AI restores lost Roman writings, transforming digital archaeology

Google DeepMind has unveiled an artificial intelligence model called Aeneas that can decode and restore ancient Latin inscriptions from the Roman Empire with remarkable precision, according to research published in Nature this week. Named after the mythical Trojan hero, the AI represents a leap forward in digital archaeology and could transform how historians study the ancient world.phys

The system can estimate the geographical origin of inscriptions among 62 Roman provinces with 72% accuracy, predict when texts were written within approximately 13 years, and restore missing segments even when the length of gaps is unknown. Trained on the Latin Epigraphic Dataset containing 176,861 inscriptions worth 16 million characters spanning from the 7th century BCE to 8th century CE, Aeneas analyzes both textual and visual elements of ancient writings.bioengineer+2

Revolutionary Collaboration Between AI and Historians

A collaborative study involving 23 professional historians demonstrated Aeneas's real-world effectiveness. When working alone, historians achieved a 39% character error rate in restoring damaged texts. With Aeneas's assistance, that error rate dropped to 21%. The AI was deemed useful as a research starting point in 90% of cases and increased historians' confidence by 44%.marktechpost+2

"What makes them unique is that they are written by the ancient people themselves across all social classes on any subject. It's not just history written by the elite," said study co-author Yannis Assael, a researcher at Google's AI lab DeepMind.phys

The model excels at identifying "parallels" – inscriptions with similar wording, formulas, or cultural contexts – a process that traditionally requires extensive manual searches through vast library and museum collections. Thea Sommerschield, an epigrapher at the University of Nottingham who co-designed the model, compared studying inscriptions to "solving a gigantic jigsaw puzzle" where 90% of pieces are missing.phys

Practical Applications and Case Studies

To validate its capabilities, researchers tested Aeneas on the famous "Res Gestae Divi Augusti," Emperor Augustus's autobiographical inscription that has sparked scholarly debate about its composition date. The AI identified two possible dating periods – late 1st century BCE and early 1st century CE – precisely matching the ongoing historical debate.phys+2

In another case study, Aeneas analyzed a votive altar from ancient Mainz, Germany, correctly dating it to around 211 CE and identifying its origin in Germania Superior province. The system revealed previously unnoticed linguistic connections between this altar and earlier local inscriptions.bioengineer+1

Addressing Archaeological Challenges

With approximately 1,500 new Latin inscriptions discovered annually, the AI addresses a growing challenge in classical studies. These ancient texts, ranging from imperial decrees and legal documents to tombstones and graffiti, often survive as fragmentary remains damaged by millennia of weathering and human activity.phys

The tool is freely available to researchers through the Predicting the Past platform, with open-source code and datasets. DeepMind has also developed educational curricula integrating Aeneas into digital humanities classrooms. The model can potentially be adapted to other ancient languages and scripts, from papyri to coinage, expanding its archaeological applications beyond Latin epigraphy.marktechpost