Researchers at Durham University have collaborated with Google DeepMind on a new AI-powered research tool designed to analyse ancient Greek and Latin inscriptions.
AI and historical evidence
The new system, called Predicting the Past Skill for Google Antigravity, enables historians to study ancient inscriptions, analyse patterns across thousands of texts, and generate visualisations. This can be achieved by having interactive conversations with the AI Skill without the need for any coding or separate technical tools.
The Skill is grounded in the team's earlier specialised models, Ithaca and Aeneas, which were trained on large historical datasets that are now directly linked in with Google Gemini.
This makes advanced historical analysis, based on solid and specialist evidence, more accessible and quicker.
Uncovering insights - interactively
The tool has been tested on a series of real-world case studies from the Greek and Roman worlds. These include a Latin tablet from Bath in the UK, an altar from Mainz in Germany, and damaged tablets from Dodona in Greece.
Through these examples, it shows how the tool can help to restore missing text, estimate where and when inscriptions were written, and identify regional patterns across large collections of historical evidence.
Google DeepMind collaboration
The project is the latest collaboration between Durham historian and epigrapher Dr Thea Sommerschield and Google DeepMind researchers Dr Yannis Assael and Dr Zoi Tsangalidou.
By bringing together historical datasets, specialised AI models and Gemini's reasoning capabilities, the team aims to help historians examine ancient inscriptions at scale while maintaining a strong focus on evidence-based research.
