-
@max__drake I'm old fashioned, so it still feels weird to me that there's a whole LLM machinery being used for this. Semantic clustering of text is the task people were working on that led to LLMs - word2vec & SVD and all that multidimensional stuff!Look on archive.org
❤️ 1 Favorite Mood -2 🙁
-
@max__drake I'm not even saying that plugging into LLMs isn't the easiest way to solve this (although the old fashioned way does remove the prompt injection angle, as it has a real separation between data and logic). But it def was possible before LLMs came.Permalink Look on archive.org