About two decades ago, the conference rooms of Delhi institutions like the Sahitya Akademi and India International Centre were often haunted by a tall, thin and unfailingly polite gentleman with a briefcase who always sat at the back. There, he carried on a private conference with his neighbours that began with these momentous words: “I have it!”
The nature of ‘it’ was revealed when he shyly—but proudly—opened the briefcase and took out sheaves of paper thickly inscribed with the Indus Valley script and copious notes. “This is the fish symbol, meena,” he said. “And this is the horse symbol.” What did the horse represent? Enemies, if you believe the now-tattered Aryan invasion theory. But some of his conclusions were not completely outrageous. For comparison, ripples did represent water in Egyptian hieroglyphs, and the Roman capital letter A is a rotated representation of the head of a ram—aries in Latin. No different from ‘A for apple’ in modern children’s alphabet books.
The trouble was that this gentleman gave the more inscrutable symbols the quick-and-dirty treatment. You asked, “What’s this?” He replied, “A phallic symbol.” You persisted: “And that one?” Again, “A phallic symbol.” This, too, was not wholly incredible: an extraordinary number of Roman artefacts in that category have been uncovered. But they communicated only one message: they discouraged malevolent spirits from trespassing.
Sign systems and scripts, on the other hand, communicate much more. Road signs are not scripts, but the meaning they transmit keeps millions of motorists alive every day. Scripts are even deeper, because they express languages. In the absence of parallel texts like the Rosetta stone, they can be impossible for the human mind to decipher.
That’s why Tamil Nadu Chief Minister M K Stalin has handed off the battle over the cultural origins of Indian civilisation to machines. It’s a good bet, because a lot of people from the south are good with machines. Two of them head Google and Microsoft, and the ranks of coders the world over are thickly peopled with young southern talent. Meanwhile, Vignesh ‘Metakovan’ Sundaresan, who made a splash as a speculator in blockchain-based tokens, is following a more philosophical thread: asking what Big Data means for archives, history and everyday human experience.
Stalin has made a grant of Rs 2 crore to the Indus Research Centre of the Roja Muthiah library in Chennai, where a collection goes right back to the Bibles printed off India’s first Gutenberg press at Tharangambadi. The grant will support a chair honouring Tamil epigraphist and former Dinamani editor Iravatham Mahadevan, whose name is as celebrated as that of Finnish Indologist Asko Parpola in the small world of deciphering the Indus script.
Mahadevan was associated with the library and did significant work in deciphering Tamil-Brahmi inscriptions. He had reported similar sequences of signs in late Bronze Age Indus Valley inscriptions and Brahmi inscriptions from the south in the megalithic Iron Age. Other scholars had noted non-Brahmi graffiti from the megalithic south which remain unintelligible, and argued these were survivors of the Indus script. Such data could be a starting point for new research using pattern-seeking software.
For this noble purpose, Stalin offers a jackpot of $1 million. A small price to pay to win the culture wars that have been sweeping India since the Indus Valley began to be excavated. Now, north India is celebrating Rakhigarhi in order to own the civilisation, while the south proposes Keezhadi as a more compelling argument.
But there’s a problem: where’s the ancient data on which machines can be trained? Machine intelligence has been helping archaeologists to cut through language barriers. Damaged texts read like new by the intelligent interpolation of missing characters. In 2023, a computer science student in the US won the Vesuvius Challenge, which offers cash prizes for deciphering scrolls from Pompeii and Herculaneum scorched by the eruptions from Mount Vesuvius in 79 ACE. Too fragile to even touch, they were from the library of an Epicurean scholar named Philodemus. From scans of one scroll, the neural network extracted the Greek word porphyros―a purple dye from shellfish treasured in Roman times. We still use the phrase ‘born to the purple’ to indicate the opulence of the region, where the resorts of the Roman rich were located.
Since 2022, the Ithaca neural network, based on Google’s DeepMind, which was trained on 78,608 ancient Greek inscriptions whose provenance is reasonably certain, has been unravelling Greek epigraphy at a fast clip. Lakhs of Akkadian inscriptions dating back to 3,500 BCE have been discovered in Mesopotamia, but there aren’t enough scholars alive to decipher even a fraction of these vast records of an empire. The solution is being provided by Israeli software that translates directly from cuneiform script.
These digital interventions will change our understanding of the remote past as profoundly as genetic and biological methods have. But the Indus Valley script and language present a special problem: there’s very little textual data to train machines on. Correspondences like the Iron Age parallels noted by Mahadevan could help. Any effort to root the subcontinent’s origin story in facts deserves support. Any effort that’s finally, legitimately announced by the momentous words: “I have it!”
(Views are personal)
Pratik Kanjilal
For years, the author has been speaking easy to a surprisingly tolerant public
(On X @pratik_k)