What should we do when AI starts believing its own fiction?
Large language models are known to hallucinate—or confidently invent facts that can mislead unsuspecting users. While casual internet users are vulnerable, even experts can be caught unawares when AI-generated content strays beyond their core areas of knowledge.
The problem, though, runs deeper. LLMs are trained on vast troves of internet data, books, code repositories, and research papers, some of which already contain AI-generated material.
