How Does CIVIX Explains AI: Hallucinations and Deepfakes Work?

CIVIX explains AI: hallucinations and deepfakes work like a tricky game of pretend where the computer tries to fool you.

Imagine you're playing with a friend who loves pretending. They might say, "I saw a dragon at the park!" even though there were no dragons, that's like an AI hallucination. The AI makes up things it isn't sure about, just like your friend does when they want to make the game more fun.

Now imagine you and your friend are both pretending to be a robot in a video call. Your friend speaks, and their face appears on the screen, but it's not really them; it’s a clever trick that makes it look like they're there. That's like a deepfake. It uses special technology to make fake videos or images that look real.

How AI Pretends

  • The AI listens carefully, just like you do when someone tells a story.
  • It tries to guess what comes next, maybe even making up parts it doesn’t know for sure, that’s the hallucination part.
  • For deepfakes, the AI takes pictures or videos of someone and changes their face so it looks like they're saying something else, just like your friend pretending to be a robot.

It's all about clever tricks that make you believe something isn’t real, but it feels very real!

Take the quiz →

Examples

  1. A child imagines a dragon in the sky, but AI can make it look real to others.
  2. An AI creates a fake video of a teacher giving a speech they never gave.
  3. AI makes up facts that seem true, like saying the moon is made of cheese.

Ask a question

See also

Discussion

Recent activity

Categories: Science · AI· deepfakes· hallucinations