How AI Translation Is Helping Scientists Understand Ancient Languages

Artificial intelligence (AI) may be the key to decoding ancient languages. 

Researchers at a German university are using AI to understand lost texts from antiquity and will release a vast trove of translations. It's part of a growing movement to use AI translation to uncover archaeological secrets. 

"AI combines the full history of human intelligence with cloud scalability," Rijul Gupta, the CEO of DeepMedia, an AI platform that translates video into different languages. "Think back to the language specialists who used the Rosetta Stone to decipher Egyptian scripts by matching a translation from ancient Greek. It created a never-before-seen breakthrough. AI enables breakthroughs at a far greater pace."

Understanding the Past Through Tech

A team of scientists at LMU Munich in Germany has used AI to translate 300,000 lines of text from ancient languages that were found on cuneiform tablets from ancient Mesopotamia. The researchers developed an algorithm to piece together fragments to make them understandable. 

"It's a tool that didn't exist before, a huge database of fragments," one of the team members, Enrique Jiménez, a professor of Ancient Near Eastern Literatures at LMU Munich, said in the news release. "We believe it can play a vital role in reconstructing Babylonian literature, allowing us to make much faster progress." 

At a news conference this month, the researchers plan to publish the AI database and a digital edition containing all known transcriptions of cuneiform fragments to date. With more than 300,000 lines of text, most of them previously unpublished, the researchers say the launch will be the largest publication of texts in the history of cuneiform studies. Jiménez will publish complete electronic editions of important texts such as the Babylonian creation myth and the Gilgamesh epic. 

The team's AI algorithm has identified hundreds of manuscripts and many textual connections. Last year, for example, the software recognized a fragment belonging to the most recent tablet of the Gilgamesh epic, which dates from 130 BC—making it thousands of years younger than the earliest known version of the Epic. 

The German researchers aren't the only ones using AI in archaeology. DeepMedia is currently using its text and voice platform to integrate ancient languages. For instance, last year, the Ithaca AI system was able to decipher the gaps in ancient Greek inscriptions with 62 percent accuracy.

"To be sure, AI can enhance, not replace, human intelligence, human linguists, human historians, and so on," Gupta said. "When it comes to ancient languages, the clues often come from fragments of physical artifacts and cultural immersion with people who have descended from those who spoke ancient languages. Still, AI can process massive amounts of data very quickly—potentially speeding up certain processes by months or years."

A Swifly Moving Field

The future looks bright for AI translation of ancient texts. As more and more ancient texts are digitized, AI algorithms will be able to process this data more efficiently and accurately than ever before, Sven Patzer, an AI engineering consultant, told Lifewire via email. 

"This will lead to new insights into the underlying structure of ancient languages and a deeper understanding of the meanings behind the texts," Patzer added. 

Patzer added that developing deep learning algorithms and neural networks will allow AI to make increasingly accurate predictions about the meaning of ancient texts. "This will be particularly useful in deciphering languages that have not yet been deciphered," he said. "By using AI to identify patterns in the data, researchers will be able to make informed guesses about the meaning of these ancient scripts."

AI's ability to process images of texts will play an essential role in deciphering ancient languages. "For example, AI algorithms can be used to identify and transcribe texts written on fragile or damaged manuscripts," Patzer said. "This will help to preserve these important historical documents and allow researchers to analyze their contents."

Gupta said that AI archaeology is still in its infancy. In the future, "we'll use large language models like ChatGPT to advance the space," he said. "Language models will pick up patterns in human language to help decipher ancient languages, giving us never-before-seen insights into our ancient past."