What is an AI hallucination?
Answer
When generative AI tools like ChatGPT or Gemini create false content it is often called a "hallucination". In the library we often see these so-called hallucinations when students are trying to locate a book, article, or other information resource that doesn't exist.
It's important to note that this is not an example of an AI tool working incorrectly. Large Language Models, or LLMs, are trained on data to find patterns and then create new content. Although the content is presented as though it is fact, it is often false, misleading, overly simplified, or heavily biased. There is no evidence that this can be solved by asking better questions.
Here is an example from an article in Nature on how fakes look real, Hallucinated citations are polluting the scientific literature. What can be done?
