Issues

Hallucination

When an AI model confidently generates false, nonsensical, or unverified information.

Detailed Explanation

In the context of AI, a hallucination occurs when a Generative AI model produces an output that sounds plausible but is factually incorrect, nonsensical, or disconnected from the provided context. Hallucinations happen because LLMs are fundamentally probabilistic engines—they predict the most likely next word rather than verifying facts against a source of absolute truth. Techniques like RAG are used to minimize this.

Related Terms

Fine-Tuning

The process of taking a pre-trained AI model and training it further on a smaller, specialized dataset.

Large Language Model (LLM)

A type of foundational AI model trained on vast amounts of text to understand and generate human language.

RAG (Retrieval-Augmented Generation)

A technique that connects an LLM to external knowledge bases to improve accuracy and reduce hallucinations.

Back to Glossary