Historians can Decipher Damaged Greek Texts Thanks to a New Artificial Intelligence (AI)

Summary

Historians can now use an AI program called Ithaca to decipher and date ancient Greek texts with significantly improved accuracy.

Abstract

The "Ithaca" AI, developed by DeepMind, has been designed to reconstruct and analyze ancient Greek inscriptions, providing historians with a powerful tool to fill in missing content, identify geographic origins, and date texts with precision. The program has demonstrated the ability to accurately restore up to 72% of damaged manuscripts, outperforming human historians who managed to restore only 25% without AI assistance. Ithaca's potential extends beyond Greek texts, as it has the flexibility to be adapted for other ancient writing systems. This tool is seen as a significant advancement, akin to the invention of microscopes and telescopes, which could revolutionize the study of ancient civilizations.

Opinions

Thea Sommerschield, a study co-author, emphasizes the importance of Ithaca for interpreting fragmentary or unintelligible manuscripts that have been moved from their original locations.
Yannis Assael, another co-author, likens Ithaca's impact to that of microscopes and telescopes, suggesting it will enhance our ability to investigate pivotal moments in human history.
Eleanor Dickey, a classicist at the University of Reading, cautions that while Ithaca is a significant advancement, it is still prone to errors about one-third of the time, and it remains to be seen if this technology will result in more accurate versions of ancient texts for public access.

Historians can Decipher Damaged Greek Texts Thanks to a New Artificial Intelligence (AI)

The program “Ithaca,” developed by the program’s creators, can recreate inscriptions that have been lost or destroyed

Matteo Vistocco/Unsplash

According to the Guardian’s Nicola Davis, a new artificial intelligence software (AI) might assist to fill in the gaps in ancient Greek manuscripts and pinpoint when and where they are from. In a report published in Nature by the program’s creators, it was revealed that their new artificial intelligence system could accurately recreate up to 62% to 72% of the content of lost or damaged manuscripts.

Google subsidiary DeepMind built the AI, dubbed “Ithaca” after the Greek island where King Odysseus was born. According to Carissa Wong of New Scientist, the researchers tested the algorithm using data from over 60,000 well-studied ancient Greek manuscripts dating from 700 B.C. to 500 C.E. The scientists detected how accurately Ithaca filled in the missing bits by concealing some of the sentences. Using just its resources, Ithaca could correctly restore 62% of the manuscripts it examined, identify the text’s geographic origins in 71% of the cases, and date the texts to within 30 years of their original publication date. Historians could restore only 25% of texts without the help of the software. Historians augmented AI-generated data to boost historical accuracy to 72 percent.

Historians will find Ithaca essential in recreating ancient manuscripts, according to academics. Study co-author Thea Sommerschield of Ca’ Foscari University in Venice and Harvard University told the Guardian that most of these surviving manuscripts are “fragmentary or unintelligible” and may have been transferred from their original site. In addition, radiocarbon dating cannot be used to establish the age of stone inscriptions, rendering this method inconclusive. Sommerschield claims that the Ithaca program has much more promise and has already been used to date inscriptions found in the Acropolis of Athens. Even though Ithaca was developed to read Greek inscriptions, its adaptability means that it may be used to read any ancient writing system, from Mayan to cuneiform.

According to the study co-author, Yannis Assael, much in the same way that microscopes and telescopes have widened the breadth of what scientists can accomplish today, Ithaca intends to complement and expand the ability to investigate one of the most critical times in human history.

“When people rely on it they will need to keep in mind that it is wrong about one-third of the time,” Eleanor Dickey, a classicist at the University of Reading, tells the Verge’s James Vincent, who interviewed her. Although Dickey believes it’s too early to tell if this technology will lead to better versions being made accessible to the public, she adds, “it will be interesting to find out.”