On Large Language Models for Understanding Human Language Christopher Manning 

4:30 "now at first sight this doesnt exactly seem like it could possibly be the path to artificial intelligence (AGI), however it turns out that this is a very effective task precisely because predicting words can depend on any aspect of the text meaning and any knowledge of the general world around you and so it works very well as a universal pre-training method that gives models wide understanding that can then be deployed for all sorts of general tasks...". This is why prediction, applied to language, is so powerful.

Thank You for sharing such a nice and interesting video. Really a great talk

This lecture is full of jargon to me. May I ask which university grade this lecture is given ? And how long it take me to get there ?

◇°• So; Am I Understanding Even A Part of This Process of "Natural Language Understanding"; that Is Desired of an AI Entity By Its' Programmr?◇°• The LLM (Large Language Model) Decides What is the ONE? Relevant~Correct Best Answer; to a: question; query~ problem...and NOT the Initiator of the Question?◇°• The LLM Negates "Multiple Responses" for a: Favored; Wanted; Desired~Predicted "ONE" Outcome by the Programmer; Even IF That ONE Answer Response Is NOT Correct; Or The Best Answer?!◇•☆

ChatGPT is able to learn. ChatGPT can be trained. However, ChatGPT is unable to understand😢

Fantastic, I was just going through the good professors cs224n