Now I have enough material. Let me write the article.

In the Desert, Watching Back

Somewhere in southern Peru, a 22-meter-long orca is holding a knife. It has been holding that knife for roughly two thousand years, etched into the stony ground of the Nazca Pampa by hands that belonged to a civilization that peaked and collapsed long before the Inca empire rose. Nobody knew it was there — not the Spanish colonists, not the pilots who first spotted the famous Nazca Lines from the air in the 1930s, not the generations of archaeologists who spent a century cataloguing the desert’s strange bestiary. It took a neural network six months to find it.

In September 2024, a team led by Masato Sakai of Yamagata University published a study in the Proceedings of the National Academy of Sciences announcing the discovery of 303 previously unknown figurative geoglyphs on the Nazca Pampa. The find nearly doubled the total count of known figures — from roughly 430 accumulated over almost a century of fieldwork to more than 730. The AI model that made it possible worked about sixteen times faster than human surveyors. It scanned over 600 square kilometers of aerial photography at ten-centimeter resolution, flagging candidate shapes in terrain where the human eye saw only rubble and dust.

This is a story about what was found. But it is also a story about what the finding means — because the new geoglyphs are not simply more of the same. They are a different category of object, and they suggest a different kind of relationship between the Nazca people and the images they carved into the earth.

A Century of Looking Down

The standard narrative begins in the air. In the late 1920s, commercial flights began crossing the Peruvian coastal desert, and passengers noticed something strange below: enormous figures — a spider, a hummingbird, a monkey with a spiraling tail — scratched into the ground at a scale that made them invisible to anyone standing next to them. The Peruvian archaeologist Toribio Mejía Xesspe had spotted lines while hiking through the foothills as early as 1927, but it was the aerial view that made the world pay attention.

American historian Paul Kosok flew over the lines in 1940 and noticed that certain alignments converged at the point of the winter solstice. He called them “the largest astronomy book in the world” — a dramatic claim that launched decades of debate. His collaborator, the German mathematician Maria Reiche, would spend the rest of her life studying and physically protecting the lines, sweeping the desert floor to keep them clear, sleeping in a small hut nearby, and becoming so associated with the site that she earned the title “Lady of the Lines.” Reiche believed the geoglyphs encoded an astronomical calendar. She died in 1998, and her house is now a museum.

The astronomical theory has largely fallen out of favor as a total explanation. Statistical analyses showed that the line-to-star alignments were no better than chance. But Reiche’s real legacy was preservation. She fought off developers, squatters, and the Pan-American Highway — which had already sliced through one geoglyph before anyone thought to stop it — and ensured the site received UNESCO World Heritage status in 1994.

For most of the twentieth century, though, the scholarly emphasis remained on the large, famous geoglyphs: the line-type figures. These were made by removing the dark, iron-oxide-coated pebbles from the desert surface to reveal the lighter ground beneath, creating pale lines against the reddish terrain. They average about 90 meters in length. Some stretch over 300. They depict wild animals — birds, a spider, a whale, a lizard — and they are, by design or accident, best appreciated from the sky.

This created a peculiar problem. Why would a civilization with no aircraft create images that could only be seen from aircraft?

The Question That Won’t Go Away

Theories have not been in short supply. Some are serious; some are not. The extraterrestrial hypothesis — that the lines were landing strips or signals for alien visitors — was popularized by Erich von Däniken’s 1968 book Chariots of the Gods and has no archaeological support whatsoever. It persists mostly as a cultural artifact of its own era, a projection of Space Age anxieties onto ancient ground.

The more credible proposals fall into a few camps. Johan Reinhard, a National Geographic explorer, argued in the 1980s that the lines were connected to mountain worship and water — that in one of the driest places on Earth, the geoglyphs represented a form of prayer or sympathetic magic directed at the sources of irrigation. David Johnson, studying the Nazca puquios (an ingenious system of underground aqueducts), found in the 1990s that many triangular geoglyphs pointed directly at subterranean faults that channeled water into the valley. The lines, in his reading, were a map of invisible hydrology.

Others have proposed the lines as ritual pathways — that the point was not to look at them but to walk them, in processions toward the ceremonial center at Cahuachi. The act of creation may have been the act of worship, the geoglyph a trace of devotion rather than its object.

Each of these theories contains something plausible. None of them accounts for everything. And until 2024, all of them were built on the same limited dataset: the big, famous, line-type geoglyphs that had been catalogued over decades. The small ones — the ones you could actually see from the ground — were mostly ignored.

What the Machine Saw

The collaboration between Yamagata University and IBM Research began around 2018, when Sakai’s team realized that the Nazca Pampa was too large and the remaining undiscovered geoglyphs too faint for traditional survey methods to keep pace. IBM brought computational muscle and expertise in deep learning. The resulting system used a convolutional neural network built on the ResNet50 architecture, trained on the relatively small set of previously known geoglyphs.

The challenge was real. Relief-type geoglyphs — the smaller kind, made by piling or arranging stones rather than clearing them — are often only a few meters across and virtually invisible in standard satellite imagery. They survive as faint variations in surface texture, shadows cast by shallow ridges of displaced rock. The AI was trained to detect these subtle signatures in high-resolution drone photography and aerial surveys.

Of the 303 new geoglyphs confirmed through field visits, 178 were identified directly by the AI model. Another 66 came from clusters the AI flagged, and 59 were found during subsequent fieldwork in areas the AI had directed researchers toward. On average, the team needed to screen only 36 AI-generated candidates to confirm one real geoglyph. That hit rate doesn’t sound impressive until you consider the alternative: a human archaeologist walking 600 square kilometers of featureless desert, looking for stone arrangements the size of a living room.

The figures themselves are striking. Humanoids with outstretched limbs. Decapitated heads — a recurring motif in Nazca iconography, connected to trophy-head rituals documented in the archaeological record. Domesticated llamas. Cats. Birds. Fish. And that orca, 22 meters long, gripping what appears to be a ceremonial knife — a scene echoed on Nazca pottery, where orcas are depicted as supernatural beings severing human heads.

Two Kinds of Line, Two Kinds of Purpose

Here is where the 2024 study pushes beyond discovery into interpretation, and where it gets genuinely important.

The researchers found that relief-type and line-type geoglyphs are not just different in technique. They differ in subject matter, location, and — almost certainly — social function.

The large line-type geoglyphs primarily depict wild animals: 64 percent of their subjects are creatures of the untamed world. They sit, on average, about 34 meters from the network of geometric geoglyphs and major ceremonial roads that crisscross the pampa. They were, the study argues, community-level constructions associated with collective ritual.

The relief-type geoglyphs — the small ones, the ones the AI found hundreds of — are different. They depict humans or human-modified subjects 81.6 percent of the time: people, domesticated animals, decapitated heads, scenes of interaction. And they are located, on average, just 43 meters from ancient informal walking trails. Not the grand ceremonial roads. The everyday paths.

This distribution pattern suggests that relief-type geoglyphs were built and viewed by individuals or small groups — travelers passing through the desert, marking the landscape with private or semi-private imagery. They were not meant to be seen from the sky. They were meant to be seen from the trail, at eye level, by people walking. A lonely figure in the desert glances to the side and sees a humanoid scratched into the stones. A llama. A severed head. Messages left by those who came before.

“The use of AI in research has allowed us to map the distribution of geoglyphs in a faster and more precise way,” Sakai told reporters. The understatement is characteristic of the academic register, but the implication is not understated at all. For a century, the Nazca Lines were treated as a single phenomenon — giant drawings made for unknown viewers. The 2024 data suggests they are at least two phenomena, created at different scales, for different audiences, serving different purposes.

The Visibility Paradox, Reconsidered

So what happens to the old question — why make images only visible from the air — when half the images turn out to be visible from the ground?

It doesn’t disappear. The line-type geoglyphs remain genuinely mysterious in their scale. No hill in the vicinity offers a vantage point from which you could see the hummingbird as a hummingbird. Some researchers have proposed that the Nazca people could have constructed hot-air balloons — there is exactly zero evidence for this — or that the geoglyphs were “seen” by deities or ancestors looking down from the sky. The ritual-pathway hypothesis sidesteps the question entirely: maybe nobody was supposed to see them. Maybe the seeing was never the point.

But the relief-type discovery reshapes the context. The Nazca people were clearly capable of making images at human scale, for human eyes. They chose, in the case of the line-type figures, to make something different — something that exceeded the visual capacity of its creators. That choice feels deliberate. It feels like it mattered. The small geoglyphs were conversation; the large ones were something closer to prayer, or declaration, or an act of faith in an audience that could not be confirmed.

There is a parallel in medieval cathedral construction. The stone carvings on the upper reaches of Gothic cathedrals — a hundred feet above the nave, invisible to anyone below — were made with the same care as those at eye level. When asked why, the carvers reportedly said they were made for God. The Nazca may not have shared the theology, but the impulse — to create beyond the limits of human perception, for a viewer who might not exist — seems to rhyme across millennia.

What Remains

The AI has not solved the Nazca Lines. It has done something arguably more valuable: it has reframed the question. We are no longer asking why an ancient people drew pictures no one could see. We are asking why the same people who drew pictures for everyday travelers also drew pictures that no human could fully perceive — and what the relationship between those two practices was.

There are roughly 600 square kilometers of the Nazca Pampa still to be surveyed at high resolution. Sakai’s team estimates that further AI-assisted work could uncover hundreds more geoglyphs. The desert is not done talking.

But here is what stays with me. Those 303 figures — the humanoids, the llamas, the severed heads along the walking trails — were made by people who expected them to be seen. They were communicating with future travelers, with neighbors, with the next person down the path. For two thousand years, nobody came. The trails were abandoned. The stones shifted. The messages became invisible.

Then a machine trained on pattern recognition flew overhead, and the desert answered back.

The question I cannot shake is this: if the Nazca people made their small geoglyphs to be seen by individuals on foot, and their large geoglyphs for an audience that transcended human sight — and if it took us a century of flight to find the large ones and an artificial intelligence to find the small ones — what, exactly, did they think was watching?

Sources: