A 21-year-old computer science student used AI to decipher the first word of an unopened ancient scroll that was buried in mud and ash during the 79 CE eruption of Mount Vesuvius.
Luke Farritor was able to discern the word porphyras (πορφύραc), meaning “purple” in ancient Greek, on the papyrus text that University of Kentucky computer science professor Brent Seales had “virtually unwrapped” using microtomographic scanning, a non-destructive computer imaging technique that uses X-ray technology to allow researchers to digitally see inside an object.
The scroll was excavated from the Herculaneum papyri collection — the largest surviving Greco-Roman library unearthed from the Villa of the Papyri site between 1752 and 1754. Believed to be the villa of Julius Caesar’s father-in-law, Lucius Calpurnius Piso Caesoninus, the site contained some 1,800 manuscript fragments comprising 1,000 texts. With little to no oxygen in the massive library, the heat of the pyroclastic flow from Mount Vesuvius carbonized the scrolls and other organic materials, preserving their contents for thousands of years in roughly 66 feet of mud and ash.
After the villa’s discovery, early attempts by researchers to physically unroll the incredibly fragile texts proved disastrous. But over the past two decades, Seales and his team of computer programmers at the University of Kentucky developed a way to digitally reconstruct and analyze the interiors of these unopened ancient scrolls without risking further damage. In 2016, the team used microtomography to read an ancient scroll found in the Dead Sea. At one point thought to be a piece of charcoal, the scroll was found to contain an excerpt from the sacred Hebrew Book of Leviticus.
“Computerized tomography (CT) coupled with machine learning can elicit structure and content from badly damaged material at a very high level of spatial resolution,” Seales told Hyperallergic, adding that while this technology can “be engineered for meaningful work in the lab and/or on-site,” there are limitations. The tool uses ionizing radiation, for instance, and “still requires objects to be unearthed, as it is not a substitute for the dig itself.”
After releasing his artificial intelligence software with thousands of X-ray scans obtained from two unrolled scrolls and three text fragments, Seales launched a global competition known as the Vesuvius Challenge in March of this year, as a way to pool the efforts of researchers around the world to help him decipher the ancient manuscripts. Sponsored by more than 20 Silicon Valley investors, the challenge offers participants cash prizes ranging in tens of thousands of dollars for unscrambling the X-rays to reveal the hidden texts.
Building on the efforts of Casey Handmer, a contestant awarded $10,000 for discovering a “crackle pattern” in the scans that was confirmed to be ink, Farritor was then able to train artificial intelligence on Handmer’s findings to identify readable text, making him the first person to read the scrolls in more than 2,000 years, according to the contest announcements.
Farritor submitted his research to Seales and his team of researchers, who gasped when they saw “they could immediately read the word porphyras, despite the letters being faint.”
Shortly after Farritor’s discovery, another participant, Youssef Nader, also found the same word through his own independent research. The Vesuvius Challenge awarded Farritor with a $40,000 First Letters Prize, and Nader with the second place prize of $10,000.
The challenge is not over yet, however, as the grand prize of $700,000 will go to the first contestant to identify four readable passages of at least 140 characters from inside two intact scrolls.