LLMs can't do probability

Comments

• lvxferre@mander.xyz ⁨1⁩ ⁨year⁩ ago

  Before checking the HN discussion, let me guess: lots comments boiling down to “ackshyually le hyumans cant eithar lol lmao”. Missing the rather practical point of the article - to highlight something that LLMs do extremely poorly, and that you should not rely on LLMs to do it.

  /me checks comments bingo! (Including the top comment) Although to be fair some answers show how to handle this issue - ask it to create code to output the answer, and then ask it to run it. Or, you know… run it yourself, instead of doing the code equivalent of using a desktop computer as a paper weight?

  source