DeepMind, in collaboration with historians, has developed an artificial intelligence (AI) model named Ithaca, designed to assist in restoring and dating damaged ancient Greek inscriptions. This AI system builds upon previous work by DeepMind, notably the Pythia model introduced in 2019, which focused on text restoration. Ithaca extends these capabilities by also predicting the geographical origin and dating of these inscriptions.
In tests, Ithaca demonstrated a 62% accuracy rate in restoring missing text from damaged inscriptions. When combined with human expertise, the accuracy improved to 72%, highlighting the potential of AI-human collaboration in this field. Additionally, Ithaca achieved a 71% accuracy rate in determining the geographical origin of the inscriptions and could date them within a 30-year range. These capabilities are particularly valuable, as traditional methods like radiocarbon dating are not applicable to stone or metal artifacts. The AI model has already been utilized to resolve debates among historians regarding the dating of certain ancient Athenian decrees, providing insights that align with recent scholarly conclusions. The development of Ithaca signifies a significant advancement in the application of AI to historical research, offering tools that can enhance the understanding of ancient texts and their contexts. The project also underscores the growing intersection of technology and the humanities, opening new avenues for interdisciplinary research and discovery.