eccentric
The scroll, which was virtually unsealed with the help of AI, reads: “We are no longer inferior to them, as we have certainly pushed ourselves to the limit through research and learning.”
A combination of machine learning and high-resolution CT scans has finally revealed the sealed scrolls of the Roman town of Herculaneum, which were destroyed by the eruption of Mount Vesuvius nearly 2,000 years ago.
In 2023, researchers were able to decipher some words from the char and ashes that make up the bulk of the scroll. Some of the same award-winning researchers recovered more texts from one of the scrolls, PHerc.Paris.4, and won the grand prize of $700,000 in the Vesuvius Challenge contest in early 2024.
Another two years later, the grand prize winner is now part of the Vesuvius Challenge team that successfully read the remaining portion of the unrolled scroll from end to end, as the VC team shared in Thursday’s announcement and detailed in the attached paper. [PDF].
According to the research paper, the entire scroll was able to be identified thanks to high-resolution phase-contrast X-ray microtomography performed at the European Synchrotron Radiation Facility in France. This is an improved imaging technique over previous methods used to capture previous images analyzed in prize contests.
But that’s not all. Much of the research was successful, they say, thanks to the development of a new “workflow” that scanned the scrolls, detected ink on the charred papyrus, effectively “unrolled” the scrolls by modeling their deformed surfaces, and digitally preserved those surfaces. This allows the machine learning model to identify letters across the scroll, not just isolated patches.
“Thus, the important transition marked by the current study is from exceptional local recovery to systematic scroll-scale recovery,” the team wrote. In other words, if the characteristics of the hundreds of sealed scrolls recovered from the Papyrus Villa at Herculaneum, the world’s only fully extant library from antiquity, can be described, this could mark the beginning of an explosion of new material for historians.
So what did you say?
PHerc.Paris.4 wasn’t at the center of this breakthrough either, but we had some exciting news to share in this regard, which we’ll get to later. Instead, PHerc brought about a breakthrough. In 1667, the preserved text of a previously unread scroll was read from cover to cover for the first time. This work appears to be a treatise on Stoic philosophy, focusing on ethics, a favorite subject of Zeno, Marcus Aurelius, Epictetus, and their intellectuals.
A passage later in the scroll reads, “Indeed, having exerted our utmost efforts through study and learning, we will no longer be inferior to them in any respect,” and “we have likewise accomplished things worthy of them, and possessed the same practical wisdom as they.”
The perfect ancient wisdom to first see the light of the modern world.
The team digitally unfolded the scroll, detected its ink, and transcribed the preserved text that was part of the original PHerc from edge to edge. 1667 was lost long ago during early attempts to physically open the scrolls, before archaeologists had access to advanced X-ray imaging and AI-assisted analysis.
“The outer layer was destroyed during previous attempts to open it by hand in the 19th century and again in 1969 and in the 1980s,” the Vesuvius Challenge team said, noting that the original scroll was originally 19 to 24 centimeters tall, with only the eight-centimeter-high core remaining. Nevertheless, “this is the first time that the preserved text of the rolled Herculaneum scrolls has been read continuously from end to end, rather than in isolated words or patches,” the researchers said.
In addition to fully uncovering the remains of PHerc.1667, the team was also able to extract some information from several other documents using a new workflow. One of them, PHerc.139, turned out to be a copy of the eighth volume of Epicurean philosopher Philodemus’ treatise on God. This means that once the scroll is fully unfolded digitally, scholars can expect to know what they’re looking at.
The second one, as mentioned above, concerns PHerc.Paris.4. New high-resolution images taken for this latest experiment allow the words on the scroll to be directly visible for the first time. This means there is no need to rely on algorithmic detection of individual words or phrases from CT scans. Most importantly, Paris.4’s new scan results perfectly matched those determined by the Grand Prix team several years ago, providing independent confirmation that the award went to the correct team.
There are still challenges to deciphering and deciphering the rest of the ancient library, with the team pointing to geometric challenges in predicting surfaces that can make the unfolded scans illegible, and radiometric challenges that make identifying the ink difficult due to inconsistencies in ancient recipes. Still, this is a significant advance, and the team believes they are ready to scale the X-ray and machine learning workflows they developed.
“The ideas of the ancient world, sealed in darkness for 2000 years, are coming back into the light one volume at a time,” said the Vesuvius Challenge team. As for me, I can’t wait to see what ancient secrets they discover next. ®
