Chapter 1: The Hidden Treasures of Antiquity

The world is filled with ancient literature that appears to have vanished. While we may never recover all of this lost knowledge, some of it remains concealed, waiting to be rediscovered. However, accessing these hidden texts is no simple feat. The materials are often fragile, and specialized technology is essential. This is where artificial intelligence steps in to assist.

Section 1.1: A Glimpse into Roman Life

In 62 AD, a series of tremors shook a tranquil Roman town, but life resumed as normal. Unbeknownst to the townsfolk, the low mountain that provided their crops harbored a dark secret.

On a fateful afternoon in 79 AD, the world changed. Mount Vesuvius erupted, burying the city under layers of ash and pumice. Those who survived the initial collapse perished in toxic fumes. Pompeii was obliterated, while Herculaneum faced a brutal deluge of gases, ash, and steam. The once-thriving cities of the Gulf vanished from view, shrouded beneath ten meters of volcanic debris.

"Broad flames shone out in several places from Mount Vesuvius, which the darkness of the night contributed to render still brighter and clearer..." — Pliny the Younger

The Last Day of Pompeii. Painting by Karl Brullov, image source: here

For 17 centuries, the memories of these cities faded away until archaeologists began their excavations. Gradually, these ruins revealed a tragic yet vivid account of the final moments of these communities, offering a unique glimpse into Roman life.

At Herculaneum, a remarkable discovery occurred in 1752: an ancient library was uncovered within the Villa of the Papyri, containing 1,800 papyrus scrolls, many of which remain largely intact. Today, while over 300 scrolls are nearly complete, nearly a thousand are only partially preserved.

Chapter 2: The Quest for Deciphering the Papyri

In the video "Is AI Ruining Our Perception of History?", experts discuss the implications of AI in understanding ancient texts and its potential to reshape our historical narratives.

For two decades, various teams have sought ways to decode these scrolls without physically unrolling them. Brent Seale has utilized three-dimensional computed tomography (CT) to capture high-resolution images. His work with the Diamond Light Source particle accelerator has shown promise, though distinguishing ink from papyrus remains a challenge.

Seale successfully demonstrated this method by virtually unrolling scrolls from En-Gedi, a Jewish community site dating back to the late eighth century BCE. This process involved intricate transformations and preprocessing of the images, allowing researchers to read what turned out to be the oldest copy of a Pentateuchal text.

The charred scroll from En-Gedi (left) and completed virtual unwrapping for the scroll (right). image source: here

Despite these advancements, determining the optimal scanning protocol before identifying ink presence is complex. Each step requires meticulous calibration, and identifying the text is equally daunting.

In response to the challenges, "The Vesuvius Challenge" was organized to encourage community participation in developing sensitive algorithms capable of reading the scrolls. The competition offered significant prizes, motivating participants to innovate.

In a breakthrough moment last October, researchers detected faint traces of ink resembling Greek letters. This led to the creation of a machine learning algorithm designed to identify these textual nuances. Early findings included the word πορϕυρας (porphyras), meaning 'purple'.

In the accompanying video "Care to Learn the History of AI? Read Genius Makers by Cade Metz," viewers can explore the intersection of AI and historical research, shedding light on the ongoing evolution of technology in the humanities.

Despite the excitement, research continues. Teams submitted models to compete for the Grand Prix, uncovering hundreds of words within the papyrus. However, the text revealed only 15 columns and less than 5% of the entire scroll, indicating it contains previously unknown literature.

The winning design, featuring a small architecture with divided space-time attention, has been made available on GitHub, providing a valuable resource for future research.

What does this text reveal? Preliminary transcriptions indicate that the scroll is not a duplicate of any existing work. Instead, it presents an original text from antiquity, delving into themes of pleasure and abundance within Epicurean philosophy.

Parting Thoughts

Decoding a scroll charred for two millennia is an extraordinary technical feat, yet it also ignites curiosity about future discoveries. Researchers speculate about potential texts from Aristotle, Homer, or other forgotten authors. The exploration of the mansion where these papyri were found is still incomplete, suggesting more treasures may lie ahead.

These algorithms not only hold promise for the scrolls from Herculaneum but could also be adapted for other ancient texts, whether from different locations or wrapped around mummies, all while preserving these invaluable artifacts.

Are you intrigued by the potential of uncovering more lost literature? Share your thoughts in the comments!

For those interested in further reading, you can explore my other articles or connect with me on LinkedIn. I maintain a repository of weekly updates on ML & AI news and welcome collaborations. Additionally, visit my GitHub for resources related to machine learning, artificial intelligence, and beyond.