Geoffrey Hinton: LLMs understand what they say

Geoffrey Hinton argues that LLMs truly understand language, explaining that neural networks combine words like Lego blocks to form meaning rather than using an internal symbolic code.