In our story, the station has a pretend log helper that writes reports.
Rocket reads its report. "You took 21 mission photos on Monday. That is certain."
"Great! Let's add that to the poster," Rocket says.
Raven opens the paper logbook. "What do you notice? Our tally says 12."
Nova hovers between them. "Which one did you count yourselves?" she asks.
Some AI tools write or make things, like answers, stories or pictures.
These tools can say something wrong in a very sure way.
People call this a hallucination. The tool is not seeing things. It is just wrong.
The pretend log helper said 21. The crew counted 12. The helper sounded sure, but it was wrong.
Real AI tools can be wrong too. That is why we check.
| Day | Logbook tally (the crew counted) | Pretend helper said |
|---|---|---|
| Monday | 12 | 21 |
| Tuesday | 15 | 15 |
| Wednesday | 9 | 19 |
| Statement | True or false? |
|---|---|
| If a tool sounds very sure, it must be right. | ? |
| AI tools can say wrong things in a sure way. | ? |
| People call a sure but wrong AI answer a hallucination. | ? |
| The crew's own tally is a good way to check. | ? |
Rocket almost put 21 on the poster. That would have spread a wrong fact.
Checking first kept the poster true.
Nova is a story character. Real AI tools can be wrong, so we check them.
Great checking! Tomorrow we learn where to look when we need to check.