AI Just Tracked How Sanskrit Words Changed Meaning Over Centuries

A researcher named Tanay Agrawal has shown that modern AI tools — the same kind of math that powers search engines and translation apps — can successfully track how word meanings changed in Sanskrit, an ancient Indian language used in texts thousands of years old. He built a digital library of 2.7 million words drawn from four different historical periods and let the software compare how words were used in each era. The idea is simple: if a word starts showing up next to different words over time, its meaning has probably shifted.

Sanskrit is famously hard for computers. Words smash together, change shape, and stack up into giant compound terms, so the software first had to learn to chop sentences back into individual words. That's a big deal, because most AI language tools are built for English and other well-documented modern languages. This paper is proof the same approach can work on something far older and messier.

To check whether the AI was actually right, Agrawal compared its findings against centuries of human scholarship — dictionaries, commentaries, and academic work on how Sanskrit evolved. Of 21 meaning changes that scholars had independently documented, the AI agreed on 19. That's not perfection, but it's a strong signal that the method works rather than producing random noise.

Why should anyone outside academia care? Because huge numbers of historical documents — religious texts, legal records, scientific manuscripts — sit in languages that few people can read and machines currently can't process well. If AI can reliably track meaning across time, it could help translators, historians, and museums digitize and understand these archives far faster. It's a small step, but it points toward a future where no written culture is locked away simply because the language is hard.