Skip to content
Young Makers: AI for KidsLesson 4 of 11

4 · AI can be wrong — always check!

This might be the most important lesson in the whole course, so read it twice:

AI can be wrong — even when it sounds totally sure. Always check what it tells you.

Remember from Lesson 2 that an AI chat helper is really making a smart guess about what words come next. It's trying to sound right, but sounding right and being right are not the same thing. Sometimes an AI will make up something that is completely false and say it with a big confident smile. Grown-ups have a funny word for that: they call it a hallucination (say it: huh-loo-sih-NAY-shun). It means the AI "saw" something that isn't there — like a made-up fact, a fake book title, or a wrong date.

Here's the tricky part: a hallucination doesn't look like a mistake. It looks neat and sure, just like a true answer. That's exactly why you have to check.

How to check — be a fact detective:

  • Ask a grown-up. A parent, teacher, or librarian can help you check.
  • Look in a trusted place. A real book, a kids' encyclopedia, or a trusted website. If the AI and a trusted source disagree, the trusted source wins.
  • Use AI for the easy stuff, double-check the important stuff. Spelling help or brainstorming? Less risky. A fact for your homework, or anything you'll really count on? Check it.
  • Never just trust a number, a date, or a "fact" an AI gives you without checking. Numbers are easy for AI to get wrong.

The grown-ups who write safety rules for AI (at a U.S. group called NIST) say good AI should be "valid and reliable" — fancy words that just mean correct and trustworthy. Until you've checked, treat an AI's answer like a clue, not the final answer.

Think about it. An AI tells you a "fact" for your report and sounds very sure. What are two things you could do before you write it down?

Sources

Sign in to track your progress