Steve Fenton ➜ on Nostr: I once worked on an OCR project, and the execs never grasped the fundamental problem ...

I once worked on an OCR project, and the execs never grasped the fundamental problem of OCR (which IMO applies to LLMs):

If something is wrong 20% of the time
And you don't know which 20%
You have to check 100% of it's work