Untitled attachment

Notices where this attachment appears

1. Embed this notice
   Jason Gorman (jasongorman@mastodon.cloud)'s status on Monday, 30-Dec-2024 18:10:58 JST Jason Gorman

   Playing GPT-x at chess reveals a major limitation of Large Language Models: a complete lack of dynamic reasoning.

   So many tasks require the ability to plan ahead and evaluate potential outcomes (e.g., is this a good chess move or a bad chess move?)

   It's arguably the whole point of intelligence.

   This is coupled with a complete lack of temporal reasoning. LLMs have no sense of history or ordering of events.

2. Embed this notice
   samir, a distributed system (samir@mastodon.functional.computer)'s status on Saturday, 30-Nov-2024 06:02:52 JST samir, a distributed system

   in reply to

   @hrefna Oh, I love this observation.

   I’ve been very guilty of protesting against “magic” in the past. On reflection, I think what I’m wary of is leaky abstractions. I want to be able to reason about my code/software, and I’m fine with magic as long as it doesn’t interfere with the reasoning.

   So thank you for making me dig into this.

3. Embed this notice
   Prof. Emily M. Bender(she/her) (emilymbender@dair-community.social)'s status on Friday, 10-Mar-2023 23:27:23 JST Prof. Emily M. Bender(she/her)

   in reply to

   What matters about language in the context of this tech is that language and meaning are relational, that communication is a joint activity, and that systems set up to mimic the form of language can provide the illusion that they understand, that they know things, and that they are reasoning.

   So, read this, not that:

   https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html