Google’s new AI tool helps historians decipher incomplete ancient Latin texts
Historians input a scanned image and partial transcription, and Aeneas generates likely completions, historical parallels and possible origins for the inscription
What if an algorithm could breathe new life into the worn-out words of ancient Rome? A new artificial intelligence tool developed by Google DeepMind may just be doing that.
Called Aeneas, this tool is designed to assist historians in deciphering Latin inscriptions — many of which have been eroded by time, weather or damage. Drawing from a database of nearly 150,000 Latin texts and thousands of scanned images, Aeneas offers researchers suggestions on what might be missing from a damaged stone, and when and where the inscription was likely made.
The process is collaborative though, not automatic. Historians feed the system a scanned image and a partial transcription. Aeneas then offers educated guesses for the missing text, suggests historical parallels, and proposes a likely origin. These clues do not provide answers outright, but rather guide experts toward possible interpretations.
Its development builds on DeepMind's earlier tool, Ithaca, which focused on Greek inscriptions. But Aeneas moves a step further by integrating a wider dataset and offering historical comparisons across geography and time.
During testing, Aeneas improved the accuracy of expert analysis in nine out of 10 cases. It even impressed historians with its detailed reading of the Monumentum Ancyranum — one of the most important inscriptions from the Roman era.
Still, the system has limitations. It cannot interpret text meaning on its own and works best when paired with expert input. Questions remain about how well it will perform with less studied artefacts.
Now made open-source, Aeneas is available to students, teachers, museum professionals, and academics. DeepMind is already working with Belgian schools to bring it into classrooms. The tool is offering researchers a powerful companion that can help connect fragments of the past with the present.
