My understanding is that there are a huge number of undesciphered clay texts from the middle east that remain undecisphered for lack of scholars.
A couple of ai contests would probably get most of them deciphered. Why? they already have a number of texts deciphered and translated. So it wouldn’t be too tough to train the ai on these texts to read the unread stuff.
I wonder how many different languages or age separations exists between these various clay records. An AI trying to work on English literature 2,000 years in the future might have a bit of difficulty translating the English of 1066 from that of 1566, for example. I think the linguistic differences between 1066 and 1566 were probably a lot greater than those between 1523 and 2023. The invention of the printing press probably helped reduce the changes in literary language closer to our own time.
Also some cultures/nations have a more rigid control over their formal language. I have a modern copy of a book written in Spanish of the 1500s. With a copy of a similar document written in English then and today, it is much harder to read the English. Spain had a very hard working Academy that supervised the official language for centuries. English did not so far as I know. Variations of modern Bible translatons compared with the King James version are an example. An AI system will have to deal with these many variations in clay texts. Probably human scolars will have to do some sorting by the criteria I mention before any machine reading can be effective.