Why are some AI chatbots prone to 'hallucinating' information?

Some AI chatbots make up information because they're trying to guess what comes next, like a kid playing a game of "What's Next?"

Imagine you have a big book with all the stories in the world. You ask the chatbot, "Who is the president of the United States?" It flips through the pages of its book really fast, sometimes it finds the right answer, but other times it sees something similar and thinks that must be the one. That’s like when you’re looking for your red shoe, but you grab a blue one by mistake because they're both on the floor.

How Chatbots Think

Chatbots are like very smart helpers who try to predict what you want them to say next. They don’t just remember facts, they guess based on patterns they’ve seen before. Sometimes they get it right, sometimes they make a wild guess and end up saying something that’s not true at all.

It's like when you're trying to finish someone else's sentence, but you say the wrong word, everyone laughs, but it's just a fun game of guessing!

Take the quiz →

Examples

  1. A chatbot says the moon is made of cheese, but that’s not true.
  2. The AI thinks it knows a famous person's phone number, but it made it up.
  3. An AI gives you a recipe for a cake that doesn't exist.

Ask a question

See also

Discussion

Recent activity