Google DeepMind has introduced a new artificial intelligence (AI) model, Aeneas, designed to help historians analyze ancient Latin inscriptions. The software can estimate when and where engravings were made and suggest missing text, offering researchers valuable historical information.
DeepMind’s Aeneas AI works by examining partial transcriptions and scanned images of inscriptions, then proposing possible dates, origins, and text completions. For example, if an inscription reads “…us populusque Romanus,” Aeneas might suggest adding “Senat” at the beginning to form the famous phrase “Senatus populusque Romanus” (“The Senate and the people of Rome”).
Unlike DeepMind’s earlier tool, Ithaca, which focused on Greek texts, Aeneas cross-references inscriptions with a database of nearly 150,000 Latin engravings from regions spanning modern Britain to Iraq. This allows historians to find connections between similar texts. However, the training data is limited compared to large language models like Gemini, as high-quality scans of ancient inscriptions are scarce.
In a study published in Nature, researchers tested Aeneas with 23 historians, finding that it improved accuracy in dating and locating inscriptions while also sparking new research ideas 90% of the time. The AI was also tested on the well-known Monumentum Ancyranum in Turkey, producing results that aligned with existing historical analysis.
Still, experts caution that Aeneas doesn’t interpret meaning, and its effectiveness on lesser-known inscriptions remains uncertain. Google DeepMind has made the tool open-source, aiming to integrate it into education and museum work.
To learn more about Aeneas and Ithaca, you may visit the official website here: https://predictingthepast.com.
Source: Google DeepMind