This has been posted several times before, but has not had any discussion within the past year.
In the more recent context of widely available LLMs, I am specifically fascinated by the idea of "Freddish" as prompt engineering for young kids. Children and LLMs are alike in many ways (they latch onto patterns and occasionally make faulty inferences) and differ vastly in others (the innate knowledge in a child's neural architecture is of an entirely different character than the textual dataset of an LLM).
I am curious if HN has any other insights about commonalities in the "demanding process of getting all the words just right" for when talking to both.
1 comment
[ 3.4 ms ] story [ 13.9 ms ] threadThis has been posted several times before, but has not had any discussion within the past year.
In the more recent context of widely available LLMs, I am specifically fascinated by the idea of "Freddish" as prompt engineering for young kids. Children and LLMs are alike in many ways (they latch onto patterns and occasionally make faulty inferences) and differ vastly in others (the innate knowledge in a child's neural architecture is of an entirely different character than the textual dataset of an LLM).
I am curious if HN has any other insights about commonalities in the "demanding process of getting all the words just right" for when talking to both.