I’ve yet to hear mention of the “Prompt Conundrum.” What I coin a human problem to address your favorite LLM with the proper prompt is easily dismissed as “you’re not holding it right…”. It’s not. LLMs, in their current iterations, will at best be engines of approximation, used by approximate beings striving and ever failing for perfection. Faced with the daunting task of learning the entire context of a position before asking an LLM a question, people will “do their best,” gathering context as quickly and efficiently as possible, as much as quite possibly, an entire careers worth of context, to ask the proper question. If we are to equip people with tools to provide them with approximate answers, how then do we equip them to ask the proper question to arrive at the proper answer without first training them on the data set? The “Prompt Conundrum” we all face is not merely what to ask about subjects we know little about but also why we should ask them at all in the first place?

AI is a fantastic tool for those with context.

Human first and unabashedly so …

-Adam