infrastructure debate at a local scale: data centers enable cloud services millions use daily, but their construction and cooling can concentrate energy and water demands in host communities. In water-scarce regions like central Chile, even indirect increases in water use prompt concerns about irrigation, river health and long-term availability for residents and agriculture. By offering a human alternative, organizers closed the loop between user behavior—rapid, repeated prompting—and visible, tangible labor in the community.
Policy implications include stronger local permitting scrutiny, clearer reporting on data-center water and energy use, and incentives or requirements for less water-intensive cooling technologies. Corporations often report on energy-efficiency measures and mitigation projects, but community groups argue that reporting and mitigation must be transparent, quantified, and responsive to local hydrological conditions. Regulators and utilities may face pressure to reconcile economic development tied to data centers with long-term water sustainability plans.
For users and designers of generative AI, Quili.AI also raises questions about demand-side interventions: nudges, prompts, or rate limits could curb excessive queries, and developers could provide per-request carbon or water cost estimates to inform behavior. The event suggests reputational risk for firms that expand in water-stressed regions without clear community benefits or measurable safeguards, and it highlights how civic activism can shape public discourse on digital infrastructure.
Comparison & Data
| Metric | Value |
|---|---|
| Volunteer participants | About 50 people |
| Event length | 12 hours |
| Reported prompts handled | More than 25,000 requests |
| Average requests per hour (reported) | ~2,083+ |
| Example response time (drawing) | About 10 minutes |
| Google Quilicura data center | Switched on in 2015 (company-stated efficiency) |
The table summarizes the event’s key operational metrics alongside a notable corporate milestone for context. Reported request volumes show how rapidly demand can grow: handling roughly 25,000 prompts in half a day required sustained volunteer attention and illustrates why automated systems scale differently. The 10-minute turnaround for an illustrative drawing underscores both the labor intensity of human responses and the experiential trade-offs participants sought to foreground.
Reactions & Quotes
Organizers and participants framed the experiment as both an educational demonstration and a moral reminder about resource use.
“The goal is to highlight the hidden water footprint behind AI prompting and encourage more responsible use.”
Lorena Antiman, Corporación NGEN (organizer)
Antiman and the organizing team emphasized the project’s intent to invite reflection rather than rejection of AI tools. They said Quili.AI was designed to show alternative practices and to encourage users to weigh environmental costs against convenience.
“Quili.AI isn’t about always having an instant answer. It’s about recognizing that not every question needs one.”
Lorena Antiman, Corporación NGEN (organizer)
The community response included volunteers of different ages contributing distinct skills, from illustration to culinary knowledge. Local organizers described eager international interest in the experiment, while some residents expressed pride in using local knowledge to answer global questions.
Unconfirmed
- Precise per-request water or energy costs for prompts routed through regional data centers were not published by organizers and remain unverified.
- Claims about the net impact of corporate wetland restoration projects on local water availability lack independent, publicly available hydrological assessments in this reporting.
- The long-term legal outcomes and full details of court challenges related to other data-center projects near Santiago were not resolved in the source report and require follow-up.
Bottom Line
The Quili.AI experiment turned a technical and often invisible infrastructure problem into a visible community activity, showing how everyday digital behavior can translate into local environmental pressures. By substituting human labor for instantaneous AI responses, organizers illustrated trade-offs—slower interactions and limited scale, but a clearer connection between action and resource use—intended to provoke public reflection and policy conversation.
As cloud providers continue to expand in Latin America, transparency about water and energy use, stronger local engagement, and technology choices that reduce freshwater demand will be central to reconciling economic and environmental priorities. Quilicura’s demonstration is a reminder that infrastructure decisions are not only technical or corporate but also civic and ecological, and that communities will increasingly insist on measurable safeguards and local input.