From Lidar scans to automatic papyrus deciphering, new technologies and AI are revolutionizing archaeological research. But without context analysis and the guidance of the human brain, errors are just around the corner.
Brushes and trowel (the famous English trowel, scepter of every archaeologist), whip and fedora, Heinrich Schliemann and Indiana Jones: if your imagination of archaeological research draws on these models, you will have to update yourself. Because today, undermining the throne of stratigraphic excavation, father of every buried mystery, there are them: new technologies.
The same ones that, only in the last handful of years, have revealed the existence of forgotten ruins, translated unknown languages, reconstructed illegible texts and finds impossible to decipher.
Excavations and field research
Of course, it is still early to imagine legions of robots digging in place of ancient investigators in flesh and blood and cargo pants, yet digital has already entered fully into field research, profoundly transforming the sciences of antiquity. Shovel, pickaxe and books are no longer enough today: the old reconnaissance on foot, the “traditional” excavation and the study of the finds are often integrated or even replaced by sophisticated analysis tools.
At first glance it is a series of unpronounceable acronyms, which can only be dissolved by selling the soul of the Italian language to diabolical Englishisms, but if we wanted to describe them in a single concept we could define them as powerful weapons at the service of the detectives of the past. Less fascinating than Indy’s leather equipment, perhaps, but certainly more useful than a whip, the reconnaissance vehicles often created for military purposes, the increasingly precise and powerful medical diagnostic tools, the algorithms applied to artificial intelligence are revolutionizing the sciences of antiquity and solving or reducing many of what for years seemed indecipherable puzzles.
Remote sensing
But how are they doing it? And with what repercussions? «Since the birth of digital archaeology, the use of technologies applied to research and methodology has constantly evolved: from the first databases and GIS systems (a system of digital maps that integrates spatial data, historical maps and excavation information) to 3D survey (a generic definition that includes all the technologies for acquiring three-dimensional images, including Lidar, photogrammetry and laser scanners) up to the current integration of artificial intelligence», explains Luca Sanna, archaeologist and professor of Digital Archeology at the University of Cagliari.
«This new frontier is revolutionizing research, essentially accelerating two specific areas: landscape archeology and the study of material culture in museum deposits. In the first case, Deep Learning algorithms (i.e. an artificial intelligence model that imitates the human brain to recognize complex patterns) and Computer Vision allow you to quickly scan enormous volumes of data, such as 3D images collected over large areas, to automatically and quickly identify topographic anomalies invisible to the naked eye that can hide potential sites.
In the second, artificial intelligence recomposes complex three-dimensional puzzles of fragmentary finds and deciphers ancient inscriptions.”
In short: if until a couple of generations ago archaeologists wore out their soles in vain on inaccessible terrain, for long, sometimes dangerous and often partial reconnaissances, today the majority of discoveries come from the most impassable areas of the planet.
Thanks to a powerful laser remote sensing system mounted on drones or aircraft, launched in flight over the tropical jungles of the world and the rainforests of Central and South America, over the remote frozen plateaus of Central Asia or the nearest forested areas of Switzerland and Italy.
Light meter
It’s called Lidar (acronym for Light Detection and Ranging, i.e. “detection and measurement using light”) and it reveals hidden treasures by functioning as a kind of radar 3.0, which, instead of the radio waves of its old colleague, sends thousands of laser pulses per second to the ground, of a wavelength suitable for penetrating through the densest vegetation. Based on the time it takes for each pulse to return after bouncing off the ground, the system calculates the distance to all the individual points, gradually reconstructing the profile of the ground for hundreds of square kilometres.
The final result is a very detailed three-dimensional topographic map, with millimetric details of every micro-variation in altitude, obtained without deforestation, preventive excavations or impossible reconnaissance. As if a gigantic lawnmower had razed the Amazon forest or the rain jungles to the ground, in its most recent reconnaissances the Lidar has rewritten the history of the urban and economic evolution of regions long considered uninhabited: in Guatemala and Yucatán there are now hundreds of thousands of new Mayan sites, interconnected with each other and equipped with advanced agricultural systems hidden by the jungle, while in Mexico the ceremonial complex of Aguada Fénix has been recognized as the most ancient and vast monument never built by the Mayan civilization, still with many secrets to be revealed.
In the Peruvian Andes, perched at almost 4 thousand meters above sea level in the archaeological area of Wat’a, a 15th century citadel has surpassed in height and age its more famous Inca cousin, Machu Picchu. Two thousand meters lower, but at very different latitudes, unexpected lost cities have also cropped up in the rugged mountains of southeastern Uzbekistan, along the ancient Silk Road.
Among others, Tugunbulak, a metropolis covering one hundred and twenty hectares, with houses, squares, streets, five watchtowers and a central fortress protected by a stone and mud brick wall, which thrived between the 6th and 11th centuries, in a hostile and cold environment, where no scholar expected to be able to find signs of urban life, metallurgy and agriculture dating back to over a thousand years ago.
The guidance of a human brain
But, as an old advertisement said, “power is nothing without control” and no technology can do without a human brain that guides it or knows how to interpret and verify its results. «This is a true digital revolution that does not replace the researcher, but supports and speeds up his work», confirms Sanna. «Scanners and ground penetrating radar offer only physical maps, they record “anomalies”, but the earth is a historical archive made up of context, microstratigraphies and relationships that only man can currently interpret. The past cannot be understood without direct contact and analysis of the material gained through on-site experience.”
And in fact, just like all ordinary mortals, artificial intelligence also happens to make mistakes when it is used to recognize hypothetical ancient structures in the morphological anomalies of the terrain. «In archaeology, AI fails on context: it lacks the intuition, experience and ability to interpret sites, and it also suffers from training biases», notes the expert. «Working on fragmentary data often generates visual hallucinations, because the software, trained to look for regular shapes (circles, squares, rectangles), tends to force the interpretation: it then happens that you “see” a proto-historic nuraghe in a circular enclosure built fifty years ago with dry stones by a shepherd or even in a road roundabout covered by vegetation».
Without historical sense
In addition to hallucinations, the AI also suffers from a total lack of historical sense: very good at identifying a shape, it is unable to associate a stratigraphy with the correct era, anachronistically fusing elements that are millennia apart. But not only that. «The algorithm can also err by default», continues Sanna. “The archaeological structure is there, but the AI erases it or does not detect it, because it does not fall within its rigid statistical canons or because it is mistaken for a disturbing element.” Its fallibility, however, does not detract from the appeal of artificial intelligence and Machine Learning is also applied to the study of ceramic shapes or to the deciphering of ancient texts.
This is the case of the 1,800 papyri in Greek and Latin dating back to over two thousand years ago, found in the 18th century in the Villa dei Papiri in Herculaneum (about 16 kilometers from Pompeii), crumpled and blackened by the heat of the eruption of Vesuvius in 79 AD, which remained preserved for a long time in the display cases of the National Library of Naples without being able to be read.
Until 2024, when a papyrus was virtually unrolled and translated thanks to AI, the intuition of an international team of archaeologists, philologists and computer scientists, and a CT scan. No, not a simple X-ray CT scan like the one they put the little cousin into to see if he had appendicitis, but a high-definition synchrotron radiation CT scan, a particle accelerator capable of identifying traces of ink on the surface of the specimen. Precisely from these traces, a bit like investigators who, to highlight the marks left by the culprit, rub a pencil on the sheet of notepad on which he wrote down the victim’s address, the AI managed to reconstruct the written text: more than 2 thousand Greek characters, a long note on how to enjoy life according to a follower of the Greek philosopher Epicurus.
Researchers at the English University of Nottingham did something similar for Latin inscriptions, feeding Aeneas, the AI developed for archeology by Google DeepMind, with a database of over 176 thousand epigraphs dating between the 7th century BC and the 8th AD and using it to restore the missing texts, indicate their possible provenance and estimate their age.
