Deepmind Artificial Intelligence restores the thousands of years of Roman inscriptions

In the latest change in the use of artificial intelligence in the humanities, Google Deepmind, in collaboration with a group of historians, has developed an artificial intelligence -based tool called Aeneas capable of rebuilding damaged Latin inscriptions. This tool can fill the vacancies of deleted or worn -out vocabulary in the inscriptions, as estimating the place of writing and the time range of their creation.

According to Dr. Tia Samrschild, a history researcher from the University of Nottingham and one of the developers of the project, this tool is “changing game” in inscriptions studies.

Aeneas, using a set of about 6,000 Latin inscriptions, which contains about 5 million characters in total, is trained and can only identify similar inscriptions and analyze historical communication between them by receiving the text or image of an inscription using deep learning algorithms.

The system is capable of estimating the time of writing a scroll for up to 5 years and attributing it to one of the four Roman provinces. Also in experiments on texts such as Res Gestae Divi Augusti or altar inscriptions in modern -day Germany, it has been able to reveal new information from the origin or interactions between the inscriptions.

The preliminary results of the use of this tool, published in the journal Nature, show that in 5 % of the information provided by Aeneas, the researchers have helped researchers more accurately analyzing the inscriptions.

“Aeneas has the potential to transform our understanding of ancient texts,” said Professor Mary Bard, a well -known Roman history expert from the University of Cambridge.

The project can also provide the path for the widespread use of researchers in countries without huge historical or financial resources; What is especially inspired by the academic community in developing countries.