AI deciphers 200-year-old Napoleon letter

A letter written in 1809 to one of Napoleon’s generals, which had remained undeciphered for over 200 years, was decoded by artificial intelligence in just six hours. The letter described the positions of French and allied troops and was later published in an online university archive. Without AI, the task might have taken weeks, but a model developed by OpenAI, the company behind ChatGPT, completed it quickly.

AI reveals ancient papyrus texts

AI is also helping to read carbonized papyri from the ancient Roman city of Herculanum, preserved by the eruption of Mount Vesuvius in 79 AD. These scrolls are too fragile to be unrolled without risk of destruction, so researchers scan them with X-rays and use AI to reconstruct the writing. In June, a team announced the first complete virtual unrolling of a papyrus from the National Library of Naples, revealing a Stoic treatise on ethics.

French project targets ancient encrypted texts

In France, the National Institute for Research in Digital Science and Technology (Inria) is working on a project called "Back in time." This initiative brings together historians, cryptographers, and AI specialists to create a tool that can transcribe and decipher ancient encrypted documents. These include diplomatic and military messages written in symbols like triangles and circles, which have remained unread for centuries and are stored in European archives.

AI expands historical research scale

The "Back in time" project aims to improve AI's ability to process historical texts and provide deeper insights into the past. AI is also transforming the scale of historical research, as seen in the Midrash project. This initiative, led by the École pratique des hautes études-PSL, uses AI to transcribe millions of pages of manuscripts, creating large, searchable databases on Jewish culture. The collection has been described as a "huge Facebook of the past" by one of the project’s directors.

AI uncovers hidden historical connections

Transcribed texts are being fed into large databases, allowing historians to search through previously inaccessible collections. This method has brought together fragments from the Cairo Genizah, a global collection of Jewish manuscripts. It is also being used in the e-NDP project, where AI transcribed 14,600 pages of medieval Latin records from Notre-Dame de Paris. The work revealed details about the cathedral’s daily operations, staff, and care for abandoned children. Julie Claustre, who leads the project, estimates that doing this manually would have taken "dozens of years." Automation, she says, "makes possible investigations that were not possible before, transforming the conditions of research."