Skip to main content
tech futures.Introduction to AI

Lesson 3: Trust and AI

What are hallucinations?

Sometimes AIs make up information and we call these hallucinations.

An AI hallucination is information in an AI response that is wrong.

It's important to know that AIs do not think like a human does and do not have an understanding of the world like we do. And although they learn from huge amounts of information, they can make mistakes and say things that are not true.

Some people will try and find prompts that cause hallucinations, which can help people understand strengths and weaknesses of AIs. An example of one of those prompts that caused some hallucinations in recent LLM models is shown in the chat below.

Looking at the chat above, where has the AI made a mistake? Have a think about it and when you have your own answer, click the Reveal Answer button below.

LLMs learn from huge amounts of data but that doesn't mean they are always right! LLMs can answer incorrectly to questions that might seem simple to us.