Armed with AI, a trio of young students have accomplished in a single year what scientists couldn’t in 272 years.
Luke Farritor, a student at the University of Nebraska-Lincoln and former SpaceX intern; Youssef Nader, an Egyptian PhD student in Berlin; and Julian Schilliger, a Swiss robotics student in Zurich, deciphered text on a 2,000-year-old papyrus scroll, winning them $700,000 of prize money in the process.
The scroll is just one of hundreds that were first excavated in 1752. Named the Herculaneum Scrolls after the city they were found in, the collection is the only library left from the ancient world. But their secrets remained unknown for the last 272 years.
When Mount Vesuvius erupted some 2,000 years ago, the scrolls were burned and buried, making them difficult to understand. In reality, prior attempts to interpret them had limited success and damaged some of the scrolls.
Enter artificial intelligence (AI) technology. The smart student trio employed AI and 3D mapping to identify letter shapes in a small piece of the scroll. The Vesuvius Challenge, introduced in 2023, offers a $1 million prize to citizen scientists who can translate a portion of the scroll’s text.
What the scroll said
The newly translated sections are assumed to be from the Epicurean philosopher Philodemus and include his opinions on music and eating.
Brent Seales, a computer science professor at the University of Kentucky who scanned the ancient scrolls in 2009, stated that the Epicureans discussed the impact of abundance and scarcity on pleasure.
On Monday, Nat Friedman, a tech executive and former CEO of GitHub, who grew intrigued with the scrolls during the 2020 COVID shutdown and helped fund the challenge, said on X (previously Twitter): “Today we are overjoyed to announce that our crazy project has succeeded.” After 2,000 years, we can now read the scrolls.
First-prize winner for the second time.
Farritor had already won first place in the Vesuvius Challenge’s “First Letters Prize,” which earned him $40,000 last year.
He, along with second-place finisher Nader, used AI to be the first to read a single word from the scroll. The name was translated from Greek as “purple dye” or “purple clothing.”
Several months later, Nader and Farritor collaborated with Schilliger to translate full chapters.
They read an extra 11 incomplete columns, exceeding the contest’s four-column limit. They transcribed hundreds of words in all, receiving $700,000 as the main prize.
However, the students could not have done it without the hard work of Seales and his team of researchers.
Seales created software using CT images to realistically unravel scrolls.
He and a team of academics then smoothed out the photos so they didn’t stack on top of one another like onion layers, and then used machine learning to identify ink-covered areas.
Currently, researchers have only unrolled and deciphered approximately 5% of a single scroll. According to Nature, scholars hope to discover forgotten gems from Aristotle or Homer among the remaining 280 papyri.