Hallucination

Hallucination is when an AI model produces output that sounds authoritative and well-reasoned but is factually wrong. The model might cite a paper that does not exist, describe a function that does not work as stated, or confidently provide incorrect historical dates. The term comes from the analogy to human hallucination - perceiving things that are not there.

Hallucination is the single most important limitation to understand when using AI tools. If you blindly trust AI output, you will eventually use incorrect information. This has real consequences: lawyers have been sanctioned for citing AI-hallucinated case law, students have submitted fabricated references, and developers have deployed code with bugs that AI confidently declared correct.

Always verify critical information. For code, run it and test edge cases. For facts, check the source. For important decisions, cross-reference with authoritative sources.

Anthropic: Reducing Hallucination in AI - https://www.anthropic.com/research

Wikipedia: Hallucination (artificial intelligence) - https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

NYT: "Here's What Happens When Your Lawyer Uses ChatGPT" - https://www.nytimes.com/2023/05/27/nyregion/avianca-lawsuit-chatgpt.html

Google Research: Survey of Hallucination in Natural Language Generation - https://arxiv.org/abs/2202.03629