Technology

Scientists Just Read an Ancient Scroll That Was Never Opened—Here's How

Martin HollowayPublished 2month ago4 min readBased on 2 sources
Reading level
Scientists Just Read an Ancient Scroll That Was Never Opened—Here's How

For the first time in roughly two thousand years, researchers have read a complete scroll from the ancient Roman city of Herculaneum. The remarkable part: they did it without ever opening the document. The scroll remained sealed, intact, and undamaged throughout the entire process.

This breakthrough came from the Vesuvius Challenge, a worldwide competition aimed at solving a specific problem. In 79 AD, Mount Vesuvius erupted and buried the city of Herculaneum under volcanic ash and mud. The library there survived intact, but the scrolls inside became rock-hard and nearly impossible to read. In past centuries, when scholars tried to physically unroll these scrolls, many crumbled to dust. So for nearly two thousand years, no one could read what was written inside.

Last year, a team participating in the challenge successfully extracted readable text from an unopened scroll—15 passages with more than 2,000 characters decoded in February 2024. That was significant in itself. But now, researchers have done something even more impressive: they have read an entire scroll from beginning to end.

How It Works

Think of the process as taking an extremely detailed X-ray photograph, then using artificial intelligence to find the invisible ink inside.

First, researchers use a special scanning machine called micro-CT—the same technology hospitals use for medical imaging, but much more powerful. This machine takes thousands of tiny X-ray pictures of the scroll from every angle and combines them into a detailed 3D map of the entire document. All of this happens without touching the scroll at all.

Next, computers use that 3D map to "virtually unwrap" the scroll. The software imagines what the papyrus would look like if it were gently unrolled, creating a flat digital image. But there's still a major problem: the papyrus is black carbon, and the ink is also black carbon, so they look nearly identical on an X-ray. The ink traces are so faint that humans cannot see them.

This is where artificial intelligence comes in. Researchers trained computer programs—called machine-learning models—to recognize the tiny patterns that ancient ink leaves behind on papyrus. Even though the ink is nearly invisible, it creates microscopic texture changes that the AI learned to spot. In earlier competition rounds, researchers fed the models examples of ink patterns, and the algorithms got better and better at detecting letters and words from what looks like noise to human eyes.

A 21-year-old competitor named Luke Farritor was the first person in two thousand years to successfully read a single word from inside an unopened scroll. This proved the method worked and helped the entire research community refine their approach.

What Happens Next

Archaeologists believe the Herculaneum library contains hundreds of scrolls that have never been read. Most of what we know about ancient philosophy from that time period comes from fragments or copies made centuries later. The discovery of new, complete texts could change our understanding of what ancient thinkers wrote.

The passages decoded so far came from an ancient philosopher named Philodemus and discussed the topic of pleasure. The content itself is not startling, but that is not really the point. The real significance is that the method works and can be repeated. If researchers can read one scroll this way, they can read the next one, and the one after that.

It is important to note that the system is not yet simple or automatic. Scanning each scroll takes a lot of time, and the resulting files are enormous. The artificial-intelligence models have only been trained on a small number of practice examples so far. And humans still need to carefully check the computer's work at several stages. Going from reading one complete scroll to reading hundreds of them will require more resources, better automation, and further refinement. The breakthrough announced now is a proof that the method works, not yet a routine process that is ready for mass use.

But the progress is clear. With each year of the competition, the technology gets faster and more accurate. From the first readable word to the first complete scroll took about two years. The next major milestone may arrive even sooner than that.

The scrolls of Herculaneum waited nearly two thousand years to be read. The tools to unlock them, it turns out, are arriving faster than anyone expected.