Google DeepMind Introduces Aeneas: A New AI That Helps Historians Understand Ancient Latin Inscriptions

Google DeepMind has unveiled the Aeneas model, a novel AI-driven approach to historical research. This model aids in deciphering ancient Roman inscriptions. Developed with top universities, Aeneas speeds up historians' ability to find textual "parallels" in Latin inscriptions. It excels at handling incomplete or damaged texts, giving historians new tools to understand ancient writings.

Aeneas is designed for inscriptions often missing full context. By combining text and visual data, it provides insights into Roman life and society. Google states that Aeneas achieves 73% accuracy in filling gaps of up to ten characters and can date texts within 13 years of historians' estimates. These features make it essential for exploring Roman history through various inscriptions.

AI Integration in Historical Research

The model's testing on the Res Gestae Divi Augusti, a renowned Roman inscription by Emperor Augustus, showcased its capabilities. Aeneas offered a range of possible dates, quantitatively capturing different hypotheses. This highlights the model's ability to turn historical questions into a probabilistic framework, offering fresh perspectives on age-old debates.

Aeneas excels in geographical attribution using a multimodal generative neural network that analyses both text and images. With the Latin Epigraphic Dataset (LED) containing over 1,76,000 inscriptions, Aeneas groups texts by date more precisely than other models. This AI integration exemplifies the collaboration between machine learning and expert knowledge.

Collaboration and Educational Initiatives

Developed with the University of Nottingham and partners from Warwick, Oxford, and Athens University of Economics and Business (AUEB), Aeneas represents a collective effort to use AI for historical research. This collaboration aims to help historians identify parallels on a large scale, providing new avenues for inquiry and innovation.

Aeneas is available through an interactive platform for researchers and educators, aligning with efforts to boost AI literacy. It supports restoring inscriptions with unknown gap lengths, crucial for handling severely damaged texts. Its ability to find "parallels" enriches understanding of Roman society and geography.

Future Prospects of Aeneas

The developers are committed to enhancing Aeneas' versatility through a new teaching syllabus that integrates technical skills with historical analysis. This initiative supports AI literacy and aligns with European educational frameworks by providing resources bridging technical and historical fields.

As Aeneas becomes integral to historical research, its capabilities in processing multimodal inputs and restoring unknown-length texts will be invaluable. By transforming historical analysis into a more quantitative process, Aeneas promises new insights into ancient societies, ensuring past civilisations' legacies are explored innovatively.