Give me a sense of scope and scale.

Sure.

I did not know about this collection as a student.

And it wasn't until we started developing this particular technology that I started searching around for, you know, likely candidates.

And Herculaneum came to the top immediately because it's iconic.

I mean, it was formed, the library at Herculaneum was formed by the volcanic eruption of Mount Vesuvius in AD 79.

And most people are familiar with the story of Pompeii.

They argued with

And can I ask what the scroll actually says?

Well, we've read several now.

One of them is, uh, a Philodemus work with the title and author extracted from the very center of the scroll.

A, a real miracle, if you will, because the title and author are written at the very, very core of the scroll.

And of course, we're not opening it, so we're reading it without opening it.

But, um, what we've discovered there is Philodemus wrote a book on gods, and it happens to be book eight.

So there's a series of at least eight books by Philodemus on gods.

And h- how many scrolls have you got at the moment from Herculaneum that you can use this on?

Also in this podcast...

We're using X-rays to be able to see inside, and then software, including AI, to be able to actually read the evidence of the ink from the interior.

Brent Seales is a professor of computer science at the University of Kentucky and one of the founders of the project.

You don't have just the volcanic, you know, damage, but you have aging for two thousand years, the archeology.

And then the last two hundred and fifty years, I mean, bless everyone's heart, but people did the best they could at, you know, trying to hold off, and not everything got preserved.

I mean, a lot of things are fragmentary or even completely destroyed.

So we're operating on three or four hundred of the worst of the worst, which is why we're using X-rays to be able to see inside, and then software, including AI to be able to actually read the evidence of the ink from the interior.

Um, you can map the entire internal structure using what's called X-ray tomography.

[chuckles]

Sure.

Um, you can map the entire internal structure using what's called X-ray tomography.

It's essentially a three-dimensional map of everything inside the scroll, but at a very, very fine resolution.

And once you get that, you can see where all of the folds are, not just the outer ones, right, but the interior wraps, and then the software lets us localize those folds and unwrap them.

I mean, something's charred to a cinder a couple of weeks ago, but this is two thousand years old.

Well, it is.

I mean, you don't have just the volcanic, you know, damage, but you have aging for two thousand years, the archeology, and then the last two hundred and fifty years, I mean, bless everyone's heart, but people did the best they could at, you know, trying to hold off, and not everything got preserved.

Well, Brent, do you have anything else to add to what Nat said?

Well, the artificial intelligence part, which Nat Field deployed at GitHub co-pilot, which was one of the first of aids that the world saw in coding anyway, right, related to this whole AI large language model surge that we now see.

And so he was the right person to help me extend toward a kind of AI approach that would help us make that ink that's so hard to see more visible.

And so there's this enhancement problem that was kind of unsolved and open.

And being able to use the contest to really push hard on that turned out to be a key that opened up the next phase of the project.

And so AI has really been, I would call it analytical AI, not generative AI, but the ability to enhance evidence of the ink that's not visible to your naked eye, but is still there.

That evidence is still there.

And so that's, yeah, so we need help.