Over the phone, we agree on a metaphor that starts like a tangent and ends like a warning. Inbreeding brought down the Hapsburg Empire.

At first it sounds like a reach. But then it lands. The Habsburgs, who ruled vast parts of Europe, married within the family to preserve power. It worked for a while. Until it didn’t. What followed was disease, deformity, and dynastic decline. Nayak believes generative AI may be following a similar path.

Models like ChatGPT are trained on massive datasets built from human-made content. Books, essays, art, music, journalism, software. But as more people begin using AI to generate output, that AI-made content begins to bleed into the datasets that train the next generation of models. Over time, that loop could grow tighter and more self-referential. “It’s like inbreeding,” Nayak says. “The original pool gets gradually corrupted.”

The tragedy, he adds, is not just about degraded data. It’s also cultural. “People who couldn’t write are now calling themselves authors. Those who’ve never held a brush are claiming to be visual artists. What we’re seeing isn’t the rise of creativity. It’s a flood of well-dressed mediocrity.”

But not everyone sees it this way. A neuroscientist in Pune, who asks not to be named, offers a counterpoint. “Humans learn by mimicry too. We absorb what came before us. Language. Logic. Music. Emotion. We connect dots. And when we’re lucky, something new emerges.”

He doesn’t completely reject Nayak’s Hapsburg metaphor. He just doesn’t think it signals doom. “Not every mutation is deadly. Some give rise to wings.” Where both men unexpectedly agree is not on data, but on energy.

The neuroscientist says, “The human brain is the most energy-efficient creative engine in the known universe.” He points to how we generate symphonies, code, poetry, and design using the energy equivalent of a light bulb. “Our brain operates on about 20 watts when at rest.”

By contrast, AI requires an enormous ecological infrastructure. The numbers are sobering. Training a single large language model like GPT-3, for instance, can consume over 700,000 litres of clean freshwater if done in Microsoft’s data centres in the U.S. If the same process were run in Asia, the water requirement would triple. These are not speculative estimates. They were cited by Professor Shaolei Ren from the University of California, Riverside, in an interview with The Markup.

And that figure doesn’t even account for the electricity required to run those servers, or the fossil fuels that power that electricity. We call it the cloud, but clouds don’t leave carbon footprints. Data centres do.

In a world already staggering under the weight of climate change, what is the real cost of this machine-driven creativity? And can we afford to keep building datacentres that guzzle water, electricity, and minerals at industrial scale just to write emails faster?

Nayak and the neuroscientist may differ on where AI is headed. But on this question, they are in quiet agreement. “The planet cannot afford unthinking AI deployment,” says the neuroscientist. Nayak nods. “It’s not about ethics. It’s about entropy.”

It bears to keep in mind that AI is a tool. Like processed food was a tool. It ended famines, but gave rise to lifestyle diseases. We fed more people, but paid a long-term cost in public health. Something similar may unfold with AI. It will free up time, unlock scale. It will push the average ceiling higher but at a cost.

Maybe it is time we stopped writing obituaries for creativity. Maybe the more urgent task is to examine the kind of creativity we truly value. Do we want a world flooded with content that looks right but feels hollow? Or one where originality, even when imperfect, still carries the scent of human effort?

The choices we make now will define the cultural landscape of the future. And those choices cannot be left to algorithms alone. That’s where the human hand still matters-- in asking harder questions and in insisting that quality is not just output, but intent. While machines may learn faster, only humans know why something matters. And maybe that’s what we need to hold on to.