Think you can decipher an ancient scroll? You could win a million bucks

If you like a good challenge, here's one that could have you attempting to decipher words or symbols first written down in the days of ancient Rome and unearthed in the 1700s. 

"You don't often have a chance to work on a 300-year-old puzzle and to go, as a technical person, on an Indiana Jones-style archeology adventure," Nat Friedman told As It Happens host Nil Köksal.

"This is one of those rare chances."

Friedman is a tech entrepreneur based in Silicon Valley, and one of the people behind the Vesuvius Challenge — a quest to read what was written on pages of papyrus that were fused together when Mount Vesuvius erupted in AD 79 and now resemble pieces of coal. 

They're offering up a $1 million US cash prize to anyone who can "make history by reading an unopened Herculaneum scroll for the very first time."

The scrolls belong to the Institut de France, part of a trove of hundreds that were found in the 1750s in the resort town of Herculaneum near Pompeii, Italy, in the library of a villa that's believed to have belonged to the father-in-law of Julius Caesar.

But they were too fragile to open. 

"They look like flaky, extremely lightweight, very fragile pieces of ash," Friedman said. Just squeezing them lightly in your hand, he adds, would be enough to turn them to dust.  

Friedman's team is providing scans of parts of two intact Herculaneum scrolls and other data to anyone who wants to take part in the challenge. "We're providing all the best tools and techniques that we have, that have been developed for doing this. We're open-sourcing all that code and we're providing all the data," he said. But his team's not going to tell people how to do it.

"Really any approach they want to take that works, we're happy with."

Ink invisible to the eye

In recent years, computer scientists, led by Dr. Brent Seales at the University of Kentucky, have been able to use powerful 3D X-ray technology called tomography to produce images of what's inside ancient scrolls. They did it first in 2015 with the En-Gedi scroll.

But there's a challenge with the Herculaneum scrolls.

"The ink that was used on these papyri actually doesn't really show up on the X-ray," Friedman said. "You can't see it with the naked eye."

Scientists have been trying for years to read the text on the scrolls using machine learning and artificial intelligence.

"The team that's been working on this for the last several years has discovered that machine learning models can detect the ink. And so we have a proof positive that this can work," said Friedman.

He says machine learning can detect subtle patterns in the surface of the papyrus — for example, where maybe the ink has filled in the ridges or where the writing instrument has depressed the fibres of the papyrus.

'The dream is big' 

Friedman wants someone to apply those techniques and actually get to the bottom of what the words are. 

"The dream is big," he said. "The extreme upside case is there could be a lost work in here. There are famous books from ancient Rome that we don't have — that we hear about, that are referred to. Poems of Sappho, autobiographies that are lost, epic works." 

Or, he says, it could be something completely different.

"I think almost anything would be fascinating, even if it's just a grocery list. That would be so cool to see."

But the best case scenario, he says, would be to find anything that "upends the story we've built about history and the world. Anything that really challenges us to rethink things."  

The contest calls for the extraction of four passages from the scrolls, with each passage at least 140 characters long and with no more than 15 per cent of the characters missing.

"Someone pointed out to me after we launched that we're basically asking people to find four tweets in the scrolls," Friedman said.  

He adds that wasn't actually done on purpose, but it gives him a good laugh. 

"It's fun to take AI and build the future," Friedman said. "But it's also fun to use it to get a little window into the past."