05
Lesson 5 · 2 min
Why AI makes things up
AI is built to produce a likely answer, not a true one, so sometimes it confidently invents facts.
When a chatbot invents something, like a fake book title, a wrong date or a law that does not exist, it is called a hallucination.
It happens because the model is predicting text that looks right. If it does not know the answer, it rarely says so. It fills the gap with something that fits the pattern.
Hallucinations are most common with exact details: names, numbers, quotes, links, citations and recent events. They are less common with general explanations.
Newer models hallucinate less, and many can search the web and show sources. But no model is free of it yet. Try it below: three answers, one is invented. Can you spot it?
Spot the made-up fact
Each answer sounds sure of itself. One contains something invented. Tap the one you think is fake.
Who wrote the novel “One Hundred Years of Solitude”?
Do it now
Make it admit what it does not know
- 1.Ask a chatbot about a small local topic, like a street, shop or teacher in your town.
- 2.See whether it gives details it could not possibly know.
- 3.Then use the prompt below and compare the answer.
Tell me about [SMALL LOCAL TOPIC]. If you are not sure about any detail, say "I am not sure" instead of guessing. Mark each fact as confident or unsure.
Words from this lesson
- Hallucination
- When an AI states something false with confidence, such as a made-up fact, quote or source.
- LLM
- An AI trained on huge amounts of text to predict the next word, which lets it write, summarise, translate and chat.
- Training data
- The examples, such as text, images or recordings, that a model learns its patterns from.