When AI Gets Things Wrong: It Happens More Than You Think

AI sometimes makes things up. It does not know it is wrong. It says wrong stuff in the same confident voice it uses for right stuff. So you have to learn to check.

Some examples

• You ask 'When was George Washington born?' and AI gives you the right year.
• You ask 'When was George Washington's pet dog born?' and AI might just MAKE UP a year.
• AI can mix up dates, names, and facts.
• AI can invent fake books and fake quotes that sound real.

Try it!

Ask AI for a fun fact about your favorite animal. Then look up the same fact in a kid encyclopedia or with a librarian. Was AI right? Was anything off?

AI is impressive, but it makes funny mistakes too. Picture-AIs once drew people with 6 fingers all the time. Chat-AIs sometimes invent fake books or fake countries. These goofs are reminders: AI is not magic.

Some examples

• Picture-AI used to draw hands with 6 or 7 fingers.
• Chat-AI invented fake court cases that fooled a real lawyer (oops!).
• AI has confused a chihuahua photo for a blueberry muffin.
• Some AIs answer 'how many R's in strawberry?' as 2 (it's 3!).

Try it!

Ask AI to draw a picture of 'two hands shaking.' Count the fingers. Notice anything weird?

AI doesn't 'know' things the way you do — it predicts what sounds right. So sometimes it confidently makes stuff up. The fancy word for this is 'hallucination.' That's why we always double-check important answers, especially facts and math.

Some examples

• AI confidently invents a fake book title.
• AI gets math wrong because it guessed instead of really computing.
• AI mixes up dates because two stories sound similar.
• AI says something true sometimes — and totally wrong other times.

Try it!

Ask AI an easy math question and a tricky one. See if it gets both right. Use a calculator to check.