I can genuinely see a use for local LLMs - they're quite good at summarising well formatted lists of data. But fuck me are they bad at being in any way not stupid.
@awfulwoman They're not even good at *summarizing* - an act which requires understanding of relative importance. Instead they can only guess based on which phrases were important in someone else's context, which may not match yours. Someone wrote a really good article on this a while back. LLMs can only "shorten", not "summarize".
@dalias @awfulwoman I believe you mean this article: https://ea.rna.nl/2024/05/27/when-chatgpt-summarises-it-actually-does-nothing-of-the-kind/ - not trying to upstage you, I just knew I had it in my bookmarks as soon as you mentioned it
@jwcph @awfulwoman Great, I didn't have it handy and was hoping someone would dig it up.