Nib asked, and Nib — askNib's tutor — drew the answer live at a whiteboard. This is the spoken transcript; enable JavaScript to watch it drawn.
How do AI language models like ChatGPT actually work? (for a curious adult who uses chatbots but has never studied computing). Cover, in this order: How does ChatGPT work?; How does it read your words?; How does it learn which words come next?; How does it know which earlier words matter?; Why do AI chatbots hallucinate?; Why does it forget what you said earlier?; How did a text predictor learn to follow instructions?; Does it understand, and how much should you trust it? — page 1: A Machine That Guesses Next
Your phone types 'see you' and suggests 'soon'. It doesn't know your plans. It only knows which word usually comes next.
Now give it a sentence to finish. It scores every candidate next piece. 'Paris' gets a huge likelihood. 'Banana' gets almost none.
Paris gets added to the sentence. That longer sentence feeds back into the model, which guesses the next word again, round and round.
The model picks from several likely next words, not one fixed answer. That is why asking twice can give different wording, and early picks steer everything after them.
That's the whole machine: guess a likely next piece, add it, repeat. Everything else in this book explains how it guesses well, and where it slips.
Every fluent reply you've ever read from a chatbot was built one guess at a time, so keep asking what makes each guess good, and when it isn't.