LONDON — Researchers are optimistic that a combination of artificial intelligence and human knowledge can unlock the secrets of ancient scrolls that were carbonized following a volcanic eruption nearly 2,000 years ago.
These papyrus scrolls were discovered in the 1750s at a luxurious villa in Herculaneum, a Roman town that, along with its neighboring Pompeii, was buried by the eruption of Mt. Vesuvius in A.D. 79. The villa, known as the Villa of the Papyri, contains a library that could vastly enrich our understanding of ancient ideas if the scrolls, which have been compressed into shapes comparable to candy bars, can be opened and read.
The intense heat and volcanic debris caused the town’s destruction while simultaneously preserving the scrolls in a state that makes them nearly illegible. They have turned into delicate, charred masses that cannot be physically unrolled without crumbling.
For over 250 years, scholars and scientists have explored various methods of deciphering these scrolls, most of which are kept in the National Library of Naples. In 2023, a group of tech entrepreneurs initiated the “Vesuvius Challenge,” providing financial incentives to harness machine learning, computer vision, and geometric techniques aimed at reading the scrolls.
Recently, the challenge reported a “historic breakthrough,” revealing that researchers had successfully created the first detailed image of the interior of one of the three scrolls housed at Oxford University’s Bodleian Library. Brent Seales, a computer scientist from the University of Kentucky and co-founder of the Vesuvius Challenge, expressed excitement over the successful imaging, noting that this scroll reveals more recoverable text than has previously been documented from Herculaneum scrolls.
The scanning process was facilitated by Diamond Light Source, a facility in Harwell, near Oxford, that employs a synchrotron particle accelerator to produce highly potent X-rays. Following this, scientists utilized AI technologies to compile the images, identify ink patterns signifying text, and enhance the overall legibility of the content. This innovative approach resulted in a 3D representation of the scroll, allowing for its virtual unrolling through a technique known as segmentation.
Despite the advancements, the capabilities of AI remain limited, and only a minimal amount of text has thus far been deciphered. One of the few identified words is from ancient Greek, translating to “disgust.” Scholars are being invited to contribute to the efforts aimed at comprehensively interpreting the texts.
According to Peter Toth, the Cornelia Starks Curator of Greek Collections at the Bodleian, it is important to recognize that this project is still in its infancy. He noted the need for superior imaging techniques while expressing confidence in improving both image quality and text clarity.
Toth also hopes that the technology developed could be accessible locally, negating the need to transport the other two fragile scrolls to Diamond’s facilities. He speculated, “Perhaps a portion of the equipment can be relocated,” and importantly, he reminded that there are still around 1,000 additional scrolls located in Naples that remain untouched.
