IPs and Ports are like street address and room number.
LLMs ie chatGPT are like a Galton Board. Your sentence is the bucket of balls at the top, each ball is a word or ‘token’, and the LLM is the arrangement of pegs on the board. Training the model is like moving the pegs around until the pattern you get at the bottom is desirable.