It’s been barely 6 months since GPT-3 was first released, but the scope and applications of AI continue to grow.
A team of archaeologists and computer scientists from Israel has created an AI-powered translation program for ancient Akkadian cuneiform, allowing tens of thousands of already digitized tablets to be translated into English instantaneously. The Akkadian Empire was the first ancient empire of Mesopotamia which flourished around 2000 BCE. Its script is known as the cuneiform script, which is perhaps the earliest-known writing system.
AI is now able to instantly translate the ancient writings in cuneiform. “We show the high potential in assisting scholars and interested laypeople alike, by using natural language processing (NLP) methods such as convolutional neural networks (CNN), to automatically translate Akkadian from cuneiform Unicode glyphs directly to English (C2E) and from transliteration to English (T2E),” the paper had said. The researchers have also released a handy Google Colab notebook that showcases their research.
“What’s so amazing about it is that I don’t need to understand Akkadian at all to translate [a tablet] and get what’s behind the cuneiform,” said Gai Gutherz, a computer scientist who was part of the team that developed the AI system. “I can just use the algorithm to understand and discover what the past has to say.”
“Translating all the tablets that remain untranslated could expose us to the first days of history, to the civilization of those people, what they believed in, what they were talking about, what they were documenting,” said Gutherz.
Translation, incidentally, is at the heart of the recent AI revolution. The paper which introduced the transformer, Attention is all you need, had demonstrated how the transformer could be used for a translation task of translating French to English. The transformer has caught on, and today forms the bedrock of several generative AI applications including GPT-4 and other technologies. And while AI has moved into all kinds of different directions, including generating images, creating video, and analysing data, it seems it’s still very useful for translations. Ultimately, computers match patterns, and it’s possible that with some labeled data, can help in translation of languages that not a lot of people speak.