They're known as the Herculaneum Scrolls, an ancient collection of manuscripts that were carbonized and buried during the eruption of Italy's Mount Vesuvius almost 2,000 years ago.

Discovered in 1752, the scrolls were deemed unreadable and remained so until recently when a University of Kentucky computer science professor led an effort to develop software that could virtually unwrap the scrolls.

We're unlocking the secrets of the distant past with a little help from technology from the future.

█ █ █ █ █ █ █ █ The scrolls from Herculaneum have an incredibly interesting history.

They were formed at the explosion of Mount Vesuvius in AD 79, and this cataclysm is well known mainly because of Pompeii, which is the birth of modern archaeology.

We know about the Pompeii archaeological site, but the scrolls were at an adjacent town in a library, and the explosion of the volcano actually preserved that library.

How did the papyrus survive something like that?

Well, normally papyrus would decompose over a fairly short period of time, but because the scrolls in the library were covered up with the ash and the gases and the dirt and the mud from the explosion, they were ironically preserved by that very process that destroyed everything else.

When the dust finally settled, Herculaneum and its libraries had been completely buried, a time capsule that would remain hidden for nearly 1,700 years.

They were discovered archaeologically 250 years ago when some people were trying to dig a well, and they basically hit not only the library but floors of mosaics that were from the Roman period, and they discovered that there was an entire town underneath the area.

The area pretty much was a very wealthy area, so wealthy Romans sponsoring philosophical groups, discussions.

So, in terms of its significance, it's just a center that's copying books.

They're preserving books.

So, as a library, it's preserving copies of very, very, very ancient literature.

So, when you think about the eruption of Vesuvius, every work that we have predates the eruption.

So, it's anything written before 79 going as far back as you can imagine.

A treasure trove of knowledge spared from destruction.

Could they be deciphered?

It's a tantalizing puzzle that Seales says was impossible to resist, but getting access was not easy.

The scrolls have always been at the center of a kind of currency, sort of the coin of the realm, if you will, for cultural artifacts that are highly regarded, venerated really by the royalty of Italy.

So, it's not like you can just go to the library and say, “I would like to do this experiment with your treasures, right?” So, there's sort of a political and a cultural question there about how to make that happen.

How did you navigate that?

I didn't navigate it very well at the beginning.

I made approaches to the people who held the collection and we were declined.

So, I had to build up the credibility over a long period of time with other work that we did.

And I had to show those things succeeding in order to be able to eventually work on this material.

We'd done enough work that we were gonna get an answer that was gonna be yes.

And I mean, that was incredibly important to be able to walk through the doors of the library and have the conservator hand me a scroll and say, “We would like you to work on this.” What was that moment like?

I mean, it was incredible because this material is really challenging and it's fragile and it's inaccessible.

It is not easy to be able to even be in the same room.

And then think about how that feels.

It's 2,000 years old.

This scroll existed maybe before Jesus walked the earth, right?

And you're in the same room with a document that actually has writing from a human, right?

That no one has read.

Early when they were discovered, and this is back in the 1750s, 1800s, scholars in that moment hoped that they would discover something new about the first century and about the classical period.

So, there was already a huge context when I came to this problem as a computer scientist.

What I discovered is that there were still scrolls that were not unwrapped.

They're completely intact, like a book on a shelf that had never been opened.

And so, what attracted me was the technical challenge of being able to read that book without opening it.

And so, what we did is we engineered a method to see inside and read what's inside without opening them.

And we do that with X-ray, but in a more sophisticated way than had been applied before.

We had to sort of engineer a new algorithm to use X-ray to be able to do what we call virtual unwrapping.

Seales and his team developed software that would examine these scroll fragments and use artificial intelligence to decipher the markings embedded within.

In this particular case with Herculaneum, AI has really been crucial in virtual unwrapping.

I mean, we have two really open problems that we've solved using AI.

One of them is to follow the layers of the papyrus because they're all tangled up and they're crushed.

You might think, “Well, it's a scroll.

I know what the cross-section is gonna look like.

It's gonna be like a jelly roll.

It'll be a cylinder that spirals into the middle.” That's actually not what it looks like.

The papyrus is damaged.

It's crushed.

Places are broken.

And we needed AI-inspired approaches to be able to follow those layers and pull them out.

So that really got you over the hub, if you will?

That helped us with the geometry of taking a scroll and virtually unwrapping it into something that was flat.

Right.

Ok. AI helps us do that.

But the other piece of the AI is that the ink is not very visible in the way that we have to scan using X-ray.

Right.

The evidence is there, but not to the naked eye.

Normal people looking at these images are not gonna see the evidence of the ink from Herculaneum.

But the AI can be trained to see it and then make it more visible.

And that's what we did.

At the end of the day, it really just needed more access to more data and more data and more data, more actual Herculaneum fragments.

All it's doing is it's just going over.

It learned the pattern of what's papyrus and what's ink.

And so, it just identifies ink, ink, ink patterns, ink patterns.

And then it just basically amplifies them.

But it needed that learning to progress in order to get to kind of where it is.

The method was very simple, but it needed a massive amount of training data.

You need actual carbonized fragments for it to train and learn on and getting things to work at scale, not just in a little microcosm of just a few bits and examples.

Enter the Vesuvius Challenge, a crowdsourced approach to process these fragments at scale in a worldwide contest with $1 million in prizes awarded to those who could be among the first to decode the Greek text hidden within the layers of carbonized scrolls.

When you look at the scale of the problem, it's really difficult to read a 100 scrolls with a few people working in the lab.

And so, at the right moment, we realized an opportunity for scaling up could be found through building a contest.

And so, with partnership in Silicon Valley, we built a contest framework where we could entertain contestants from all over the world who would help us do the work of virtually unwrapping a scroll.

That contest led to a few prizes that we were able to award, and one of them we called the First Letters Prize.

An incredibly high moment happened last year right here at the University of Kentucky, where one of the contestants found the first word ever read from the inside of a particular scroll that we were working on.

And that word, clearly visible, was purple.

As a first word, it's really pretty unique.

In the ancient world, the color purple was meant for royalty, it meant riches.

We thought immediately this might be a work that talks about wealth, talks about comfort, maybe a little bit of pleasure, pain, not sure.

But as a first word, it was a very sort of rich piece of vocabulary.

That word was just the uncorking of many, many words that came out of that scroll.

Ultimately, the grand prize was awarded for 16 columns of text, 500 or 600 words, that gave us a pretty good indication that the text is about Epicureanism and the philosophy around pleasure, pain, and scarcity.

It was a very human discussion about what does it mean to live a life that is responsible and that is full?

What does it mean to seek pleasure?

Is that what we should be doing as humans?

What brings us pleasure?

Does food bring us pleasure?

Music?

If things are scarce, are they more pleasurable?

These are all questions that are being asked in this text that we exposed using virtual unwrapping.

█ █ █ █