Tom Morris on Nostr: It is remarkable how many scholarly papers have been published where the methods ...

It is remarkable how many scholarly papers have been published where the methods section boils down to "we asked a commercial LLM to spit out a bunch synthetic text about a topic, then we examined what it told us about X".

"Since we have no bloody clue what data the model was trained on or how representative that is of reality, drawing conclusions about human psychology or society (etc.) is silly, but we got a paper out of it" is the acknowledgement that should also be in the paper but isn't.