Artificial Intelligence: Now AI Could Unlock Secrets From 2,500-Year Old Iranian Tablets
Machine learning and AI are helping researchers learn about Persia’s Achaemenid Empire.
Researchers at the University of Chicago’s DeepScribe project have deployed machine learning and artificial intelligence (AI) to decipher thousands of clay tablets dating back to the ancient Iranian Achaemenid Empire (550–330 BC). (TNW)
Researchers from the University’s Oriental Institute and its Department of Computer Science are collaborating on DeepScribe. It is a potentially pathbreaking breakthrough because humans and computers have been struggling to make sense of the thousands of Achaemenid clay tablets loaned to the US by Iran. The collection of tablets is known as the Persepolis Fortification Archive (PFA).
How AI will help
The researchers are training an AI model on 6,000 annotated images received from the PFA. Once trained, the algo would read the tablets that remain to be deciphered.
Asstt. Prof. Sanjay Krishnan of the Department of Computer Science applies deep learning and AI techniques to data analysis, including video and other complex data types. He said of DeepScribe’s Iranian tablet project: “It’s a good machine learning problem because the accuracy is objective here, we have a labeled training set and we understand the script pretty well and that helps us. It’s not a completely unknown problem.”
However, time is of the essence here as the tablets are now on their way back to Iran. Discovered in the 1930s in the ancient city of Persepolis, four consignments of the tablets have already been delivered in 1948, 1950, 2004 and 2019 to Iran.
Researchers are therefore currently digitizing high-resolution images of the tablets. The size of this still-expanding database is 60 terabytes. It was the source for the creation of a dictionary of Elamite – the language of the tablets.
In testing post-training, Krishnan’s model successfully read cuneiform signs (on tablets outside the training set) with 80% accuracy.
Further research and training are likely to improve the accuracy of the model further.
In promising possibilities, DeepScribe could evolve into a broad-based AI tool. Other archaeologists could also use it, while it could also decode other cuneiform languages.
Latest Alternative Investment News
Flowcarbon, co-founded by Adam Neumann, known for WeWork, is a climate technology company seeking to formalize the market for carbon offsets on the blockchain. The startup announced Tuesday its $70…
One hapless investor lost about 3 billion won ($2.4 million) in the LUNA and UST implosion earlier this month. A Korean social media personality going by the name of “Chancers”…
Scientists have long held the view that earthquakes are impossible to predict, being random events. But the potential for damage these natural events have, including the loss of life, have…
Digital Assets: Swiss Venture Firm CV VC Launches African Blockchain Early-Stage Fund; Issues African Blockchain Report
CV VC announced Monday at Davos the launch of its African Blockchain Early-Stage Fund to bring the advantages of the blockchain and its own expertise to enable Africa to transact…