Conversation

Notices

1. Rob Landley (landley@mstdn.jp)'s status on Thursday, 28-Sep-2023 06:15:56 JST Rob Landley

   New paper shows that LLMs trained on "A is B" can't infer that "B is A"... Which was a known failure mode of "neural networks" back in 1988.

   https://garymarcus.substack.com/p/elegant-and-powerful-new-result-that

   "In math, when one make a conjecture, a simple counterexample suffices. If I say all odd numbers are prime, 1, 3, 5, and 7 may count in my favor, but at 9 the game is over... In neural network discussion, people are often impressed by successes, and pay far too little regard to what failures are trying to tell them."

   • xianc78 likes this.
   • 『VRISKA』 (vriska@lizards.live)'s status on Thursday, 28-Sep-2023 06:28:13 JST 『VRISKA』

     in reply to

     @landley they call it artificial intelligence because it's claims to intelligence are artificial

     Disinformation Purveyor :verified_think: likes this.