Generative AI for Everyone

Lesson 1

AI Refresher

Five minutes to re-anchor the few ideas the rest of this course leans on.

The one idea that matters most: next-token prediction

A chatbot generates text by predicting the next small piece, one step at a time, based on patterns it learned from data. It is autocomplete scaled up enormously.

the
Temperature 0.90
Top-k 0 (off)
Top-p 1.00

Temperature →0 = greedy/repetitive; →2 = flat/random. Top-k keeps the k tallest bars; top-p keeps the smallest set summing to p (nucleus). Grey bars are excluded, then re-normalized. (Toy seeded model — illustrates the loop & sampling.)

Everything practical in this course traces back to this. It is why AI is brilliant at drafting, why it sometimes states false things with confidence, and why how you ask changes what you get.

Three consequences you will use constantly

What this makes AI great at

  • Drafting, rewriting, summarizing, brainstorming
  • Producing many plausible options fast
  • Anything where a strong first draft saves you time

What you must watch for

  • Confident-sounding answers that are simply wrong
  • No built-in sense of what it does not know
  • Output quality swings hugely with the context you give
  1. Plausible, not guaranteed. AI is a probabilistic predictor, not a database of facts. Trust it where plausible is good enough; verify where exactly correct matters.
  2. Narrow, not general. It is dazzling inside its trained patterns and brittle outside them. (The capability ladder, from AI Literacy: we are firmly on the narrow rung.)
  3. Context is the lever. The biggest driver of output quality is the context you provide: your documents, your data, your examples. You will lean on this in every build lesson, and go deep on it in the Prompt Engineering course.

That is the whole refresher

Three ideas: next-token prediction, plausible-not-guaranteed, context-is-the-lever. Hold those and the rest of this course is hands-on from here.

What you can now do

  • Restate next-token prediction in one sentence
  • Decide where to trust AI (plausible is fine) versus verify (exactness required)
  • Explain why context is the single biggest lever on output quality

Check your understanding

1 / 2
next-token-prediction

What is the single mechanism behind how a chatbot generates a response?

What's next

Refreshed. Now let us get reliable output on purpose: Talking to AI.

Test your understanding

Prof is ready

Prof will ask you questions about quick AI refresher for the gen-AI course — not explain it. You'll be surprised what you don't know until you have to say it.

Finished this lesson?

Read through the lesson first (0/20s).