Researchers have used artificial intelligence and advanced scanning technology to read the complete text of scrolls from Herculaneum, a Roman city buried by Mount Vesuvius in 79 AD. The breakthrough recovers philosophical writings lost for nearly two millennia.

The charred scrolls, preserved in the volcanic ash that destroyed Pompeii and Herculaneum, had resisted all previous attempts at reading. Their carbonised state made physical unrolling impossible without destroying them. Scientists employed X-ray phase contrast tomography, a scanning method that reveals density variations within the brittle papyrus, combined with machine learning algorithms trained to recognize ancient Greek letters.

The recovered texts include works by Epicurus and other Stoic philosophers whose writings survived only in fragments or through quotations by later authors. The full scrolls provide unfiltered access to their original arguments on physics, ethics, and the nature of pleasure and pain.

This accomplishment represents a leap forward in recovering lost knowledge from antiquity. Herculaneum's library contained hundreds of scrolls, many still unread. The AI-powered approach could unlock thousands of additional texts that conventional methods cannot access.

The research builds on earlier progress. In 2019, researchers read partial text from a single scroll using similar technology. This latest effort demonstrates the technique now works reliably on multiple documents at scale. The machine learning models learned letter patterns from the few previously deciphered sections and applied that knowledge to new scrolls.

Limitations remain. The AI struggles with rare letters and unusual formatting. Damaged areas still block complete reconstruction. The process requires specialized equipment and computational resources beyond what most institutions possess. However, the methodology proves transferable to other collections of ancient carbonised documents found at archaeological sites worldwide.

The discovery underscores how modern computing transforms access to ancient texts. Rather than relying solely on physical restoration, digitization combined with artificial intelligence opens