A jargon-free explanation of how AI large language models work

The design of Large language models (LLMs) such as ChatGPT to predict the next word requires a vast amount of text for their training. Built on a neural network trained using billions of terms of everyday language, unlike software, this makes it challenging for any individual to understand their internal workings fully. The article aims to make significant portions of this knowledge accessible to a broader audience by explaining training, word vectors, and the transformer. 

Source: ARS Technica