New name MyDD is now RAILE (rhymes with “rail”). Same team, same mission, same login. Why the new name? →

What a Hallucination Is (and Why AI Sounds So Sure)

Transcript

A chatbot can be totally confident and totally wrong at the same time. There's a word for that.

It's called a hallucination: smooth, specific, confident text that isn't true.

Here's why it happens. A chatbot writes by predicting the next likely word from an enormous pile of writing. Nothing in that checks whether the sentence is true.

When a lot has been written about something, its guesses are usually right. When hardly anything has, it's still guessing. Take Ismay, Montana. Population 17. Almost nobody writes about it. A chatbot might tell you 19, as of the 2020 census. Nineteen is real. It's the 2010 count. A real number on the wrong label.

And it says that in the same steady voice it uses for everything else. That's why AI sounds so sure. Confidence is a writing style, not proof.

One more thing. It isn't lying. Lying means knowing the truth and choosing to say something else. A chatbot doesn't know the truth. It knows what sounds right.

Get these videos delivered straight to your inbox

Every new video, as we release it. Unsubscribe anytime.

No spam. Unsubscribe anytime.

Try it tonight

Watching helps. Practicing sticks.

World 1 of RAILE Learn plays in the browser. Your kid catches their first AI lie in about 12 minutes, and you have something real to talk about at dinner.

World 1 is free. No signup, no card.

Prefer reading? Get the RAILE newsletter.