Google DeepMind launches AI skill for historians working with ancient inscriptions

Predicting the Past connects Gemini with Ithaca and Aeneas so historians can restore, map and compare ancient Greek and Latin inscriptions without coding. Google DeepMind has introduced the Predicting the Past Skill for Google Antigravity, built with Thea Sommerschield from Durham University. Google DeepMind has introduced the Predicting the Past Skill for Google Antigravity , an AI research tool built with Thea Sommerschield from Durham University to help historians and epigraphers analyze damaged ancient inscriptions through natural language. The skill connects Gemini with Google DeepMind’s specialist models Ithaca and Aeneas, which were developed to restore, date, place and contextualize ancient Greek and Latin inscriptions. The new version shifts that work from separate computational tasks into a conversational workflow for historical research. Google DeepMind said in a LinkedIn post that historians and epigraphers can now “query, cross-analyze, and map massive collections of ancient data as naturally as speaking with a colleague—with no coding required.” The tool is designed for researchers working with fragmentary texts, uncertain dates and unclear places of origin. Google DeepMind says the Predicting the Past Skill is available to try through Google Antigravity. The launch follows nearly a decade of collaboration between Google DeepMind and epigraphers, including Ithaca in 2022 and Aeneas in 2025. Sommerschield, a historian and epigrapher at Durham University, has co-led Google DeepMind’s work in this area. AI moves into the historian’s workflow Ancient inscriptions are central sources for historians, but many survive in damaged or partial form. They can include imperial decrees, votive dedications, everyday transactions and personal appeals, often with missing text, uncertain dating and disputed origin. Google DeepMind says earlier work with the epigraphy community identified three barriers for AI-assisted historical analysis: researchers need flexible visualizations for individual inscriptions, more advanced multi-text analysis without specialist coding, and large language models grounded in evidence and domain expertise. The Predicting the Past Skill is built to address those barriers by linking Gemini’s interactive reasoning with the outputs of Aeneas and Ithaca. Rather than asking a general-purpose model to work from scratch, the skill draws on specialist inscription models and presents results in a form historians can inspect. Google DeepMind says the tool can support restoration, attribution, contextualization, mapping and comparison across large collections of ancient data. From a stolen ring to a Roman cult Google DeepMind tested the skill across three case studies from the Greco-Roman world, starting with a Latin curse tablet from Roman Britain. The first case focuses on Tab.Sulis 97, a curse tablet from Aquae Sulis, the Roman settlement at Bath. The tablet was recovered from a votive deposit at the hot-spring sanctuary of Minerva and was written by a woman named Basilia, who cursed whoever had stolen her silver ring. Bath has produced hundreds of similar curse tablets, making it one of the major sources for this type of inscription. Google DeepMind says Aeneas placed the tablet within the chronological and geographical ranges proposed by historians, while producing an explanation of how that conclusion was reached. The point is not just that the model gave a date or place. Google DeepMind says the explanation began to resemble epigraphic commentary, using textual features to support the historical attribution. The second case moves from one inscription to a wider corpus. Google DeepMind used a votive altar from Mainz, or Mogontiacum, dedicated in 211 CE by the provincial official Lucius Maiorius Cogitatus to the Aufaniae, a group of Germanic mother-goddesses. Similar dedications are found across the Rhine and Danube provinces, often left by Roman soldiers and administrators. Google DeepMind says the skill analyzed regional patterns across related inscriptions and traced how religious practices spread through the movement of people across the Roman Empire. Dodona case maps people, not just texts The third case study uses lead oracular tablets from Dodona in northwest Greece, a sanctuary where visitors asked divine guidance on topics including business, travel, family affairs and religious obligations. Thousands of tablets survive from Dodona, many in highly fragmentary condition. Google DeepMind says the Predicting the Past Skill used the collection to move beyond individual inscription attribution and reconstruct a wider community of people who came to the sanctuary. That case shows the tool being used less as a restoration engine and more as a way to explore connections across a large historical dataset. Google DeepMind says researchers can examine Dodona not just as a collection of texts, but as a network of connected individuals moving through the ancient Mediterranean. The case studies also show the limits Google DeepMind is trying to address. A single inscription may need restoration and dating, while a corpus may require comparison, mapping and pattern detection across many damaged objects. Google DeepMind says the Dodona work used Dodona Online, a project providing critical reeditions of the oracular tablets as I.Dodone Online. Ithaca draws on a version of the Searchable Greek Inscriptions database from the Packard Humanities Institute, while Aeneas was trained on data from the Epigraphic Database Roma, the Epigraphic Database Heidelberg and the ETL repository for the Epigraphic Database Clauss Slaby. The Predicting the Past Skill is now available through Google Antigravity. Google DeepMind says it can help epigraphers analyze patterns and produce visualizations “in a matter of minutes,” with the work grounded in specialist inscription models and datasets rather than unsupported AI output. Subscribe to the ETIH newsletter Sign up with your email address to receive news and updates. First Name Last Name Email Address Sign Up We respect your privacy and will not pass your email address on to third parties. However, we will occasionally send you promotional messages on behalf of our advertisers. Thank you!