AI and Open Source Crack a 2,000-Year-Old Mystery

In March 2023, the Vesuvius Challenge set an ambitious goal: decipher the Herculaneum Papyri, a collection of scrolls carbonized and buried by the eruption of Mount Vesuvius roughly two millennia ago. Because the artifacts are so fragile, any attempt to physically open them would reduce them to dust. The only route to their contents runs through advanced imaging and machine learning.

University of Kentucky computer science professor Brent Seals explained to NPR that the underlying technology—virtual unwrapping using tomography and X-rays—has existed for decades. A 2015 breakthrough enabled reading a scroll from the Dead Sea Scrolls collection, but the Herculaneum texts resisted the same approach. The ink simply did not show up in the scans. “We needed an AI-based approach to be able to see that ink,” Seals said.

Once high-resolution scans of the papyri were captured, technologists worldwide competed to analyze them. The winning team—Youssef Nader, Luke Farritor, and Julian Schilliger—never met in person until after the contest ended. They connected entirely through the challenge’s Discord server, each contributing a distinct piece of the solution.

Three Students, One Pipeline

Farritor had already won the First Letters Prize for spotting the first visible characters. He reached out to Nader after Nader was named runner-up in that category. At the time, Farritor was focused on segmentation while Nader had built AI-based ink detection models. Schilliger joined soon after, automating the segmentation process at scale—work he credits in part to GitHub Copilot.

“There are so many benefits to using GitHub as a student, from a free GitHub Pro account to the use of GitHub Copilot in Visual Studio Code,” Julian said. “For the Vesuvius challenge, I had to write pipeline most of the time. I’d write one piece of code that I needed to use to achieve the next goal and if I knew what I wanted to write, I could use the auto completion tool to help me write faster. It was a huge time-saver!”

The collaboration spanned time zones, requiring deliberate schedule overlaps. Nader recalled a message from Schilliger, sent a night before the submission deadline, reporting that his software was segmenting roughly 1,600 cm² of scroll material. The pair spent much of the night hunting for the scroll’s title. Nader also described the near-miss moment: after nearly giving up, he ran one final experiment in the morning, and it worked. The process involved tracing the ancient writer’s strokes and inferring letters from subtle ink deposits.

For papyrologists, the result was a milestone. Gianluca Del Mastro, who met Farritor at the First Letters Prize ceremony in Kentucky, said he expected someone older. “It made me realize we have entered a new world of information technology in which it’s possible to make new discoveries even if you are very young,” he said.

The team published their winning code on GitHub, viewing open access as a natural extension of the contest’s collaborative spirit. As Nader put it: “Housing our code on GitHub was the only thing that made sense so the community can continue to build and to have easy access to collaborate and push progress forward.”

From Contest to Career

After the win, Del Mastro’s team sponsored flights to Naples, Italy, so Nader and Schilliger could view the scrolls in person—the first time the two teammates had met face-to-face. At a conference there, they connected with the professors who had evaluated their work. All three winners remain in contact and hope to collaborate again.

The outcome also reshaped at least one career path. Schilliger accepted a full-time role at the Vesuvius Project, where he continues decoding scrolls and exploring the intersection of code and classical history. He is quick to frame the victory in context: “Youssef, Luke, and I won this grand prize, but this is only a small piece in the ongoing efforts to decode the scrolls. Lots of people have worked on this before 2023 and there is plenty left to be done in 2024.”