11 · Practice: safe, smart builder

Fill in the missing word from Parts 5–6. Spelling is forgiving — get the idea right.

1. 1. Because an AI-powered robot's decision is a ___, it can be confidently wrong — which is why we design when it must ask a human.

   Check

2. 2. For the highest-stakes action, engineers often use a plain, readable ___ instead of a learned guess, so the safety behavior is predictable.

   Check

3. 3. Keeping a person able to confirm or override a robot's risky decisions is called keeping a human in the ___.

   Check

4. 4. Machines doing work on their own is called ___; it can remove some jobs while creating new ones, like building and training the robots.

   Check

5. 5. When a robot's AI works better for some people than others because of who was in the training data, that's a fairness or ___ problem.

   Check

6. 6. When you fix a confused model, you don't add more rules — you ___ by giving it better, more varied examples.

   Check