Portuguese 
English 
Spanish 
4 min of reading 
1 comment 
Advances in AI Enable the Interpretation of Ancient Texts That Were Once Indecipherable, Helping Archaeologists and Historians to Better Understand the Past of Humanity
For centuries, reading cuneiform texts involved meticulous manual transcription work. Experts spent years analyzing individual tablets, dealing with variations in ancient writing styles that complicated translation.
Now, a new artificial intelligence (AI) system is evolving this process, making deciphering faster and more efficient.
Researchers from Cornell University and Tel Aviv University developed ProtoSnap, an AI model capable of recognizing and reconstructing cuneiform characters with high precision.
-
Artificial intelligence for robots with human-inspired hands advances and expands machine learning capabilities in the new generation of robotics.
-
The world’s richest man says that saving money for retirement could become a thing of the past with the advancement of artificial intelligence and humanoid robots.
-
China’s Court decides that companies cannot replace employees with artificial intelligence and surprises the world with a measure that could curb mass layoffs and redefine the global market.
-
Powerful artificial intelligence from the University of Warwick finds over 100 hidden planets in NASA data and reveals rare worlds
The system takes into account the regional and temporal differences in writing, offering a new approach to interpreting texts from ancient Mesopotamia.
The advancement is already increasing the number of translated documents and revealing new information about the economic and social life of the time. However, according to experts, this technology can go even further.
Challenges of Cuneiform Writing
Unlike modern alphabets, which consist of a limited number of letters, cuneiform presents over a thousand signs. Additionally, its form changed over the centuries, making reading a significant challenge even for experts.
“Even with the same character, the appearance changes over time, so it is a very challenging problem to automatically decipher what the character actually means,” explains Hadar Averbuch-Elor, a computer scientist at Cornell Tech.
To solve this difficulty, the researchers employed a diffusion model, a type of generative AI. The system compares the image of a cuneiform character with a known prototype and adjusts it in the correct context, like a puzzle piece.
With this approach, ProtoSnap can consider, copy, and reproduce cuneiform writings more quickly and precisely than ever.
Expanding Access to the Past
Currently, there are about half a million cuneiform tablets preserved in museums around the world. However, only a small portion has been fully translated.
With the automation of the transcription process, ProtoSnap can accelerate the availability of these texts, providing new insights into the law, economy, and daily life of Mesopotamia.
“At the core of our research is the objective of increasing the available ancient sources for us tenfold,” says Yoram Cohen, an archaeologist at Tel Aviv University. “This will allow us, for the first time, to manipulate big data, leading to new measurable insights about ancient societies — their religion, economy, social and legal life.”
Artificial Intelligence and Deciphering Ancient Writing
ProtoSnap is an example of how artificial intelligence has been revolutionizing the reading of historical documents. In 2023, researchers studied AI to decipher a Greek papyrus carbonized nearly 2,000 years ago in Herculaneum, a Roman city buried by Vesuvius’s eruptions.
These scrolls, discovered in the 18th century, were considered impossible to open without destroying them. However, a group of computer scientists and papyrologists created an algorithm to identify invisible ink traces in the layers of papyrus. The result was the recovery of entire passages of texts that had been hidden for centuries.
“It was incredible, ” recalls Federica Nicolardi, a papyrologist at the University of Naples. “I thought, ‘So this is really happening.’”
AI in Confirmation of History
The use of artificial intelligence for ancient texts is becoming increasingly common. Machine learning models have already helped restore damaged Greek inscriptions, translate Akkadian tablets, and even determine the origin and dates of historical documents.
The Itaca system, developed by the University of Oxford, was recently used to review decrees from Classical Athens. Based on historical evidence, scholars debated whether these texts were written around 420 BC or 446/445 BC. When tested, Itaca confirmed the date of 421 BC, reinforcing the revision of the historical chronology.
“Although it may seem like a small difference, this change in date has significant implications for our understanding of Classical Athens’s political history,” explains Jonathan Prag, a professor of Ancient History at the University of Oxford.
Meanwhile, in Korea, researchers are using AI to process documents written in Hanja, an ancient writing system used for centuries in the region.
These archives contain records about the reigns of 27 Korean kings, spanning 500 years of history. With the help of AI, scholars are identifying governance, economic, and diplomatic patterns of the time.
Challenges and Limitations
Despite advancements, the use of artificial intelligence in translating historical texts still presents challenges. In some cases, neural networks may generate misleading translations or fill in gaps with plausible but incorrect information.
This phenomenon, known as AI hallucination, can lead to distorted interpretations of the documents.
For this reason, experts emphasize that the technology should serve as a complement to human work, not a substitute.
With information from ZME Science.
Agora temos uma “janela” ampla para conhecer a fundo parte da antiguidade. Ótima notícia.