AI glossary
Hallucination
A hallucination is an answer from a language model that sounds confident and fluent but is not supported by its sources or is simply false.
A model writes text that is likely, not text it has checked. When it does not know, the likely-sounding answer can be an invented citation, a function that does not exist or a wrong figure.
You cannot remove hallucination, but you can make it rarer and easier to catch: give the model the source documents, ask it to quote them, let it say "I do not know", and measure how often answers are unsupported.
A common mistake is to trust the model's own confidence. A wrong answer is phrased just as calmly as a right one, so a check has to come from outside: a source, a test or a person.
Learn it hands-on
These free lessons teach it with commands to run in a lab and checked tasks. Read them whole with no account.