Hacker News Reader

I’m guessing one stone is not enough training data.

Not as such. AI isn't magic, it's fuzzy pattern matching based on existing data. We couldn't just show it some stone out of nowhere and have it "think" or "reason" about what to do with it.

If on the other hand you had a model otherwise trained on a bunch of hieroglyphics<-->English data, then sure probably.

Well I know that it can't reason. But we know two translations and hieroglyphics are the third. So how close could it get to "solving" the hieroglyphics to Greek pattern match?

How would that be enough context and how would you backtrack the reasoning to check for correctness / completeness

That a fun and interesting question tho! I would bet on « absolutely not, get over it »