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 8: Understanding and Trust
Does it understand, or just match patterns? The answers feel smart, but experts still disagree on this.
Skeptics say it's fluent without checking reality. Others say good prediction forces useful internal models. Both points carry weight, and the debate continues.
You don't need to settle that question. Either way, it sounds equally confident when it's right and when it's wrong.
Remember, it's built to sound plausible, not to check facts. An invented citation reads exactly like a real one. The prose gives you no warning.
So check hardest where errors are likely and costly: facts, numbers, quotes, citations. Find the original source, confirm the number, make sure the paper exists.
As a draft, it's wonderful: outlines, explanations, rewrites, summaries of text you paste in. It gives you the first version. You do the checking.
So hold both ideas: it may be doing something that deserves a respectful question mark, and it can still be confidently wrong. Use it as a brilliant, fast draftsperson, and keep your own hand on the facts.