(Web Desk) : A University of Nebraska undergraduate student has made history by using AI to read a section of a 2,000-year-old scroll.
Luke Farritor was the first person to decipher a word from ancient scrolls as part of the Vesuvius Challenge.
The Vesuvius Challenge is a competition for people to use modern technology to decipher the secrets of ancient rolled-up papyrus scrolls that hail from an ancient library in the Roman city of Herculaneum and, as a result of Mount Vesuvius erupting in 79 CE, were fossilized into carbon and are incredibly fragile.
The Vesuvius Challenge announced during a press conference that the 21-year-old computer-science major had won the “First Letters” prize of $40,000, after successfully deciphering and reading more than 10 characters in a 4-square-centimeter area of a scroll.
Farritor became the first contestant to submit the required number of legible letters in the competition. During the press conference, he shared his excitement about the initial moment of spotting the letters.
“I saw these letters and I just completely freaked out,” Farritor said. “I freaked out, almost fell over, almost cried.”
“I took a screenshot. I immediately sent it to JP Posma, who sent it to everyone else. I sent it to my family. My mom called and she was like, ‘Hey, like this is the first thing that you sent me that really looks like the letters. This is really cool,’ ” he explained.
Farritor first made the discovery late one night and said he knew he needed to “improve” the photo. “I was like… let’s keep going until it got to something that looks a lot like the image you’re seeing today,” he said.
Since Farritor reported his findings first, he was awarded the main prize, however, the second-place winner, Youssef Nader, also discovered the word in the same area and was awarded the $10,000 cash prize.
Farritor’s use of AI to read letters from the Herculaneum scrolls came as they were considered too fragile to unfurl and, according to the Vesuvius Challenge website, they would “turn to dust” if handled improperly.
