Artificial intelligence (AI) is accelerating new discoveries in every industry, from video production to medicine, and is currently expanding what we know about ancient times.
On Tuesday, researchers at the University of Pisa in Italy announced that they had successfully used AI to decipher a papyrus scroll discovered in Herculaneum, a town near Pompeii. Herculaneum was also destroyed in the eruption of Mount Vesuvius in 79 AD. This scroll is one of 1,800 volumes preserved. A Papyrus villa once owned by Julius Caesar's father-in-law, the estate was buried in mud and ash.
Also: AI breakthrough allows scientists to read Roman scrolls once buried in Mount Vesuvius
The scrolls are carbonized and too fragile for humans to touch, so hands-free scanning technology must be used to decipher them. According to ANSA, the researchers used infrared hyperspectral imaging and optical coherence tomography (OCT) to see through the charred papyrus.
By identifying and translating 1,000 words, about 30% of the scroll, researchers discovered the final resting place of the Greek philosopher Plato, a garden on the grounds of Plato's Academy in Athens. The document also reveals that Plato was sold into slavery in either 404 or 399 BC, not 387 BC, as historians had believed before Tuesday's discovery.
This release demonstrates the potential of this technology to further refine our historical knowledge of this period and its most prominent figures.
The discovery follows the February results of the Vesuvius Challenge, a global competition launched in March 2023 to decipher the entire Herculaneum scroll collection. Scrolls, the last complete libraries from antiquity, can yield a fascinating piece of history.
The project's main technology uses a combination of computed tomography (CT) scans and machine learning to decipher the content written on the scrolls without physically destroying them. ZDNET spoke with University of Kentucky researcher Brent Shields, one of the people behind the challenge, earlier this year to learn more about the project's findings.
Related article: How AI hallucinations can help develop life-saving antibiotics
So this process has three steps: scanning, segmentation, and ink detection. Researchers will create a micro-CT scan of the inside of the scroll, split the scan into individual pages, and use machine learning to decipher what is written on them.
The scroll and its ink have been carbonized by volcanic eruptions, so it's essentially black on black, making it difficult for computers to tell them apart. Researcher Stephen Parsons teamed up with Seales to train an ML model to read carbon text. The team then created Volume Cartographer, an open source software that understands text.
The researchers said they applied the model to full scrolls rather than fragments, created more “ground truth data” to improve the model's accuracy, and worked to uncover the patterns used to detect ink. He said he is still working on some challenges, such as reverse engineering the model. Go to the Vesuvius Challenge website.
Also: Are you ready to bring AI to your workplace? Google has a new course to help you get started
In this year's challenge, we will attempt to transcribe the entire scroll by scaling both the segmentation and scanning parts of the process. “Our goal for 2024 is to read 90% of Scrolls 1-4, and we are offering a $100,000 prize to the first team to reach this milestone,” the site says.
The University of Pisa's discovery proves that AI can discover things that cannot be known in any other way. Challenge researcher Michael McOsker told ZDNET that this approach could help uncover the equivalent of about 200 new books in the Herculaneum Library.
But the possibilities don't end there. The researchers also believe the technology could be applied to other fields, where machine learning could improve what CT scans and MRIs can detect, such as tumors in medical images. They are currently working on a virtual rendering of a medieval manuscript housed in the Morgan Library in New York.
