Tuesday, March 17, 2026

Update on the Herculaneum scrolls

As background, I've posted several times on the Herculaneum scrolls, here here and here.  That last link in particular is a fairly pain-free Youtube video about what the Big Deal is.

And a Big Deal it certainly is.  In short: when Mt. Vesuvius buried the Roman town of Pompeii in 79 AD, it also buried it's more prosperous neighbor Herculaneum.  One of the (very) rich Romans who lived in Herculaneum was likely the father-in-law of Julius Caesar, and had one of the biggest libraries in the Empire.  The extreme heat of the lava flow carbonized the scrolls (books).  Researchers have been using CAT scans to image the carbonized rolls and have been applying AI to "unroll" the scrolls virtually and distinguish between carbon-based ink and just plain old scroll carbon.  They are starting to read scrolls that have been lost for 2000 years.


 Like I said, this is a Big Damn Deal.

If this interests you, there is a must read essay on what's been happening over the previous 18 months, the progress that's being made, and the challenges that are still present.  This part is really, really interesting:

So the central question has shifted from whether text could be recovered at all to whether it could be done routinely. At the current pace, processing the full Herculaneum library would take several years. The Vesuvius Challenge Master Plan, published in July 2025, outlines a series of steps intended to compress that timeline. These include improved surface extraction, deeper automation, and tools designed to reduce manual intervention at every stage.

According to Schilling, the problem is not that current methods fail outright, but that they require too much human steering.

“It’s not as fast or effective or cheap as it should be,” he told me. “Right now, we have solutions that work but that require human input.” What researchers want instead is a “global optimal solution” — a system that can isolate papyrus surfaces, unwrap them, and detect ink reliably across many scrolls without constant correction.

We're not there yet, but people are starting to figure out how to get there.  And it looks like there are a bunch of scrolls that were entirely lost over time that we will be able to read:

These scrolls are believed to contain Greek prose that largely vanished elsewhere, including philosophical works from the Epicurean tradition that were rarely recopied because they conflicted with Christian doctrine.

Very, very cool


 

No comments: