A human-powered chatbot rather than one with artificial intelligence? This community in Chile demonstrates why

Around 50 residents in a community outside Chile’s capital spent 12 hours on Saturday doing something unusual: acting as a fully human-powered chatbot.

Working from laptops inside a small community center in Quilicura, on the outskirts of Santiago, volunteers answered questions, wrote replies and even hand-drew images for users around the world — all to spotlight the environmental toll of artificial intelligence data centers operating in the region.

The project, called Quili.AI, fielded more than 25,000 requests during the day, according to organizers.

But unlike tools such as ChatGPT or Google’s Gemini, responses weren’t instant.

When one user asked for an image of a “sloth playing in the snow,” the site replied in Spanish, asking for patience and reminding the user that a human was working on the request. About 10 minutes later, a penciled sketch appeared: a cartoonish sloth clutching a snowball, ready to throw it.

Slowing down AI

“The goal is to highlight the hidden water footprint behind AI prompting and encourage more responsible use,” said Lorena Antiman of environmental group Corporación NGEN, which organized the event.

Volunteers rotated through shifts answering queries. For culturally specific questions — like how to make Chilean sopaipillas, a traditional fried pastry — responses came quickly. When no one knew the answer, participants walked around the room asking others for help.

“Quili.AI isn’t about always having an instant answer,” Antiman said. “It’s about recognizing that not every question needs one.”

She added that the project is not meant to reject AI altogether but to prompt reflection about the impact of “casual prompting,” especially in water-stressed areas.

Data centers and drought

Quilicura has become a growing hub for data centers built by global cloud companies including Amazon, Google and Microsoft.

AI systems require powerful chips that consume large amounts of electricity and, in many cases, significant volumes of water for cooling. Usage levels vary depending on equipment and location, but environmental groups say the demands are substantial.

Google has described its Quilicura facility, opened in 2015, as the most energy-efficient data center in Latin America and says it has invested in wetlands restoration and irrigation projects in the nearby Maipo River basin. However, the company has also faced legal challenges over water use tied to other proposed projects.

The concerns come as Chile grapples with a decade-long drought that experts say has worsened environmental pressures and contributed to recent deadly wildfires.

A human alternative

For participants, the slower, handmade approach was part of the message.

Instead of algorithms delivering instant certainty, Quili.AI offered something different: human curiosity, shared knowledge and occasional “we don’t know.”

In a world racing toward ever-faster artificial intelligence, the experiment suggested that sometimes the most sustainable answer is simply to slow down.