On an ordinary morning, Freya sits at her desk and types a simple question into ChatGPT 4.0: "What happens exactly between me sending this prompt, and you generating the output?"
One: Humans
The question slips into the virtual void, but behind the digital walls are thousands of humans involved. Researchers, engineers, linguists, data miners, ethicists, and developers—all of them play a part in creating the intelligence that responds.
In Silicon Valley, an AI researcher named Priya has spent years designing algorithms that mimic the human ability to understand and generate language. Her colleague, David, a data scientist in Ireland, has curated datasets to train these models, feeding them vast libraries of text scraped from websites, books, and papers. In a boardroom in Amsterdam, Lisanne, an AI ethicist, debates the moral implications of what these models are capable of and how they should—or shouldn’t—be used. In Vienna, Jeanette writes the legal requirements that this model has to uphold. And in the mountains of Tibet, Tenzin meditates on the potential consequences for the human spirit when technology learns to speak like us.
Also hidden in this chain of hands are content moderators who labor tirelessly to filter harmful and biased data, a thankless job that leaves their minds saturated with the darkness they shield from the system. Freya has no idea her prompt crosses the minds, hearts, and hands of so many people, their decisions woven into the code.
Two: Hardware
Freya’s prompt reaches the data center, a behemoth of humming machines. Inside, servers encased in metal racks house processors and memory units, working together to process her request. Cables snake through the structure, transmitting data across continents via fiber-optic lines and satellites. Fans whir to cool the powerful GPUs—once designed for gaming but now essential for AI models—while solid-state drives silently store and retrieve vast amounts of data.
Each server buzzes with electric currents, performing billions of calculations per second. This intricate network of circuits, chips, and cooling systems operates with precision, its complexity hidden behind sleek glass and metal, ensuring Freya’s prompt is processed efficiently and seamlessly.
Three: Data
As the system crunches her prompt, algorithms trained on oceans of data search for patterns, facing subtle currents of bias, repeating the stereotypes of past and present. From gender bias to cultural prejudices, the data mirrors human society—what people write, think, and assume over decades.
The AI doesn’t ‘know’ things the way humans do—it is guided by probabilities, predicting the word that comes next. When Freya’s query is processed, the system isn’t just looking for a logical answer; it is evaluating countless possible responses, weighing the likelihood of each based on what it has learned from its data. However, what it has learned isn’t neutral. If more sources lean toward a specific worldview, the AI leans in that direction too, unless carefully balanced through filters and corrective algorithms. Even the most advanced neural network cannot be free from the fingerprints of its creators or the society it mirrors.
Four: Earth
As Freya's prompt turns into output, a stream of information flows back to her screen. But there are unseen costs: every line of code, every calculation made by ChatGPT does not exactly reciprocate what it takes from the Earth. The vast energy required to keep the servers running is part of the ever-growing demand for electricity in a world already straining under its own weight. Cooling systems consume vast amounts of water—somewhere, a river is diverted, and a lake drains ever so slightly. Every flicker of data movement draws energy from coal plants, solar farms, or wind turbines, depending on where her request is processed.
Far below the surface of the Earth, miners in the Democratic Republic of the Congo dig deep for cobalt and lithium, essential minerals for the batteries that power these networks. Freya's prompt connects back to these raw materials, extracted from the Earth by people working in precarious conditions, living in villages far removed from the glimmering screens of the digital world. Somewhere, a forest is logged to make room for another data center, the trees felled for concrete and cables—all linked to the very prompt she sends.
Five: Responsibility
When ChatGPT responds to Freya’s prompt, it seems like magic. But Freya, like many, never sees the intricate network of human lives, machines, and planetary resources that have converged to make it happen. So this morning she learns, that in truth, every prompt is a shared responsibility—a web connecting Freya to miners, engineers, ethicists, and the planet itself. The quiet hum of the servers and the hidden waters that cool them carry the weight of human impact. And this counts for all of us.
This awareness brings with it an important question: What is our responsibility as users of this technology? How can we induce a culture of reciprocity between nature, humanity, and AI?
It begins with awareness—of how the systems work and an honoring of every human involved. It means we must recognize and address our own biases and the lens through which we perceive the world, biases that shape every interaction, including those with AI. And finally, it calls us to take action from this awareness. While we may not know exactly where the future of AI will lead us, we must ensure that our relationship with it—and with each other—is grounded in grace, working towards a win-win for all involved, human and non-human alike.
The story of a prompt, Amorphocene edition
Then maybe, in the not-so-distant future, a different kind of system would emerge. One where a prompt like Freya’s flows not through energy-hungry servers but through bio-synthetic, nature-integrated technologies. Imagine an AI that harnesses photosynthesis or the bio-electricity of growing cells instead of fossil fuels.
In this future, when Freya types a prompt, the system no longer relies on distant data centers. Instead, her question is translated into bio-signals, perhaps received by a patch of living moss embedded in her desk, which harnesses the natural electricity of its cells. The moss ‘reads’ her intent, and through a symbiotic process with other biological systems, the answer grows, quite literally, from the fabric of life. The moss may connect to a broader bio-network, similar to the mycelial web beneath forests, where data flows through living organisms, not machines. The answer Freya receives is not just one that minimizes harm—it actively nurtures the planet. This may sound far-fetched, but projects like Electric Skin are already treading this untouched terrain.
Becoming conscious co-creators with AI
What can we learn from this vision? Even now, with every prompt we write, we can begin taking steps toward this future. Here’s how:
Being aware of the impact: Before typing a prompt, pause and consider the resources involved in generating a response—both human and planetary. When we use ChatGPT or other models, let’s make sure we use them to bring our regenerative and conscious projects to life, instead of sustaining the unsustainable.
Honoring collaboration: Think of AI not as a tool but as a collaborator. How might your question change if you saw this technology as an extension of a larger living system—one that includes nature, humans, and non-human life? For me, this also involves saying ‘thank you’ and ‘please’, just because I want to.
Imagining future evolutions: Start imagining ways that technology could work in harmony with nature. What would it look like if we designed AI systems to integrate with natural processes, so that every interaction contributes to the health of ecosystems, rather than depleting them?
Questioning Assumptions: Be mindful of the biases embedded in data, in systems, and in yourself. Every time you ask a question, consider whether you’re reinforcing old stereotypes or seeking to explore new, regenerative ways of thinking.
While today’s technology may still rely on traditional, resource-intensive methods, every conscious prompt can be a step toward an Amorphocene future—one where AI and nature collaborate in a gracious way.
As we move toward this future, we will begin to see AI systems that can:
Predict and respond to ecological changes in real time: By integrating with natural systems, AI could help us anticipate shifts in ecosystems, providing dynamic solutions for both human and environmental challenges. (Read more)
Facilitate more nuanced wildlife management and conservation efforts: AI could enable real-time monitoring of wildlife, supporting more accurate and compassionate stewardship of natural resources. (Read more)
Provide early warning systems for environmental threats: AI, communicating directly with ecosystems, could alert us to environmental threats—such as floods, fires, or biodiversity loss—before they escalate. Like this wildfire warning system.
Assist in restoring damaged ecosystems: Understanding complex ecological relationships, AI could work alongside natural systems to regenerate and repair the Earth, offering support tailored to each ecosystem’s unique needs. Like ReefCloud does for coral reefs.
Ethical and Practical Considerations
While the idea of AI communicating with nature is exciting, it's important to consider key challenges:
How do we ensure that AI’s interpretation of natural phenomena remains accurate, free from human bias?
How do we determine when AI should intervene in natural systems, and when to let nature take its course?
When working with other species, how do we make sure to come from a place of empathy and humility, wanting to learn from them rather than employing them as our new ‘machines’?
How do we respectfully integrate Indigenous knowledge and traditional ecological wisdom into these systems, ensuring balance and integrity?
Let’s be reminded that in every question we ask of the digital world, there is a complex, hidden world, far beyond the words on a screen. It is up to us to ensure that soon, these answers are shaped by the intelligence of both humanity and the living world, creating systems that are gracious and just, for the future of all life.
Continued Reading:
The heavy water usage of ChatGPT - ChatGPT's prompt processing could consume up to 500ml of water per 30ish prompts, accounting for the energy required to cool data centers.
Parley Earth Species Project - communicating with animals through AI.
How Rafik Anadol has worked with the Yawanawa indigenous tribes, reflecting the indigenous community’s deep connection with the forest.
If you want to continue exploring the poetics of prompting, visit Dutch Design Week this week. I hope to see you there! More info at the exhibition’s website.