There are 2 kinds of usecases that software automates. 1) those that require accuracy and 2) those that dont (social media, ads, recommendations). Further, the

There are 2 kinds of usecases that software automates. 1) those that require accuracy and 2) those that dont (social media, ads, recommendations). Further, there are 2 kinds of users that consume the output of software. a) humans, and b) machines. Where LLMs shine are in the 2a usecases, ie, usecases where accuracy does not matter and humans are end-users. there are plenty of these usecases. The problem is that LLMs are being applied to 1a, 1b usecases where there is going to be a lot of frustration.