Ancient Roman history is undergoing a transformation at the hands of a Deepmind AI.
Google DeepMind's groundbreaking AI, Aeneas, has made significant strides in the field of ancient Roman history by accurately restoring and dating ancient Latin inscriptions, including providing insights into the Res Gestae Divi Augusti.
Aeneas, named after the legendary hero who journeyed from Troy to Italy, analyses incomplete and damaged inscriptions by restoring missing text, predicting when and where the inscriptions were created, and identifying parallels across thousands of Latin texts in its extensive database.
In the case of the Res Gestae Divi Augusti, a chronicle of Emperor Augustus’s deeds traditionally undated in surviving inscriptions, Aeneas helped pinpoint its date to around 15 CE, shortly after Augustus’s death in 14 CE. This contributes to understanding its political implications by firmly situating the text historically, which is crucial for interpreting Augustus’s propagandistic presentation of his rule.
Aeneas achieves this by being trained on over 176,000 inscriptions from multiple epigraphic databases covering millions of characters. This allows it to detect historical patterns across time and geography rather than relying on simple phrase similarity. It can attribute inscriptions to specific Roman provinces with 72% accuracy using both text and visual data.
The AI’s generative modeling transforms epigraphy research from partly guesswork and hunches into more precise historical inquiry. Aeneas improves the precision and speed of its analyses, what once took weeks of comparative database consultation now takes minutes.
Moreover, Aeneas detects orthographic clues, syntactic structures, and common linguistic formulas in different regions of the Empire. It also processes images of inscriptions in addition to plain text, suggesting the number of letters missing in a fragment and the expected shape in stone.
Aeneas was validated with over 176,000 Latin inscriptions and was found useful for dating and locating inscriptions by over 90% of epigraphists who tested the tool. The project is a collaboration between Google DeepMind, Oxford, Warwick, Nottingham, Athens, and Cambridge universities.
Renowned historian Mary Beard at Cambridge University considers Aeneas a transformative tool that democratizes access to historical reconstruction. Yannis Assael, a computer scientist at DeepMind and co-author of the study, views Aeneas as a new way of thinking about the past.
In a matter of seconds, Aeneas can identify common names, regional writing patterns, or shared rhetorical structures between different authors and eras. It offers new connections that were previously impossible to detect due to their volume or complexity.
In conclusion, Aeneas is a powerful tool enhancing historians’ ability to decode texts like the Res Gestae Divi Augusti, clarifying their dates and contexts and thus deepening understanding of Roman political narratives and history.
Aeneas, leveraging artificial-intelligence technology, identifies common names, regional writing patterns, or shared rhetorical structures between different authors and eras, offering new insights that were previously difficult to detect due to their volume or complexity.
By being trained on over 176,000 inscriptions from multiple epigraphic databases, Aeneas can attribute inscriptions to specific Roman provinces with 72% accuracy using both text and visual data, significantly improving the precision and speed of historical inquiry in the field of ancient Roman history.
