AI Resurrects 3000-Year-Old Babylonian Hymn with Cutting-Edge Technology

Researchers from LMU Munich and the University of Baghdad have successfully reconstructed a poem titled «The Babylonian Hymn» using artificial intelligence. This ancient work had been lost for over 2,000 years, according to Decrypt.

The poem celebrates Babylon and the god Marduk, having been composed around 3,000 years ago, with its last analysis occurring in 100 BCE.

The team involved in this project reported that they pieced together the work from a total of 30 clay fragments uncovered over many years, employing AI to unify the text.

«We utilized a specialized AI-based program to analyze and match fragments of the text based on combinations of cuneiform characters,» said Enrique Jimenez, a professor of ancient Near Eastern languages at LMU Munich.

The expert, alongside his colleagues, applies methods rooted in natural language processing to identify to which text the fragments belong.

Using the Electronic Babylonian Library platform, which houses 1,402 manuscripts, the researchers primarily utilize n-gram matching as their reconstruction technique, supplementing it with other approaches like vocabulary overlap and searching for the longest common text sequences.

Jimenez emphasized that the rediscovered poem was of considerable significance to its contemporaries and was included in Babylon’s educational curriculum.

In an article for the journal Iraq, he and co-author Anmar Fadhil proposed that the creator of the poem likely belonged to the priestly class of Babylon, as the poem describes priests as «free citizens.»

The hymn extols the city’s natural resources and beauty, expresses respect for foreigners, and supports the poor.

«It does not demean foreigners among them. The humble they protect, the weak they support. Under their care, the poor and destitute can flourish. They bring comfort and favor to orphans,» the composition states.

Jimenez noted that artificial intelligence is becoming indispensable for researchers, particularly for the restoration of damaged or fragmented texts.

«While languages like Akkadian and Sumerian are still underrepresented in language models, we are actively working on enhancing computational tools for the study of the ancient Near East,» the expert commented.

Previously, scientists from DeepMind and Ca’ Foscari University in Venice developed an AI algorithm called Ithaca to restore damaged ancient texts.

In October 2021, Google AI experts recreated destroyed paintings by Gustav Klimt using artificial intelligence.