I linked to the Vesuvius Challenge a couple of times in this newsletter, and follow the developments closely since its first announcement two years ago. The Vesuvius Challange is a prize granted to the first to develop a working AI-algorithm able to decipher ancient scroll that were burried when the italian vulcano Vesuvius erupted back in AD79, the same eruption which destroyed Pompeji.
Nearby that famed, doomed city of Pompeji there was the town of Herculaneum, roughly 20 miles away from Pompeji and located in the west (at 21h) 10 miles away of the Vesuvius, while Pompeji itself is located at south-east (at 16:30h) of the volcano, roughly 15 miles away, so Hercaluneum was hit first. The town was buried under 30 metres (98 ft) of volcanic material from pyroclastic flows.
And burried with that town was the Villa of the Papyri: "The villa's name derives from the discovery of its library, the only surviving library from the Graeco-Roman world that exists in its entirety. It contained over 1800 papyrus scrolls, (the so called) 'Herculaneum papyri'."
Those "bundles of scrolls were carbonized by the intense heat of the pyroclastic flows. This intense parching took place over an extremely short period of time, in a room deprived of oxygen, resulting in the scrolls' carbonization into compact and highly fragile blocks. They were then preserved by the layers of cement-like rock."
The Villa of the Papyri is the only ancient library to survive to the present day, and in one of those weird ironic turns of history, these papyri are the only to survive, precicely because they were preserved in their carbonized state and if Herculaneum and the villa wouldn't have been burried under volcanic debris, "these rolls never would have survived the mediterranean climate and would have crumbled or been lost".
With "over 1800 papyrus scrolls" suspected in that villa, only 800 of which are excavated until now, it is bound to be a goldmine for history and philosophy. Only one lost book from Aristoteles in it, and it's worth any effort and any cost, but there's a problem. You can't just take those carbonized papyri and unscroll them because then you get a pile of dust, as pictured in this image from an early atempt in 1865:
A breakthrough in 2015 introduced a new technique called virtual unwrapping, in wich x-rays and computer vision create scans of those ancient scrolls, but we’re still far away from being able to decypher them. The researchers who started the challenge scanned “two of those scrolls using X-ray CT at a resolution of 8 μm“, which is 8 millionth of a meter, or 8 thousandth of a millimeter. Each file has 5 terabyte of data while one scroll only contains 10000 letters. Those scans were done with the Diamond Light Source particle accelerator near Oxford. (Yes, you read that right, those burned, carbonized, ancient scrolls were imaged in an particle accelerator, which ranks vey very high in the ultimate awesome nerdstuff list of all times.)
The Vesuvius Challenge now was about developing an algorithm that can automatize the process of reading those X-ray CT scans and turn those ultra-HD-images into readable text. Casey Handmer, who did participate in the challenge and won the "First Ink"-Prize for his work, wrote about his approach to the challenge in his blog where you can find a ton of high res images of those X-ray CT scans.
They look like this:
To my non-papyrologists eyes, this looks like a lot of cracks and noise with some fuzzy patterns that may or may not be letters. Here's Wikipedia on the difficulties with virtual unwrapping and deciphering those X-ray CT scans:
These techniques, while successful at isolating the layers of the papyri, had difficulty detecting text clearly due to the complex geometry of the sheets, such as the criss-cross structure of the papyrus fibres and the sheets, pleats, holes, tears, and contamination from the extensive damage. One potential source of error might be the 3D volumetric scan itself or the flattening procedure used to read it since the algorithms are not able to perfectly prevent distortions in the reading of these papyri.
So the quest was to train an algorithm on ancient greek handwriting and then apply this algo to those scans full of scratches, holes, tears and damages.
Now, on "Monday, Nat Friedman, a US tech executive and founding sponsor of the challenge, announced that a team of three computer-savvy students, Youssef Nader in Germany, Luke Farritor in the US, and Julian Schilliger in Switzerland, had won the $700,000 (£554,000) grand prize after reading more than 2,000 Greek letters from the scroll."
Here’s those 2000 letters:
These letters read with the new AI-method likely came from Philodemus, an ancient epicurean philosopher who lived 110-35 BC. From The Guardian:
“It probably is Philodemus,” Fowler said of the author. “The style is very gnarly, typical of him, and the subject is up his alley.” The scroll discusses sources of pleasure, touching on music and food – capers in particular – and whether the pleasure experienced from a combination of elements owes to the major or minor constituents, the abundant or the scare. “In the case of food, we do not right away believe things that are scarce to be absolutely more pleasant than those which are abundant,” the author writes.
“I think he’s asking the question: what is the source of pleasure in a mix of things? Is it the dominant element, is it the scarce element, or is it the mix itself?” said Fowler. The author ends with a parting shot against his philosophical adversaries for having “nothing to say about pleasure, either in general or particular”.
There's 800 scrolls residing in library in Naples, with more than 1000 waiting to be unearthed now from the Villa of the Papyri: "As of 2012, there were still 2,800 m2 left to be excavated of the villa. The remainder of the site has not been excavated because the Italian government is preferring conservation to excavation, and protecting what has already been uncovered."
Historian Bob Fowler and papyrologist Richard Janko are “convinced that the villa’s main library was never found, and that thousands more scrolls could still be underground”. It will take decades to excavate them (if that's even possible) and x-ray those scrolls. This AI-breakthrough is only one small step in this historic, gigantic quest to uncipher these scrolls from the only ancient library surviving to this day.
But just imagine William Baskerville from Umberto Ecos Name of the Rose discovering Aristoteles' lost chapter on comedy from his Poetics in it:
William of Baskerville: (after finding the secret room of books in the tower) Adson, do you realize, we’re in the greatest library in the whole Christendom? (WHOOOHOOO!) How many more rooms? Ah! How many more books? No one should be forbidden to consult these books freely!
As i wrote in my review of Name of the Rose: “when William of Baskerville puts out a scream of joy when he discovers the treasure trove of lost books in the secretive library, i felt so understood“.
It looks like AI is providing a crucial step in finding "many more books" from ancient philosophers of yore, found in "secret rooms" burried under volcanic ashes, to be “consulted freely” by us mere mortals today. All made possible by cool young folks using particle accelerator scanners and AI-tech.
Sometimes living in the future feels great.
Further readings:
WSJ: The World’s Smartest Young Minds Just Cracked a 2,000-Year-Old Mystery: “(Luke Farritor) used some of the money to splurge on a gift for his brother: a vintage poster for the Talking Heads concert movie Stop Making Sense. With his share of the $700,000 grand prize, he says he’d like to take his mother to Paris. He’s also thinking about building an autonomous boat to circumnavigate the globe.”
Live Science: AI is deciphering a 2,000-year-old 'lost book' describing life after Alexander the Great
Nature: First passages of rolled-up Herculaneum scroll revealed