en.Wedoany.com Reported - A new artificial intelligence model is helping researchers crack historical ciphers, including a mysterious handwritten book of symbols that has remained unreadable for over 400 years in the Vatican Library. Notes on the inside cover suggest it contains secret medicinal recipes—knowledge kept confidential at the time to avoid suspicion of witchcraft.
It is estimated that about 1% of materials in archives and libraries worldwide are fully or partially encrypted. These historical ciphers vary greatly in complexity: simple systems replace individual letters with symbols, while more difficult variants add meaningless decoy symbols to prevent decryption. In some cases, the original language is also unknown, making research a time-consuming process that still relies on trial and error.
"It's like detective work—each symbol, each pattern, and each partial decryption brings us closer to a person's secret and a lost historical world," said Beáta Megyei, a professor of computational linguistics at the University of Stockholm. Despite AI assistance, decoding the Vatican Library manuscript remains laborious. She and her team now hope to make the technology more efficient. "This opens up exciting possibilities for rare and non-standard writing systems," the researcher said.
Encrypted documents may conceal diplomatic secrets, rituals of secret societies, or medical knowledge—information still missing from historical narratives. Their decryption could fundamentally change our understanding of well-known figures or entire eras. Cécile Pierrot, a cryptologist at the French National Institute for Informatics Research, spent six months decrypting a letter written by Charles V 500 years ago. The content was surprising: the most powerful ruler in Europe at the time was evidently deeply concerned about assassination plots.
Pierrot said she typically needs a day to transcribe a two-page letter with unknown symbols. AI is beginning to accelerate this process. For example, Michelle Waldispühl, a professor at the University of Oslo, used the AI platform Transkribus to decrypt a coded letter written in 1637 by the nobleman Sigismund Heusner von Wandersleben. The tool is trained on multiple centuries of different languages, fonts, and handwriting styles. It identifies text blocks and then scans text character by character. Although manual correction is still required, it provides significant assistance in partially encrypted letters.
However, existing AI transcription platforms still encounter bottlenecks when manuscripts are encrypted using fictional symbols, astrological signs, or unconventional fonts. Therefore, under the transnational Descrypt project, Megyei, Waldispühl, and their colleagues are developing their own AI tools for such special cases. "We are developing more adaptable models that are trained and tested on a broad library of fonts, alphabets, and symbols," Megyei said. AI offers particularly valuable support in terms of scale, speed, and pattern recognition. "What excites me is not just the possibility of solving a specific historical puzzle, but the prospect of developing methods that can help researchers tackle a wide variety of different scenarios."
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
