Hallucination
In the field of artificial intelligence, Hallucination refers to AI models (such as chatbots) generating information that sounds plausible but is factually incorrect or completely unfounded.
How it works
This phenomenon typically occurs in generative AI such as large language models (LLMs): when the model infers from patterns in its training data but lacks sufficiently accurate information or cannot tell fact from fiction, it ends up 'making up' content. Understanding and reducing hallucinations is essential to ensuring the reliability and trustworthiness of AI output.
Examples
- 1.When asked about a historical event that never happened, an AI chatbot may fabricate detailed dates, figures, and storylines.
- 2.When generating code snippets, an AI coding assistant may reference a library function or API that does not exist.
- 3.When summarizing a long article, an AI summarization tool may insert key conclusions or data that never appeared in the original text.