Google DeepMind’s Aeneas AI Deciphers Ancient Latin Inscriptions

Aeneas Bridges the Gap to Ancient Inscriptions

Google DeepMind’s Aeneas is a specialized AI system designed to help historians unlock the secrets of ancient Latin inscriptions. Building on the success of its Greek-focused predecessor Ithaca, Aeneas analyzes weathered stone texts to recover missing fragments and provide context. By combining partial transcriptions, high-resolution scans, and a curated database, it offers a new suite of hypotheses for epigraphic research.

Aeneas accepts two inputs: a scanned image of an inscription and any available transcription, even if fragmentary. Internally, its deep neural network processes visual features and linguistic patterns to estimate when and where the engraving was made. The system then proposes likely text completions for damaged sections, modeling Latin syntax and common formulas.

• Leverages deep neural networks trained on almost 150,000 Latin inscriptions and thousands of high-quality scans.
• Estimates dates and geographic origins for fragmentary engravings by analyzing partial transcriptions and images.
• Suggests probable word fill-ins to complete damaged Latin text segments.
• Cross-references an extensive inscription database to propose historical parallels and related artifacts.

Real-World Validation Accelerates Research

DeepMind validated Aeneas in a study with 23 professional historians, comparing traditional workflows against those augmented by the AI. Results, published in Nature, showed that Aeneas sparked new research ideas on 90% of the test inscriptions and improved the accuracy of dating and localization tasks. Participants reported that the tool integrated smoothly into their analysis routines.

In testing on the Monumentum Ancyranum, a well-known inscription at Ankara’s Temple of Augustus, Aeneas produced date and origin estimates aligned with established scholarship. It also identified parallels to other Roman public decrees, demonstrating an attention to detail comparable to that of a seasoned epigrapher.

Open Source Meets Education

Released as open-source software, Aeneas is freely accessible to teachers, students, museum curators, and academic researchers. DeepMind is collaborating with Belgian secondary schools to embed the tool in history education, hoping students will engage directly with digital epigraphy during museum visits and fieldwork.

While Aeneas does not translate or interpret meaning autonomously, it equips historians with rapid hypotheses and contextual leads. Questions remain about its performance on obscure or less-studied inscriptions, but the model’s open framework invites ongoing refinement. As a companion in the epigrapher’s toolkit, Aeneas sets the stage for AI-assisted discoveries in the humanities.