AI Unlocks 2,000-Year-Old Vesuvius Scrolls, Revealing Ancient Philosophical Texts

An entire lost library sealed in ash for two thousand years may finally speak again.
Researchers have decoded five percent of one scroll using AI, with plans to unlock 90 percent of four scrolls by year's end.
Mark

So they won $700,000 for reading one word? That seems like a lot of money for "purple."

Mimi

Not just one word—that was Farritor's early breakthrough. By the time they won the grand prize, the three of them had collectively decoded about five percent of a single scroll. That's multiple passages, multiple columns of text.

Luke

But let's be precise about what "five percent" means. They've successfully read five percent of one scroll out of four that were scanned. And there are 800 scrolls total. So we're talking about a tiny fraction of the library.

Mimi

True, but the point is they proved the method works. They showed that AI can distinguish ink from papyrus in a way humans couldn't. That's the breakthrough.

Mark

What does the scroll actually say? Is it interesting?

Mimi

It's a philosophical text about pleasure and how to live well. The author—probably a philosopher named Philodemus—criticizes people who have nothing to say about enjoyment. It's witty, actually. He's throwing shade at his opponents.

Luke

We should note that the author's identity is inferred, not confirmed. The organizers say "probably Epicurean philosopher Philodemus," but that's based on the content and context, not on a signature or explicit attribution.

Mark

And the next step is to read 90 percent of all four scrolls by the end of 2024?

Mimi

That's the goal. If they hit it, they'll have a much larger sample of texts to work with, and they can start planning how to scale up to the full 800.

Luke

Which is ambitious. They've gone from five percent of one scroll to a target of 90 percent of four. That's a big jump. We don't know yet if the method will scale that cleanly or if new problems will emerge.

Mark

Why does this matter so much? It's old philosophy.

Mimi

Because we've lost so much of the ancient world. Only three to five percent of Greek texts survive. This library was buried by a volcano and preserved in ash. It's a complete snapshot of what people were reading and thinking in the first century. That's rare.

Luke

And it came from a villa that belonged to someone connected to Julius Caesar. The site is mostly unexcavated. So there could be thousands more scrolls down there.

Mark

So this is just the beginning.

Mimi

Exactly. This is proof of concept. If it works, they could unlock an entire lost library.

  • Eight hundred ancient scrolls have resisted every human attempt to read them for nearly two thousand years — crumbling at the touch, their ink invisible against blackened papyrus.
  • A million-dollar competition forced an unlikely convergence: a Berlin PhD student, a Nebraska SpaceX intern, and a Swiss robotics engineer each attacking the problem from different angles until their methods fused into a breakthrough.
  • AI pattern recognition did what human eyes could not — detecting microscopic differences in texture and density to distinguish ink from papyrus inside CT-scanned scrolls that were never physically unrolled.
  • Only 5% of one scroll has been decoded so far, but it already contains a philosophical text on pleasure and music, likely by Epicurean thinker Philodemus, from a villa that may hold thousands more manuscripts.
  • The next phase targets 90% of four scanned scrolls by end of 2024, with the ultimate ambition of giving voice to an entire lost library preserved — improbably — by the same disaster that destroyed it.

For two thousand years, the volcanic ash of Vesuvius held a library hostage — its scrolls blackened, brittle, and silent. Now, through the patient vision of artificial intelligence, three researchers have coaxed the first whispers of ancient Greek philosophy from carbonized papyrus, winning $700,000 and opening a door that catastrophe had sealed in 79 CE. What emerges is not merely recovered text, but the possibility of recovering lost dimensions of human thought — an Epicurean voice speaking again across two millennia of silence.

Three researchers this week claimed a $700,000 prize for using artificial intelligence to read a 2,000-year-old scroll charred beyond recognition when Mount Vesuvius erupted in 79 CE. The scroll belongs to a collection of roughly 800 Greek manuscripts — now held in Paris and Naples — that have resisted every attempt at decipherment for nearly two millennia. They crumble when touched. Their ink and papyrus are, to the human eye, nearly indistinguishable.

The Vesuvius Challenge, launched by computer scientist Brent Seales and GitHub founder Nat Friedman, offered prize money to anyone who could decode at least four passages of 140 characters each with 85% accuracy — using only CT scans of the scrolls' interiors, never unrolling them. The task seemed almost impossible.

The winning team brought together Youssef Nader, a PhD student in Berlin; Luke Farritor, a student and SpaceX intern from Nebraska; and Julian Schilliger, a Swiss robotics engineer. Working separately and then together, they trained machine-learning algorithms to detect subtle variations in texture and density that distinguish ink from papyrus. Farritor identified the first Greek word — the term for purple. Nader decoded several columns. Schilliger built a tool for precise three-dimensional mapping of the papyrus surfaces. Together, they unlocked roughly five percent of one scroll.

What they found was a philosophical text, likely by Epicurean thinker Philodemus, reflecting on music, food, and pleasure — and criticizing opponents, possibly Stoics, for having nothing meaningful to say about enjoyment. The scroll came from a villa believed to have belonged to Julius Caesar's father-in-law, a site still largely unexcavated.

Scholars estimate only 3 to 5 percent of ancient Greek texts have survived to the present. This library — preserved by catastrophe itself — represents a rare and extraordinary window. "It is the only library to come to us from ancient Roman times," said Federica Nicolardi of the University of Naples. The next phase of the competition aims to recover 90 percent of all four scanned scrolls by the end of 2024, with the longer ambition of reading all 800. An entire lost library, sealed in ash for two thousand years, may finally be ready to speak.

Three researchers this week claimed a $700,000 prize for using artificial intelligence to unlock text from a 2,000-year-old scroll that was charred beyond recognition when Mount Vesuvius erupted in 79 CE. The scroll is one of roughly 800 Greek manuscripts that were carbonized in the volcanic blast that buried the Roman town of Pompeii. For nearly two millennia, these papyri—now stored in Paris and Naples—have resisted every attempt to read them. They crumble when handled. They are blackened. The ink and the papyrus look, to the human eye, nearly identical.

The Vesuvius Challenge, launched a decade ago by computer scientist Brent Seales and GitHub founder Nat Friedman, offered a million dollars in prize money to anyone who could crack the code. They began by subjecting four scrolls to high-resolution CT scans, creating three-dimensional digital maps of the interior surfaces without ever physically unrolling them. The task was to decipher at least four passages of 140 characters each, with at least 85 percent accuracy. It seemed almost impossible.

The winning team consisted of Youssef Nader, a PhD student in Berlin; Luke Farritor, a student and SpaceX intern from Nebraska; and Julian Schilliger, a Swiss robotics engineer. Working separately and then in concert, they developed machine-learning algorithms trained to distinguish ink from papyrus by recognizing subtle variations in texture and density. Farritor was the first to succeed, identifying a single Greek word—the term for purple—which won him a preliminary prize. Nader followed by decoding several columns of text. Schilliger contributed a tool called Volume Cartographer that enabled precise three-dimensional mapping of the papyrus surfaces. Together, their work has revealed roughly five percent of one scroll.

What they found was a philosophical text, likely written by Epicurean philosopher Philodemus, discussing music, food, and the pursuit of pleasure. In one passage, the author criticizes unnamed ideological opponents—possibly Stoics—for having nothing meaningful to say about enjoyment. The scroll came from a villa believed to have belonged to Julius Caesar's father-in-law, a site that remains largely unexcavated and may contain thousands of additional manuscripts.

The implications are staggering. Scholars estimate that only 3 to 5 percent of ancient Greek texts have survived to the present day. These scrolls represent an entire library preserved by catastrophe—a window into philosophical thought, daily life, and intellectual culture that would otherwise be lost. "Some of these texts could completely rewrite the history of key periods of the ancient world," said Robert Fowler, a classicist and chair of the Herculaneum Society. Federica Nicolardi of the University of Naples called it "the start of a revolution in Herculaneum papyrology and in Greek philosophy in general. It is the only library to come to us from ancient Roman times."

The organizers have already set their sights higher. The next phase of the competition aims to recover 90 percent of all four scanned scrolls by the end of 2024, and to establish the foundation for reading the entire collection of 800. If the technology continues to improve at its current pace, an entire lost library—one that has been sealed in ash for two thousand years—may finally speak again.

Some of these texts could completely rewrite the history of key periods of the ancient world.
— Robert Fowler, classicist and chair of the Herculaneum Society
This is the start of a revolution in Herculaneum papyrology and in Greek philosophy in general. It is the only library to come to us from ancient Roman times.
— Federica Nicolardi, University of Naples Federico II
Vuoi la storia completa? Leggi l'originale su CBS News ↗
Contattaci Domande frequenti