AI Deciphers Text in 2,000-Year-Old Herculaneum Scroll

Using X-ray imaging and machine learning, researchers have virtually unwrapped a charred Roman scroll, revealing ancient Greek text for the first time in two millennia.

The scroll, known as PHerc. 172, was carbonized during the eruption of Mount Vesuvius in 79 CE and is part of a collection discovered in Herculaneum in the 1750s.
A high-powered X-ray synchrotron at the UK’s Diamond Light Source created 3D images of the scroll, which were digitally unwrapped using advanced AI techniques.
Researchers identified the ancient Greek word 'διατροπή,' meaning 'disgust,' appearing twice in the text, believed to be related to Greek philosophical works.
The scroll contains more recoverable text than any previously scanned Herculaneum scroll, with scholars confident they can eventually read it in its entirety.
This breakthrough highlights the collaboration between libraries, computer scientists, and classical scholars, combining AI and humanities to unlock ancient knowledge.