A study published in Nature introduces Aeneas, an AI tool developed by Google DeepMind in collaboration with researchers specialized in epigraphy. According to The Economic Times and The Economist, Aeneas is capable of predicting missing words in Roman inscriptions and assisting scholars in interpreting and restoring ancient texts.
One of the key features of Aeneas is its ability to process images of Latin text as well as transcribed inscriptions. This allows the tool to handle worn and damaged artifacts, providing suggestions for missing passages and assisting in the interpretation of texts that have deteriorated over centuries.
“Aeneas helps historians interpret, attribute, and restore fragmented Latin texts,” said Dr. Thea Sommerschield from the University of Nottingham during a presentation about the tool, according to a report by Adevarul. Dr. Sommerschield co-developed the AI system along with Dr. Yannis Assael, an AI specialist at Google DeepMind.
The AI tool was trained on a vast database of approximately 200,000 Latin inscriptions, totaling 16 million characters. Using this extensive dataset, Aeneas can identify subtle linguistic patterns, grammatical structures, common formulas, and abbreviations across all types of inscriptions, which aids in accurate textual reconstructions.
In tests, Aeneas analyzed inscriptions from historical artifacts, such as the Res Gestae Divi Augusti, a monumental inscription detailing the achievements of Emperor Augustus. The tool suggested two periods of drafting for this text—either the decade of 10 BCE or between 10 and 20 CE—both corresponding to the timeframes debated by historians. “It was one of those moments that left us speechless,” Dr. Sommerschield remarked, referring to Aeneas's findings, according to LIFO.
Historians who have tested Aeneas reported that it has transformed their work, particularly in identifying similar inscriptions, which is crucial for understanding the historical context of the texts. According to The Economist, Jonathan Prag, co-author and professor of ancient history at the University of Oxford, stated that Aeneas could be run on the existing corpus of inscriptions to see if interpretations can be improved.
According to Adevarul, in an extensive test involving 23 historians, Aeneas was used to analyze Latin inscriptions, and the tool provided useful context in 90% of the cases, improving their confidence in key tasks by 44%. “The similar texts Aeneas retrieved completely changed my perception of an inscription,” noted one volunteer.
Aeneas can determine the place and period in which a text was created with a reduced margin of error. It can attribute an inscription to one of the 62 Roman provinces and estimate its dating with an accuracy of 13 years. This capability is particularly important, given the complexities involved in dating and locating ancient inscriptions.
Moreover, Aeneas can suggest completions for missing passages, making accurate suggestions for filling in the inevitable gaps in worn or fragmentary texts. In tests on known texts with artificially obscured parts, it correctly identified segments among its top 20 predictions 46.5% of the time. According to The Economist, when the length of the missing segment is unknown, the accuracy in predictions drops to 32.7%.
The tool's ability to recognize and link inscriptions through deeper historical connections is also noteworthy. It identifies complex linguistic and historical correlations, effectively turning each text into a historical fingerprint. This allows historians to situate inscriptions within their broader historical context, facilitating new insights and research directions.
“Aeneas's parallels completely changed my perception of the inscription,” wrote an unnamed historian who tested the tool. “It noticed details that made all the difference for restoring and chronologically attributing the text,” as reported by Engadget.
Despite its impressive capabilities, Aeneas is intended to be a collaborative tool rather than a replacement for human expertise. “What the historian can't do is assess these parallels in a matter of seconds across tens of thousands of inscriptions, and that is where AI can come in as an assistant,” explained Dr. Assael, according to MIT Technology Review.
The integration of AI tools like Aeneas into historical research signifies a new era in the humanities. By automating time-consuming tasks and providing quick access to vast amounts of data, such tools enable historians to focus on deeper analysis and interpretation. “This is the great challenge we set out to tackle,” Dr. Sommerschield said, as reported by LIFO.
“Aeneas is already available online, providing a compelling case for incorporation into researchers' workflows,” according to Dr. Charlotte Tupman. The tool not only assists in restoration tasks by identifying missing text but also specifies the place and chronology to which inscriptions belong, offering useful starting points for further research.
As more inscriptions are digitized and added to databases, the effectiveness of Aeneas is expected to improve. “We have seen that the more inscriptions we have, the better the model works,” noted Dr. Assael, according to Agencia SINC. The team hopes that tools like Aeneas can help "connect the past" by revealing patterns and connections that might otherwise remain undiscovered.
The use of AI in historical research challenges scholars to rethink traditional methods and embrace new technologies. While some may view such tools as controversial, there is growing recognition of the potential benefits. “Experimenting with tools such as Aeneas and reflecting on these questions can benefit research and the future understanding of source materials from past societies,” stated an article in Nature.
Produced with the assistance of a news-analysis system.