Reading Scrolls Once Buried by Vesuvius
The developments we make in Deep Learning, Machine Learning, and Artificial Intelligence are extraordinary. I am enthralled at how this niche of the computer world benefits us and advances us in science and technology. On October 19, 2019, I had the opportunity to attend a lecture at the Getty Villa Museum in Malibu. At the lecture, I had another epiphany about Artificial Intelligence that gave me a completely new appreciation for this science of big data.
Of all the 840 museums and art galleries in my home town of Los Angeles, my absolute favorite museum is the Getty Villa. Nestled near the shoreline of Malibu, the Getty Villa is dedicated to the study of the arts and cultures of ancient Rome and ancient Greece. As I am currently writing a book about ancient Rome, I find myself there often. It feels like a second home. The Getty Villa is a re-creation of a Herculaneum private residence now referred to as Villa dei Papiri, once owned by Julius Caesar’s father-in-law Lucius Calpurnius Piso Caesoninus. It is referred to as the Villa Dei Papiri because the house contained a library filled with papyrus scrolls.
Today, Herculaneum is a beautiful little town located on the Bay of Naples, on the southwestern side of Italy. Herculaneum’s first known inhabitants included Sabini tribes from the sixth century BCE who spoke Oscan. By the fourth century BCE, the ancient Italic Samnites inhabited Herculaneum. Eventually, the area fell under Rome’s control around 200 BCE.
On October 17, 79 CE, Mount Vesuvius erupted, violently gushing forth a deadly cloud of gases that reached a height of 21 miles in the air. It was not just a cloud. The volcano savagely emitted rock fragments, pumice, hot ash, gases, and particles from within the mountain. Here is a dramatic video that does a marvelous job of recreating the damage to Pompeii during the eruption.
Though the eruption of Mount Vesuvius is often associated with Pompeii, the damage it created was far-reaching. The explosion of Mount Vesuvius covered Herculaneum with 30 meters of ash. The eruption was so massive that it lasted between 18 and 20 hours. For perspective, the eruptions of Mount Vesuvius released 100,000 times the thermal energy of the Hiroshima Nagasaki bombings. It is hard to believe that any human survived this initial eruption. However, people did survive, including Pliny the Younger, who was a historian, lawyer, and Roman magistrate.
We are grateful for all of the writings that Pliny wrote about this explosion because his works validated the discovery of Herculaneum in 1752 when excavators commissioned by the Bourbon royal family were digging for wells in that area. The French royal family then commissioned archeologists who stepped in and began their excavation. There, in the villa, the archeologists unearthed an area that was void of oxygen when the volcano erupted. There, in the oxygen-deprived area, archeologists found almost 1,800 carbonized scrolls. This revelation was interesting because before discovering that they were scrolls, many excavators thought that they were charred tree branches damaged from the eruption of the volcano. Sadly, many of the scrolls were thrown away in the act of unknowing ignorance.
The Getty talk on October 19 was fascinating. Three individuals who have done extensive work with these papyri spoke. Classicist Dr. David Blank at UCLA and Dr. Richard Janko of the University of Michigan discussed the efforts that they took to read the the layers of the papyri, and Dr. Brent Seales of the University of Kentucky discussed how he used advancements in artificial intelligence and MRI imaging to completely virtually unwrap the scrolls and read them. The Getty Museum live-streamed the program, if you want to watch it here:
Though the eruption of Mount Vesuvius in 79 CE was immensely destructive, this powerful act of nature ultimately preserved the history of Herculaneum literature.
However, the challenge at hand was how to read them. The volcanic ash and gases carbonized the scrolls, which made it impossible to see what was in the papyri because archaeologists and scientists could not unravel them. And so began hundreds of years of attempts to read what was inside of the scrolls.
In 1752, King Charles VII of Naples commissioned the study of these scrolls and hired curator Camillo Paderni to commence work on them. Paderni sliced the scrolls in half and carefully removed the layers to copy the readable text. The problem with this method is that it destroyed the scrolls. So, in 1756, a Vatican priest named Antonio Piaggio, who was also a conservator of ancient manuscripts at the Vatican Library, created an invention that would unroll the scrolls five millimeters a day. Called the Piaggio Machine, it was the first successful apparatus to unroll the treasured scrolls to see what text was inside the papyrus. The first scroll took four years to unroll. Once Piaggio and his team unrolled it, they quickly copied the text before the scroll decayed. Again, this process would have been perfect had it not destroyed the scrolls.
Later versions of Piaggio’s machine pulled the edge of the papyrus up using silk threads drew the leading edge of the papyrus up using silk threads. Once the papyrus would reach the top of the frame, they would cut the papyrus, draw it and store it. With this machine, Piaggio opened 17 papyri.
Father Piaggio died in 1796. In 1798, the King Ferdinand IV of Naples, who now possessed the royal scrolls, had to flee from the Jacobins. And so, he and his family escaped to their second home in Palermo to avoid persecution during the French Revolution. The papyri were to follow them. However, the boxes were never sent nor opened. When the Revolution ended in 1799, the boxes were supposed to go to a company owned by the King. However, instead, the boxes were sent to the dock of Portici to be received by Reverend John Hayter. The King’s administration had put Hayter in charge of unrolling and deciphering the scrolls. Hayter briefly allowed a Chemist Diatono La Pere to experiment on the scrolls with atomized gas. But, this process destroyed them as well. In 1802, Hayter decided to do the work himself. After reaching out to Piaggio’s workers, he unfolded and partly deciphered 200 papyri. After four years, Hayter had to pause his work due to the invasion of Napoleon’s army. You can now find pieces of these papyri, known as the “Oxford Facsimiles of the Herculaneum Papyri,” at the Bodleian Library at Oxford University in England.
In 1816, French mechanic Pierre-Claude Molard and French archeologist Raoul Rochette used a replica of Piaggio’s machine to unroll one papyrus. The scroll was destroyed due to their failed efforts.
In 1819, Humphry Davy used chlorine to unroll 23 manuscripts. Though the scrolls were partially destroyed, parts of the papyri were salvaged.
In 1877, the papyrus was taken to the Louvre, where they attempted to unravel it. Unfortunately, that was unsuccessful as well and left only a quarter of the papyri left.
By the 1900s, only 585 papyri had been completely unrolled. Only 209 of the 1,800 that they found were unrolled in parts.
In 1969, Marcelo Gigante founded the International Center for the Study of the Herculaneum Papyri. Gigante created the center to continue in its excavations and to read the schools. At this point, they were starting a new method called the OSLO method. That method required destroying the scrolls as well, as they picked apart the scrolls into tiny fragments.
This destruction continued until 1999 when a team from Brigham Young University lead by Susan Booras and Steve Booras began using ultraviolet light to read the scrolls using technology that NASA had developed to view distant and dark planets. This method proved to be very useful, and by June 4, 2011, 1,600 Herculaneum papyri had been digitized!
By 2007, Dr. Seale’s team from Kentucky, along with the Institute de Papyrology began analyzing the scrolls with x‑rays in nuclear magnetic resonance.
Two years later the Institute of France, along with the French National Center for Scientific Research imaged two Herculaneum papyri with micro-computed tomography in order to read the insides of the scroll. By doing this, the teams were able to estimate that the scrolls were between 36 and 49 feet long, a little less than the length of a bowling lane!
Using micro-computed tomography, in 2009, Seales headed up a team where they virtually unwrapped scroll, using 3D scans. It was the first radiograph of papyri. They had created a way to unwrap the scrolls virtually and collect the data from inside of these ancient papyri. From that point, reading and understanding the scrolls became a process of acquisition, segmentation, texturing, flattening, merging and visualization, and scholarship. The papyri no longer needed to be opened and destroyed.
By 2015, they were virtually unwrapping the scrolls. The process began with acquiring the three-dimensional scan of the deformed manuscript. This three-dimensional process includes obtaining a scan of a set of cross-sectional images of the internal part of the scroll. As a three dimensional scroll, the viewer can see the layers of the scroll. However, the viewer still cannot see the text. In order to see the text, they carried out the virtual unwrapping.
The virtual unwrapping used a process called segmentation. In this process, the software would move through the scroll and trace the shape of a single scroll wrap. The three-dimensional model would produce a flat segment of the scroll.
Then using that segmented portions, they implemented the next step called texturing, which would extract the ink from the data. That included running another pass through the scroll looking for bright pixels. The bright pixels represent dense material in the scroll, which indicates traces of ink consisting of iron or lead.
However, because the papyri were curved, they then had to flatten the digital image.
They repeated the process of segmentation, texturing, and flattening multiple times on one small section of the scroll. These processes of steps created multiple segments of the papyri, which could then be pieced together like a puzzle. Once they placed the segments together, they would merge them creating one combined image that showed the full text.
Then, using this process to gather the text, it created what Seales referred to as “word soup.” It became the perfect resource for gathering metadata and confirming the text. With this metadata, they were able to use artificial intelligence and machine learning to predict what the writing on the scrolls. Now, without cutting up the papyri, they could see the internal structure of the scrolls.
Some of these scrolls go back as far as the fourth and third century BCE, with many of them written by Epicurus. The discovery of the written word from thousands of years ago is exciting, indeed. It is fascinating, thought-provoking, and even funny. Zeno of Sidon wrote one particular scroll titled “In reply to Craterus’ essay ‘against Zeno’s essay on geometric proofs,’” which Janko of refers to as the oldest known reply to a hostile book review. Though there were several pieces of scientific and mathematical works, the majority of the collection is by Philodemus of Gadara, who was an Epicurean philosopher.
Finally, after two thousand years, we can use our advancements in technology to peek into the past. In my opinion, when Seales saw the text as depicted through machine learning it made the last two thousand years seem like a tiny spec of time in the grand scale of our histories. It was a moment when our most momentous histories converged with our most valued technologies. The past converged with the future. It was a moment that our world became that much more enlightened by the insight of our ancient philosophers and the brilliance of our current scientists. Across the span of two thousand years, two separate teams of brilliant minded individuals came together and opened up a new understanding of human nature, scholarship, philosophy, and mathematics. That is the power of history, and that is the power of science!