AI Unwraps Ancient Scrolls

A newly developed machine learning algorithm has, for the first time, decoded parts of a burned manuscript recovered from the ancient Roman city of Herculaneum. The breakthrough was made by 21-year-old Luke Farritor, a University of Nebraska computer science student, as part of a $1M competition to digitally restore the scrolls. A single full word, translated as "purple," was revealed.

Situated just west of Mount Vesuvius, the city fell victim to the same massive first-century eruption that buried nearby Pompeii. Like its neighbor, much of the city was preserved under a blanket of ash. Roughly 1,800 manuscripts were accidentally found in the mid-18th century—charred but intact—housed in a single villa potentially owned by the father-in-law of Julius Caeser.

Farritor was awarded $40K for being the first to identify a single word—the competition's $700K grand prize goes to the first team to decode four separate passages (read more here).