Discussion about this post

User's avatar
SkinShallow's avatar

I think the prompt is good, but perhaps I already knew too much about my chosen area so I kept catching the robot ignoring important local considerations entirely (eg difference between Scottish and English legal systems leading to difference in supporting professions too eg in regards of qualifications) and then confabulating/hallucinating quite a bit when caught and confronted. Which is normal AI behaviour, nothing to do with the prompt -- but it significantly affected my trust in its deliberations.

U feel it used very general info in its attempts to hone down the requirements, ignored non standard paths, much of it was obvious, oddly ignored information it has about me already (or perhaps couldn't integrate it -- I keep forgetting those things don't UNDERSTAND the world ;) and thus was of limited use for my specific purpose.

So in a tl;dr sense: a good prompt crashing against the typical problems with LLM chat bots (no niche knowledge in its training data, limited niche knowledge available online; no ability to understand anything or genuinely generate new knowledge about the world based on such understanding)

Expand full comment

No posts