A 21-year-old undergraduate student at the University of Nebraska has won $40,000 (£32.8K) for being the first to read a word from one of the ancient Herculaneum scrolls as part of the Vesuvius Challenge — a competition with $1,000,000 (£821K) in rewards for those who can use contemporary technology to decipher the words of these scrolls.
When Vesuvius erupted in 79 AD, the scrolls known as the Herculaneum papyri, which were kept in the library of a private villa close to Pompeii, were carbonised and buried.
This ancient library's sole copy was hidden underground for nearly two thousand years beneath 20 metres of volcanic mud. They were unearthed in the 1700s, and although the eruption had in some respects preserved them, they were so delicate that, if handled improperly, they would crumble to dust.
For many centuries, the challenge has been to read these scrolls without actually opening them. This is where a number of brilliant students and artificial intelligence come in.
Researchers used artificial intelligence to peek deeply inside the delicate, charred remains and were able to retrieve the initial word from one of the inscriptions.
What is the Vesuvius challenge?
The Vesuvius challenge was launched in March to accelerate the reading of the scrolls, which are among hundreds found in the library of the villa believed to have belonged to a senior Roman statesman—possibly Lucius Calpurnius Piso Caesoninus, Julius Caesar's father-in-law. The unopened scrolls now belong to a collection maintained by the Institut de France in Paris.
Professor Brent Seales, a computer scientist at the University of Kentucky and his team published thousands of 3D X-ray photos of two rolled-up scrolls and three papyrus fragments to kick off the challenge. The project, which is supported by Silicon Valley investors, offers cash prizes to scholars who are able to decipher the language from the carbonised scrolls.
They also made available an artificial intelligence programme that they had trained to decipher the scrolls' letters based on the slight alterations the ancient ink had made to the papyrus' structure.
Luke Farritor from Nebraska and Youssef Nader from Berlin, two computer science students who took on the Vesuvius challenge, improved the search method and separately discovered the same ancient Greek word in one of the scrolls: "oc," which means "purple." Farritor, who discovered the word first, earned $40,000 (£32.8K), while Nader took home $10,000 (£8.2K).
These discoveries were inspired by Casey Handmer, who was the first person to find substantial, convincing evidence of ink within the unopened scrolls. He first discovered a “crackle pattern” which looked identical to ink by staring at the CT scans for hours.
After Handmer’s breakthrough, Farritor started developing a machine learning model on the crackle pattern. The model got better with each new crackle discovered, revealing more crackles in the scroll in a cycle of discovery and improvement.
Stimulated by Casey and Farritor discoveries, Nader then went over the top entries for the Ink Detection contest on Kaggle, which aimed to enhance Stephen Parsons' method of machine learning using isolated pieces.
To adapt these models to the scrolls, he employed a domain transfer strategy that involved unsupervised pretraining on the scroll data and fine-tuning the fragment labels.
What’s next for the Vesuvius challenge?
The race to decipher the surrounding text is what’s next for researchers.
Three lines of the scroll, each comprising up to 10 letters, were now readable, according to Dr Federica Nicolardi, a papyrologist at the University of Naples Federico II.
She expects that more lines will soon be readable.
There are at least four text columns visible in a recent passage.
