Skip to content(if available)orjump to list(if available)

Researcher uses AI to make texts that are thousands of years old readable (2023)

Researcher uses AI to make texts that are thousands of years old readable (2023)

8 comments

·March 19, 2025
Visit

pimlottc

Nothing in the article even mentions artificial intelligence. Sounds more like a standard deterministic algorithm.

sapphicsnail

It talked about digitization but that could mean a lot of things. Are they just putting pictures of the fragments up? Is there a transcription of the fragments. Can the algorithm transcribe a fragment or does someone have to do that and then feed it to the algorithm? I don't know as much about cuneiform tablets but people get PhDs just to be able to recognize what letters are on a papyrus fragment. I'd be surprised if a computer could do that already.

mistrial9

I looked quickly at a github dot com repo of the same name to investigate this question. It seemed like a portable GUI desktop app that showed catalogs of images with some relevant shortcuts, written in C++.. at least ten years old

jeroenhd

The AI part seems to be part of the OCR algorithm. OCR has been using AI techniques for ages, so there's nothing new here, other than that this project is making good progress the more valid data they can decode with help of this, the more training data they have and the better subsequent OCR will be able to process near illegible tablets.

WalterBright

I am heartened by AI now being able to decipher my grandma's letters.

aaron695

[dead]

WalterBright

> around 200 scholars worldwide have had access to the online platform for their research projects. Now it is to be made available to the public as well.

I wonder - why hide it from the public and only allow 200 scholars access to it?

I recall the random guy who figured out how to read the charred scrolls of Herculaneum.

teleforce

There is a very big stake at risk if you let loose prematurely any old ancient manuscripts for those who are in authority.

For example imagine if the earlier ancient version manuscripts of holy books would say the totally opposite about the later version of the widely accepted holy books. Or the newly discovered ancient holy book manuscripts prophesized exactly in details of an upcoming prophet, but he belongs to the different later religion that the previous religion (of the newly discovered ancient religious manuscripts) has been denying all along for centuries. The list can go on but I think you get the point.

True story, researchers got personal death threat when they discovered statistically that Indus valley script are most probably alphabets for language not random symbols. By the way Indus valley script still has not been deciphered until now but those who do probably going to get Nobel prize to their names.