The story of the Herculaneum papyri is one of historical significance and scientific persistence. Since their discovery in the 1700s, these scrolls had been too delicate to unroll and read, given the risk of damage. That all changed when University of Kentucky computer scientist Brent Seales proposed an innovative approach using high-resolution scanning and AI.
The method required three steps: scanning, segmentation, and ink detection. A particle accelerator scanned the scrolls to capture fine details, which allowed the scrolls to be digitally “unwrapped.” Then, Seales’s lab created machine-learning models to detect faint ink patterns. This research laid the groundwork, but it took time to bring it into practice—until tech investor Nat Friedman approached Seales with an idea.
In a stroke of inspiration, Friedman suggested a competition. Drawing on his fascination with Roman history, he believed the Vesuvius Challenge could attract talented young minds like Farritor’s.
For Farritor, the competition was irresistible. As a SpaceX intern with an interest in archaeology, the chance to blend history and technology was too good to pass up. “Projects like these should have more incentives,” he said.
Breakthroughs in Teamwork and Technology
The Vesuvius Challenge officially began in March, drawing a crowd of over 3,000 participants, most of them students with a passion for AI and computer science. Prize funds incentivized collaboration by rewarding progress at various milestones, creating a cooperative environment. A single discovery—a faint crackling of ink on the scrolls—sparked a cascade of insights. Farritor used this breakthrough to train his machine-learning model, eventually detecting a word in Greek: “porphyras,” meaning “purple.”
This achievement made Farritor the first person in 2,000 years to read a word from the Herculaneum scrolls, winning him a $40,000 prize. His teammate, Nader, earned a $10,000 award and poured it into cloud-computing resources, fueling the team’s AI work.
As the competition neared its deadline, Farritor and Nader decided to collaborate with Julian Schilliger, a robotics student who had been experimenting with virtual scroll unwrapping. Their entry stood out from the rest, impressing experts who reviewed the submissions.
The result? The scrolls revealed discussions on music, food, and the pursuit of happiness—topics with universal appeal. “It’s amazing that these words, so ancient, are still relevant today,” Friedman said.
Future Possibilities
The Vesuvius Challenge is far from over. The next goal is to decode 90% of the four scanned scrolls, with another large prize attached. Scholars believe more treasures may lie buried, waiting to be unearthed by future advancements in AI.
For now, Farritor is excited to continue exploring the possibilities this technology offers. Recently, he took a job working with Friedman, starting the day after his team won the Vesuvius Challenge.