Researchers are now leveraging artificial intelligence to read charred ancient scrolls, unlocking whispers from a distant past. When Mount Vesuvius erupted in AD 79, it scorched the library of a luxury villa in Herculaneum and buried the city of Rome under a thick layer of ash and pumice. The catastrophic event threatened to erase countless texts, yet nearly two millennia later, scientists trained in machine intelligence managed to glimpse the first legible word from one of these fragile scrolls. This breakthrough, reported in relation to Guard, signals a new era of paleography where technology and humanities meet to revitalize ancient manuscripts.
The milestone was announced by Brent Seals, a computer scientist based at the University of Kentucky. In March 2023, Seals and his team unveiled the Vesuvius project, a competition designed to accelerate the decoding of characters from charred parchment. Supported by investors with deep roots in Silicon Valley, the challenge offers substantial cash prizes to researchers who can extract readable words from the complicated scrolls. The program aims to push the boundaries of what is possible when AI systems are trained to recognize the subtle traces of ink that survive on papyrus and parchment under extreme conditions (attribution: University announcements; program sponsors).
“This marks the first time a text has been recovered from one of these rolled, intact scrolls,” remarked Stephen Parsons, a research fellow within the university’s Digital Recovery Initiative. Since that initial success, teams have identified additional letters from the same ancient artifact, widening the scope of what can be retrieved from a single scroll and hinting at broader possibilities for the entire collection.
In tandem with the breakthroughs, Seals and colleagues released thousands of three-dimensional X-ray scans of two scrolls and three papyrus fragments, along with an AI model trained to interpret the faint signals left by ancient ink on papyrus. The technology relies on discerning minute structural changes in the plant-based material caused by the distribution of ink, enabling a reading path that was previously unthinkable for charred documents (citation: AI research consortium).
The unopened scrolls are among hundreds that form part of the Institute of France’s holdings in Paris, discovered in a villa library long associated with a prominent Roman statesman, potentially Lucius Calpurnius Piso Caesoninus, Julius Caesar’s father-in-law. The provenance of these texts adds a layer of historical intrigue to the scientific achievement, underscoring the intersection of cultural heritage and cutting-edge technology (institutional archives; historic attribution).
Two computer science students helped refine the search process and independently uncovered the same ancient Greek term within a scroll: “πορφύραc,” interpreted as “purple.” The winning word, identified by Luke Farritor and Yusef Nader, carried a prize of $40,000 for Farritor and $10,000 for Nader, highlighting how crowdsourced collaboration can accelerate discovery in high-stakes research (award records; project notes).
With early word reads in hand, the challenge now shifts toward completing the full text. The Naples II segment of the project has yielded progress, with papyrologist Federica Nicolardi and colleagues reporting that three lines of the scroll, containing up to ten letters, are now readable, and additional lines are anticipated as AI techniques and imaging continue to mature (portions of scholarly commentary; research updates).
These developments come amid broader debates about ancient caffeine myths and other long-standing questions in the field, illustrating how modern tech is reshaping our understanding of historical artifacts. As researchers continue to pair advanced imaging with AI-driven interpretation, the method stands to transform the study of illuminated manuscripts, papyrus, and wax-sealed scrolls, offering new pathways to preserving and understanding human history for contemporary audiences in North America and beyond (contextual analysis; cultural implications).