AI Unlocks Lost Roman Scroll Revealing Ancient Philosophical Treatise
Researchers have used AI-driven virtual unwrapping to identify the author and title of a charred scroll buried by Mount Vesuvius nearly 2,000 years ago. The scroll, recovered from Herculaneum, contains the ancient philosophical treatise On Vices by Philodemus. Advanced imaging and machine learning techniques enabled scholars to read the text without physically opening the fragile artifact, marking a breakthrough in recovering lost classical literature and advancing digital preservation methods.
For the first time, artificial intelligence has enabled researchers to identify the author and title of an ancient document sealed inside a charred scroll for nearly two millennia without physically unrolling it. This remarkable achievement centers on a scroll recovered from Herculaneum, an ancient Roman town buried by the catastrophic eruption of Mount Vesuvius in 79 CE.
Known as PHerc. 172, the scroll is part of a rare collection housed at Oxford’s Bodleian Libraries. Using high-resolution X-ray tomography scans combined with sophisticated machine learning algorithms, scholars virtually “unwrapped” the carbonized papyrus to reveal the text inside. The scroll was identified as On Vices, a philosophical treatise by the Epicurean thinker Philodemus. This work, written in the first century BCE, explores the nature of virtues and vices, offering guidance on living a virtuous life—essentially an ancient form of self-help literature.
The significance of this discovery extends beyond recovering a lost text. The scrolls from Herculaneum represent one of the only surviving libraries from the classical world, but their fragile, charred state has historically made physical unrolling impossible without destruction. AI-powered virtual unwrapping techniques now allow researchers to safely access these invaluable cultural artifacts.
The breakthrough began with the use of X-ray tomography in 2015 to scan ancient scrolls, creating detailed 3D images that could be digitally unwrapped. Researchers at the University of Kentucky developed the Volume Cartographer software, which uses micro-CT imaging to detect faint traces of carbon-based ink. Because the ink contains no metal, neural networks were trained to recognize subtle ink patterns on the carbonized papyrus.
These advances culminated in the 2023 Vesuvius Challenge, an open-science competition that crowdsources AI-driven decoding of unopened scrolls. Participants utilize convolutional neural networks and transformer models to reconstruct text, with the first word from an unopened scroll successfully read in October 2023. The project continues to refine these technologies to unlock more texts.
Importantly, AI serves as a powerful tool to highlight potential ink regions, but human scholars interpret the results to confirm coherent words and phrases. This collaboration between machine intelligence and human expertise is transforming classical studies and digital preservation.
Beyond recovering lost philosophical works, this technology establishes a scalable system for digitizing and decoding ancient texts, offering unprecedented insights into the classical world. It exemplifies how AI and advanced imaging can unlock historical knowledge previously thought inaccessible.
This pioneering approach not only preserves delicate artifacts but also opens new avenues for research in archaeology, history, and literature. As AI continues to evolve, its applications in cultural heritage preservation will expand, enabling discoveries that reshape our understanding of human history.
AI Tools Built for Agencies That Move Fast.
QuarkyByte’s AI-powered analytics and imaging insights can accelerate your projects involving complex data reconstruction and pattern recognition. Explore how our solutions enable breakthroughs like virtually unwrapping ancient texts, unlocking hidden knowledge, and preserving cultural heritage with precision and scalability.