Re:Large language models reduce public knowledge sharing on ...
by creer on Sunday October 13, 2024 @07:00PM
(#41832504)
And even for trivial questions, there is a lot out there that the doc developers ignored or hid.
For me, the current state of affairs is the difficulty of aiming search engines to the right answer. The answer might be out there, but figuring out HOW t oask the questions, which are the right keywords, etc - basically requires knowing where the answer is. LLMs have potential in rephrasing boththe question and what they might have glanced at here and there - even if obvious in hindsight.
Re:Large language models reduce public knowledge sharing on ...
by skydhash on Sunday October 13, 2024 @07:25PM
(#41832653)
Don’t use search engines unless you know what you’re searching. Which means I start with another material first (books, wiki, manual,…) which will give me enough ideas of what I want will be. Starting with a search engine is like searching for a piece of a jigzaw with only the picture on the box, you have to know it’s shape first and where it would go.
Re:Large language models reduce public knowledge sharing on ...
by xxs on Monday October 14, 2024 @07:02AM
(#41836351)
It's a very similar situation with concurrent programming. Knowing which instruction does a CAS, or the exact memory semantics of a particular load/store/etc., doesn't do much on its right own.
I've answered dozens of questions on stackoverflow.com with tags like SIMD, SSE, AVX, NEON. Only a minority of these asked for a single SIMD instruction which does something specific. Usually people ask how to use the complete instruction set to accomplish something higher level.
Documentation alone doesn't answer questions like that, you need an expert who actually used that stuff.