LLMs aren’t world models

Comments

• hendrik@palaver.p3x.de ⁨2⁩ ⁨days⁩ ago

  Oh well, LLMs are notoriously bad at chess. Other tasks as well. Maths, especially counting... But I'd say this is because they're not very intelligent. And not trained to be a chess computer. So a clever middle-school kid will outperform them at chess. Or a kindergarden kid at counting letters in words. But that doesn't directly lead to the conclusion they're not world models. That'd conclusion needs some qualifying, quantification and a mathematical proof. This is just opinion.

  source
• ninjabard@lemmy.world ⁨2⁩ ⁨days⁩ ago

  Or maybe don’t use LLMs since they’re poisoning the planet and the Internet?

  source
• ulu_mulu@lemmy.zip ⁨2⁩ ⁨days⁩ ago

  Ofc they’re not, they’re not made for knowledge not intelligence, they’re made to “predict” language patterns based on probability.

  source