MIT graduate Jiaming Luo developed an AI that successfully deciphered ancient scripts. Along with his colleagues, Luo developed a model that was taught by patterns in how languages change over time. This was done by feeding the machine words in a lost language and in a known related language, where the AI would then try to align the words in the lost language with its known language counterpart.
With that being said, the model was tested on two ancient scripts: Ugaritic, and Linear B, a language that was on the ruins found in Crete. These two languages were already deciphered. The algorithm was able to correctly translate with remarkable accuracy. The model’s 67.3% accuracy gave researchers an interesting idea to work on: can artificial intelligence be capable of translating undeciphered texts? Alizeh Kohari discusses current research on the matter here!
Image credit: The Trustees of the British Museum