Google DeepMind’s Aeneas AI Revolutionizes Analysis of Ancient Latin Inscriptions

Aeneas: A New AI Tool for Deciphering Ancient Latin

Google DeepMind has introduced Aeneas, an innovative artificial intelligence software designed to assist historians in interpreting ancient Latin inscriptions. This tool can analyze weathered stone engravings to determine their original date and location, offering valuable context to scholars.

How Aeneas Works

Aeneas processes partial transcriptions of inscriptions together with scanned images. From these inputs, it suggests possible dates and places of origin, and proposes potential restorations for missing text segments. For example, when presented with a damaged inscription beginning with "... us populusque Romanus," Aeneas might suggest the preceding word "Senat" to complete the phrase "Senatus populusque Romanus," meaning "The Senate and the people of Rome."

Comparing Aeneas to Ithaca

While Aeneas shares some similarities with Google's earlier archaeological AI tool, Ithaca, which focused on ancient Greek texts, Aeneas extends its capabilities by cross-referencing inscriptions against a vast database of nearly 150,000 Latin engravings. These records span regions from modern-day Britain to Iraq, helping identify parallels and contextual clues from other texts.

Training Data and Model Limitations

The training set for Aeneas includes thousands of inscription images and a large corpus of Latin texts. Despite this, the dataset is much smaller compared to the massive corpora used to train general-purpose language models like Google’s Gemini. Due to the scarcity of high-quality inscription scans, specialized AI like Aeneas is essential for this niche task.

Supporting Historical Research

The development team emphasizes that Aeneas is not meant to automate epigraphy but to augment historians' workflows by providing multiple hypotheses for further investigation. A validation study involving 23 historians showed that Aeneas inspired new research ideas in 90% of cases and improved accuracy in dating and locating inscriptions.

Real-World Application and Future Prospects

Aeneas was tested on the Monumentum Ancyranum, a well-known inscription in Ankara, Turkey, where its results closely matched expert analysis. However, questions remain about its performance on less-studied inscriptions and its long-term usefulness in historical research.

Accessibility and Educational Use

Google DeepMind has open-sourced Aeneas and made its interface freely available to educators, students, museum professionals, and academics. The team is collaborating with Belgian schools to incorporate Aeneas into secondary history education. One of their aspirations is for Aeneas to be used onsite at museums or archaeological sites to assist with newly discovered inscriptions.

Thea Sommerschield, an epigrapher involved in the project, expressed excitement about the tool’s potential, envisioning it as a constant companion for researchers exploring ancient texts.