Skip to content
Train Your Own ModelLesson 10 of 12

10 · Build it responsibly & honestly

You can now build something genuinely powerful. How you use that power is what makes you a good maker — not just a capable one. Three habits to carry out of this course:

1. Be honest about what your model can (and can't) do. Your model makes guesses, and it can be confidently wrong. So don't oversell it. If you made a "is this fruit fresh?" demo, say "it's right about 9 times out of 10," not "it always knows." Claiming a model is more accurate or more fair than you actually tested is a kind of fibbing — and it can lead people to trust it where they shouldn't.

  • ✅ "I trained this on 200 photos; it works best in good lighting."
  • ❌ "My AI is never wrong."

2. Know its limits — and don't use it where it can hurt. A thumbs-up classifier is a great toy. But a model trained on your few examples should never be used to make serious decisions about real people — who's "trustworthy," who looks "suspicious," who's "in" or "out." Those uses can be unfair and harmful, especially because of the bias we covered in Lesson 8. The people who write AI safety rules, and UNICEF's guidance for AI that affects kids, stress transparency and accountability — meaning a human stays responsible for what the model is used to do. That human is you.

3. Respect people's data and consent. The examples you collect are data about people:

  • Use your own examples, or things that aren't private — not secret photos of classmates.
  • Ask first. If you want to train on a friend's face or voice, get their okay (and a grown-up's). People have a right to say no.
  • Don't post or share a model in a way that exposes someone's private images or info. UNICEF lists protecting children's data and privacy as a core requirement for AI — and that includes the data you gather.

The maker's promise: I'll be honest about what my model does, I'll use it where it helps and never where it harms, and I'll respect the people whose examples I learn from.

You've gone from "AI is magic" to training, testing, and thinking hard about a model of your own. That's real maker skill — use it to build things that are clever, fair, and kind.

Think about it. Name one good use and one harmful use for a face-recognition model someone could train. What makes the difference between them?

Sources

Sign in to track your progress
10 · Build it responsibly & honestly · ElementaryMBA