AI Breathes Life into a 3000-Year-Old Babylonian Poem

Researchers from LMU Munich and the University of Baghdad have utilized artificial intelligence to reconstruct a poem known as the «Babylonian Hymn,» which was lost over 2000 years ago, as reported by Decrypt.

This ancient text pays tribute to Babylon and the god Marduk, having been written about 3000 years ago. The last known study of it occurred in 100 BC.

The project team stated that the work was compiled from 30 clay fragments unearthed over many years, with AI being employed to piece the text together.

«We employed a specialized AI-based program to analyze and align the text fragments based on combinations of cuneiform signs,» explained Enrique Jimenez, a professor of Ancient Near Eastern Languages at LMU Munich.

Jimenez and his colleagues are using natural language processing techniques to identify the texts that the fragments relate to.

By utilizing the Electronic Babylonian Library, which houses 1402 manuscripts, the researchers apply n-gram matching as their primary reconstruction method. They also incorporate other approaches, such as vocabulary overlap and searching for the longest common text sequences.

Jimenez highlighted that the poem discovered held significant value for its contemporaries and was likely part of Babylon’s educational curriculum.

In a paper published in the journal Iraq, he and co-author Anmar Fadhil suggested that the poem’s creator was likely a member of Babylon’s priestly class, as the poem describes priests as «free citizens» in one of its sections.

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

«It does not belittle foreigners among them. The humble are protected, and the weak are assisted. Under their care, the poor and downtrodden can thrive. They offer comfort and kindness to orphans,» states the composition.

Jimenez noted that artificial intelligence has become indispensable for researchers, particularly for restoring damaged or fragmented texts.

«Although languages like Akkadian and Sumerian are still underrepresented in LLMs, we are actively working to enhance computational tools for studying the ancient Near East,» the expert added.

Earlier, scientists from DeepMind and Ca’ Foscari University of Venice developed an AI algorithm named Ithaca for recovering damaged ancient texts.

Additionally, it’s worth noting that in October 2021, specialists from Google AI recreated destroyed paintings by Gustav Klimt using artificial intelligence.