Module 3: Knowing what not to trust — Knowing what NOT to trust is huge. We'll learn how to spot when AI makes things up.
- Each of these AI-generated claims contains a specific type of hallucination. Match the claim to the hallucination type.
- — Ask yourself: did the AI invent something, mix things up, use old data, or reason incorrectly?
- AI tells you: "The Great Wall of China is visible from space with the naked eye." What type of hallucination is this?
- Conflation — the AI merged a popular myth with factual statements, presenting it as verified truth. — The Great Wall exists and space exists — but the AI mixed a common myth into its factual output.
- AI recommends: "Take the highway because it has fewer traffic lights, so you'll arrive faster for your 3 PM meeting." But the highway has a 2-hour construction delay today. What type of hallucination is this?
- Bad Logic — the reasoning sounds logical but ignores critical real-world context. — The logic chain had a valid premise (fewer lights = faster) but missed a crucial variable.
- AI states: "Slack is free for teams of any size with unlimited message history." But Slack changed to a 90-day limit on free plans in 2022. What hallucination type is this?
- Old Information — the AI gave accurate info from its training data, but the reality has changed. — This was once true. The AI's training data is simply outdated.