I have just returned from a long trip to Egypt. The pyramids, temples and museums, especially the recently opened Grand Egyptian Museum, were exceptional. So were the hieroglyphics and frescoes that seemed to have been made yesterday, even though they were more than three thousand years old in the Valley of the Kings in Luxor.
What also surprises me is everything that has been added to it since then. The walls of these tombs and temples are covered in centuries of graffiti. Greek and Roman visitors engraved their names on the stone. Coptic Christians left the crosses and inscriptions behind when they turned the royal burial chambers into unofficial holy places. Nineteenth-century European travelers on the Grand Tour scratched their names at Luxor Temple.
What was not known until very recently is that there are approximately thirty inscriptions in ancient Indian languages spread across six tombs in the Valley of the Kings.
A man named Sekai Kuran scratched his name eight times in five different tombs nearly two thousand years ago. In the tomb of Ramesses IX, his inscription is five to six meters above the entrance, and no one has yet known how he got there. The Tamil text simply translates as: “Sikai Kuran came here and saw.”
These findings were presented at the International Conference on Tamil Inscriptions in Chennai in February by Ingo Strauch of the University of Lausanne and Charlotte Schmid of the French School of Asian Studies. About twenty of the inscriptions are in Tamil Brahmi, which is the oldest known script for writing the Tamil language. The rest is in Sanskrit, Prakrit and Gandhari-Kharushthi, indicating the arrival of visitors from all over the subcontinent.
A Sanskrit inscription identifies its author as a messenger of King Ksharatha, a dynasty that ruled western India in the first century. In one tomb, Sanskrit and Tamil graffiti appear to interact with a nearby Greek inscription, suggesting that scribes could read in all three languages.
We have known for some time that ancient Indians traveled through the Roman world. The trading colonies of Berenike on the Red Sea, the Indian artefacts of Socotra, and the ubiquitous Tamil-Brahmi pottery in each, paint a vivid picture of a connected ancient world in which goods and people moved between the Malabar coast and the Mediterranean for trade.
The Valley of the Kings carvings add a more intimate layer. Coran had traveled inland to an ancient sacred site already in his time. He was a tourist like me, although, unlike me, he left his mark for posterity.
The inscriptions were not hidden. In 1926, French researcher Jules Pellet published a catalog of more than two thousand graffiti in these tombs. Over the next century, generations of Egyptologists relied on it, but the question of what the unrecognized scratches were was never asked, because those who knew Indian languages did not study Egyptian pharaohs.
Strauch happened to know the languages that turned those ignored scratches into sentences. He noticed the writing while on a tour, photographed it, took it home, and learned what other scholars had experienced.
This is a reminder that what seems like a complete record can remain incomplete for a hundred years, simply because the right reader has not yet arrived.
It’s also a case where a certain type of AI has something to offer. The headline-grabbing examples of artificial intelligence (AI) in research tend to be the niche ones. The Vesuvius Challenge used machine learning and new imaging to read the charred Herculaneum manuscripts buried by the eruption that destroyed Pompeii in 79 AD. It has long been considered impossible to open them without destroying them.
The larger language models that most of us use do something different. They are not as deep as human experts or specialized AI models trained on specific tasks. It is wider in width but less deep. The model read across Egyptology, Tamil literature, and Indian Ocean trade does not belong to any of these fields. This is exactly what can make it useful as a connector of disparate ideas.
Strauch made his discovery because he was in a tomb where he saw the scribblings on the wall. But if the question had been asked, the AI might have suggested years ago that the signs Billett ignored resembled Indian scripts and were worth showing to someone who could read them.
Old World catalogs and colonial-era reports certainly contain more ancient tourists waiting to be noticed.
The same applies to various sciences. We sit on vast archives of data, samples and notes collected for one purpose and never re-read for another.
Anirban Mahapatra is a scientist and author. His latest book is When Medications Don’t Work. The opinions expressed are personal
