A language model only manipulates language
Quote
this is a language manipulation system. That's what it does it, it parses language, and it puts out and it outputs language. Anything else that you attribute to this thing? is not possible.
Episode 131: Toilet Scroller (episode page, opens in a new tab) 00:56:03 (play audio from this timestamp, opens in a new tab)
Speaker-labeled
Summary
Don't anthropomorphize LLMs. They parse and output language, and anything you can't clearly express in language is beyond them. People are shoehorning them into tasks they can't do (e.g. AI-classifying music was a coin flip).