A University of Nebraska-Lincoln student was at a party in August when he received a text that would help him decipher a nearly 2,000-year-old message.
In the previous few months, Luke Farritor had joined a worldwide competition to translate ancient Roman scrolls that were damaged by a volcanic eruption in A.D. 79. The 21-year-old computer science major had developed an artificial intelligence program to detect the charred Greek letters written on papyrus.
The text message he received at the party included an image from one of the scrolls. Farritor sat down in a corner to review the picture and uploaded it to his AI program before returning to the party. When he was walking back to his dorm room around 1 a.m., Farritor pulled out his phone from his pocket and was shocked at what he saw.
His AI program had detected about a dozen letters from the image.
“I was completely amazed,” Farritor told The Washington Post. “I freaked out a little bit, jumping up and down, yelling, screaming.”
Papyrologists later translated the Greek letters into a complete word — “porphyras,” an ancient Greek word for purple.
The Vesuvius Challenge, a project created by University of Kentucky computer science professor Brent Seales to decipher the Herculaneum scrolls, has since awarded Farritor $40,000 for his discovery. He is believed to be the first person in nearly 2,000 years to be able to read a part of the scrolls.
The scrolls are “something that people said you would never be able to read because it’s too hard to extract the text,” Seales said at a news conference Thursday, explaining that he has tried to decipher the letters for about two decades. “And yet today we’re talking about exactly that.”
Thousands of years ago, the Herculaneum scrolls were stored in a library near Pompeii that burned and was buried under volcanic mud and ash after Mount Vesuvius erupted in A.D. 79. More than 600 of the scrolls were found in the 1750s, but people worried that opening them would cause them to fall apart, Seales said. Historians think the scrolls — which now look like burned logs — might have belonged to Julius Caesar’s father-in-law, Lucius Calpurnius Piso Caesoninus, the University of Kentucky said in a statement.
In 2002, Seales said he and a team of researchers developed an X-ray-like computer program that allowed them to see inside of unopened documents. Seales had used the program to read passages from unopened Hebrew books and was optimistic the technology would eventually reveal the ink inside the Herculaneum scrolls.
In 2009, Seales visited one of the four European institutions that own the Herculaneum scrolls, the Institut de France, to scan one. But unlike other ancient inks that included metal, the ink on the Herculaneum scrolls was made of charcoal and water, giving it the same density as the papyrus and making it impossible to see via the computer program, Seales said.
Seales had encountered a roadblock. But as AI improved in 2016, he started building a program that could detect the presence of ink on documents, even those with the watery charcoal type. After Seales produced the clearest scan yet of a scroll in 2019, he and his team received grants from science and arts organizations and funding from Silicon Valley investors to bolster their studies.
Still, Seales thought he needed help. He and his investors launched the Vesuvius Challenge in March, offering more than $1 million to participants who could decipher images of the scrolls. Seales, 59, and his researchers provided their data and images of two scrolls through the challenge’s website.
Farritor, who was born around when Seales began his research, said he heard about the competition on a science podcast and was one of about 1,500 people to sign up. Farritor said he was interning with SpaceX between March and July but worked on decoding the scrolls at nights and on weekends.
As Farritor was working on his own attempts to unlock the mystery, Casey Handmer, an entrepreneur from California who also participated in the challenge, saw dark spots and lines on the scrolls and concluded in early August that those areas were actually ink. Handmer shared his findings with other participants, leading Farritor to hone his AI program on revealing letters from the portions of the scroll with ink. Soon, he saw parts of letters appearing — a sign that his program was close to being able to detect full letters from the images.
Farritor was at his friend’s house party in Omaha later that month when the Vesuvius Challenge’s organizers released another image of a passage. When he saw later that night that his AI program had detected about six letters, Farritor said he texted a screenshot of the passage to his mom and the challenge organizers to share the news. The next week, he said he continued using his AI program to uncover about 10 more letters.
Federica Nicolardi, a papyrology professor at the University of Naples Federico II, said at Thursday’s news conference that she was shocked when she saw Farritor’s images, though researchers didn’t know the context of the discovered word. The Greek characters “πορφύραc” were ultimately translated to the word “porphyras.” Nicolardi said the word could describe the color of clothes or refer to purple dye, which was valuable in ancient Rome.
But researchers said the biggest revelation is that, using modern technology, it’s possible that anyone can read the scrolls. The challenge is offering $700,000 to the first person to decipher four passages from them.
Farritor, who wants to work for start-ups after graduating in May, has been staying up late to study more scrolls. He already has an idea of where he’ll spend his $40,000.
“I’m going to buy more computers,” he said, “and win the grand prize.”
{Categories} Research / Studies,_Category: Applications,*ALL*{/Categories}
{URL}https://www.washingtonpost.com/nation/2023/10/17/herculaneum-scrolls-contest-translated-deciphered/{/URL}
{Author}unknown{/Author}
{Image}https://www.washingtonpost.com/wp-apps/imrs.php?src=https://arc-anglerfish-washpost-prod-washpost.s3.amazonaws.com/public/A54V337AHZBBJIFP25LIBQ4VGI.jpeg&w=1440{/Image}
{Keywords}{/Keywords}