alt Hacker NewsI haven’t experienced this - or maybe the training data for aviation is limited. But ask a an llm for aircraft performance / flight planning data and it’s scary how bad the advice /feed back is.
Replies
uncivilized • today at 2:03 PM
Most Hacker News are working on simple JavaScript applications. LLMs quickly become unusable on any sort of specialist discipline except for the math marketing releases we’ve seen recently.
For anything where "the answer" wasn't already in the training data you just get some arbitrary correlated tokens out, as that's all a LLM can do. Of course the meaning of these tokens is just random.
(And even for things that were in the training data you don't have any guaranty they will be reproduced correctly, there is just some chance something meaningful comes out, or it doesn't, it's random.)
The whole idea to use a next token predictor as "answer machine" is completely flawed. This can't work like advertised, and that's by construction.