Conversation

Notices

1. Embed this notice
   Sick Sun (sun@shitposter.world)'s status on Monday, 10-Jun-2024 09:17:52 JST Sick Sun

   • miauz genyau
   • condret :verified: :ancom:
   • Rua
   • anemone
   @mia @condret @ruawhitepaw @anemone when I think about my knowledge of many topics I realize that I just know a big bunch of facts and more facts get added I can make bigger and bigger amalgamations of facts that make it sound like I know real things about topics like space explanation, but really I'm just reproducing what an LLM does and Im just being honest about it
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Monday, 10-Jun-2024 09:59:28 JST Sick Sun

     in reply to

     • miauz genyau
     • condret :verified: :ancom:
     • Rua
     • anemone
     • Pete Zaitcev
     @zaitcev @anemone @condret @mia @ruawhitepaw something that approximates a human will be a combination of LLM and reasoning imo
   • Embed this notice
     Pete Zaitcev (zaitcev@shitposter.world)'s status on Monday, 10-Jun-2024 09:59:29 JST Pete Zaitcev

     in reply to

     • miauz genyau
     • condret :verified: :ancom:
     • Rua
     • anemone
     @sun @mia @anemone @condret @ruawhitepaw
     While true in a certain way, it is not a productive view. It's like saying that love is a chemical reaction. The reason why AI of Google produced the absurd results is that it ilacks a world model. In theory it should be possible to train a model that approximately your thinking, but so far LLMs were fundamentally dissimilar to you.
     Linux Walt (@lnxw37j1) {3EB165E0-5BB1-45D2-9E7D-93B31821F864} likes this.