>>24623230 (OP)
It’s already worked. They won the first stage prizes for virtual unrolling. The current problem is that it’s too slow and requires too much hands on interpretation.
>>24623726
The counter reaction to “AI” (LLMs) has turned retarded as well. They’re perfectly good at doing a well defined problem. There have been mass deciphering projects running for literally decades, and they work in tandem with citizen science projects so human input and interpretation helps guide the program for the most plausible signs (or galaxy types, stars, black holes, protein chains in other projects).
Armed with the existing corpus of ancient writing it’s also quickly identifying any known scraps of texts or plausible ones. Most of the fragments belong to already known texts because the most popular shit got copied and a copy of the popular is what you are most likely to find.
