What is AI, ML, and an LLM - really
AI is the broad idea of machines doing tasks that normally need human judgement. Machine Learning is one way to build AI - instead of writing rules by hand, you show a system millions of examples and it learns the pattern itself. A Large Language Model (LLM) is a specific kind of ML system trained on huge amounts of text, whose only job is: given some text, predict what text comes next. That's it. Everything else - chatting, coding, reasoning - is that one skill used cleverly.
During training, the model reads enormous amounts of text and adjusts billions of internal numbers (parameters) so that its next-word predictions get statistically closer to real text. It never "understands" in a human sense - it's compressed statistical structure over language. At inference (when you chat with it), those learned weights are frozen; the model just runs prediction, one token at a time, using whatever text is in front of it (the prompt) as its only source of "memory."
Open Claude or ChatGPT. Type one word, then keep hitting "continue" or asking "what's the next most likely word" mentally before reading the actual output. Notice: it's not "thinking" ahead - it's building the response piece by piece. Try asking it to write a sentence and then ask it "why did you choose that specific next word" - see how it explains its own pattern-following.
What is an LLM literally doing, at each step, when it responds to you?