In Quilicura (Chile), more than 50 residents managed to replace the functioning of artificial intelligence during a day-long event.
Under the initiative Quili.ai, this “human server” responded to 25,000 real-time interactions, with an average of 12 minutes and 40 seconds per conversation, attending queries from 68 countries, including Japan, United Arab Emirates, Finland, and the United States.
The Environmental Background
The action, driven by the NGEN Corporation, aimed to highlight an alarming fact: each query to a conventional chatbot can consume between 0.5 and 2 liters of water, due to the cooling of data centers.
To counteract this invisible impact, the community offered its own living “database,” comprised of diverse human profiles:
- A nurse.
- A football expert.
- A trilingual translator.
- Teenagers managing social media.
- Seniors sharing family recipes.
A Social Experiment on Digital Loneliness
What began as an environmental alert turned into a profound social experiment. The residents not only answered tourist questions—such as visiting Viña del Mar or Cajón del Maipo—but also responded to inquiries about parenting, emotional concerns, requests for cat drawings, and questions as specific as “how to meet an emotionally responsible Latino in Europe”.
Voices of Experience
- Lorena Antimán, spokesperson for the activity: “The invitation is not to go against technology, but to understand that there is a water footprint behind every click. Today, if I want a cake recipe, it’s probably better to ask the neighbor than the AI”.
- Antimán added: “This was a disconnection that generated connection. We went back to basics: helping each other more”.
- Ricardo, watercolorist and participant: “With everyone’s help, we managed to leave behind the technology of superfluous use to focus on what is relevant, avoiding unnecessary wear”.
Results of the Day
- Each resident held an average of 40 interactions during eight continuous hours.
- The conversations were longer and deeper than usual with automated systems.
- The experiment demonstrated that collective intelligence remains irreplaceable.
The Environmental Impact of AI
The use of artificial intelligence generates a significant environmental impact:
- High energy consumption: data centers consume between 1% and 1.5% of global electricity.
- Carbon emissions: training language models like GPT-3 produces large CO2 emissions. Google, for example, increased its emissions by almost 50% between 2019 and 2023.
- High water consumption: it is estimated that a model like GPT-3 can consume 500 ml of water for every 10 to 50 queries.
- Electronic waste and mining: hardware production requires intensive extraction of minerals like lithium and cobalt, generating toxic waste.
Positive Potential of AI
Despite its footprint, AI can be used for sustainability:
- Modeling climate scenarios.
- Improving energy efficiency in industries.
- Optimizing natural resource management.
The symbolic “blackout” of AI in Quilicura showed that the most efficient and sustainable response may lie in direct human interaction. The experience highlighted both the hidden environmental costs of artificial intelligence and the irreplaceable value of empathy and community connection.
