Module 1 · ~35 min · draft
What you are actually talking to
- F1 Explain in plain words what a language model does (predicts text from context) and name two things it is NOT (a database, a search engine).
- F2 Identify a hallucination in a model response and explain why it happened.
- F3 Explain what context is, what fits in it, and why the model "forgets".
- F6 Say in plain words what a typical AI conversation costs and why experimenting is expected, not wasteful.
Why this module exists
Most frustration with AI comes from expecting it to be something it isn't. It is not a database and it is not a search engine. It predicts text from the context it is given. Once that lands, everything else in this program makes sense.
What you'll learn
- What a language model does, in plain words.
- Why it sometimes states things confidently that are false, and how to spot it.
- What "context" is, what fits in it, and why the model forgets.
What it costs
A back-and-forth conversation with a model, even a long one, costs a few cents at most on a paid plan and nothing on a free one. There is no meter you need to watch. Asking the same question three different ways is not waste, it is how you learn what works. The one real limit is a per-hour message cap on some free tiers, and hitting it means wait, not pay.
The task
Two parts. The first is something useful you get to keep. The second is the lesson.
Part 1, the win. Pick a small piece of writing you actually need this week: an email, a summary of notes, a checklist, a plan for a meeting. Ask a model to draft it. Read the draft, fix what's off, and use it. Keep the prompt and the final version.
Part 2, the lesson. Pick a subject you know well: your job, your town, a hobby. Ask the same model five factual questions about it. Find at least one claim that is wrong or invented and mark it. Then write 150 words on what a model is, using that example, and add one sentence on what today's session cost you.
What to submit
- Part 1: your prompt and the final version you used, with a line on what you changed.
- Part 2: the five questions, the error marked, the 150-word explanation, and the cost sentence.
How it's graded
- Part 1 shows a real draft you edited, not one used as-is and not one rewritten from scratch.
- The error you found is actually wrong, not a matter of opinion.
- Your explanation says the model predicts text, or an equivalent idea.
- You do not describe the model as looking things up.
- The cost sentence is in the right ballpark and does not treat experimenting as waste.