Wednesday, May 24, 2023

Oh bother

It seems that people have been trying to use ChatGPT to translate Linear-A, but there are just too few texts to allow translation.

Oh, well.  I've thought for some time that they would just be ancient Minoan shopping lists, anyway. 

2 comments:

Jonathan H said...

This reminds me of translation algorithms on Star Trek - the problem is that human language has too many rules, and even more exceptions, to handle without a large sample to generate a large rulebook.
I don't know what the minimum data size would be for a project like this, however having known some linguists, it has to be not only large but also diverse in specific ways to capture a usable picture of the language.

HMS Defiant said...

As I recall, that was the 'big reveal' in A Canticle for Lieberwitz.