Confidently Wrong
In our last lesson we learned that AI predicts words — it does not truly understand them. That leads to a very important problem: sometimes AI predicts the wrong words with great confidence. Confidence means sounding sure. Imagine someone giving you directions in a very sure, calm voice — and they are completely wrong. You might follow those directions all the way to the wrong place! Let's learn why AI does this and what to do about it.
Sounding Sure Is Not the Same as Being Right
When AI writes an answer, it is always picking the words that seem most likely. It has no way to feel uncertain the way you do. You know that feeling — when you are not sure of an answer on a test, your stomach feels a little nervous. You slow down and think harder. AI does not have that feeling. Whether its answer is right or wrong, it writes in the same confident, smooth style. The words flow just as easily when AI is making a mistake as when it is correct.
AI can be completely wrong and still sound 100% sure. Sounding confident is not proof that something is true.
Here is an example. Imagine you ask AI: 'What is the tallest mountain on Mars?' AI might give you a very detailed, confident-sounding answer. It might even give you a measurement in exact meters. But if that measurement is wrong, AI will say it just as confidently as if it were right. This is called an error — or sometimes a hallucination (a word we will learn more about in Lesson 7). For now, remember: always check important facts yourself, or with a trusted grown-up.
Flashcards — click each card to reveal the answer
This is not a reason to be afraid of AI. It is a reason to be a smart, careful user of AI — just like you are careful when you cross the street. You look both ways even when you feel pretty sure it is safe. The same careful thinking applies to AI answers.
If AI tells you something important — a medicine, a safety rule, a historical date, or a person's name — always double-check it. A confident tone does not mean a correct answer.
AI tells you a fact in a very calm, sure-sounding way. What does that mean?
Which is the BEST way to handle an important fact from AI?
The Confidence Experiment
- Ask a family member or friend to play this game with you.
- You are going to take turns saying two facts — one true and one made-up — in the SAME confident, sure voice.
- For example: 'Elephants are the largest land animals. Elephants can fly.' Say both in the same calm, certain way.
- After each round, guess which fact was true.
- Was it easy to tell just from how sure the person sounded? Usually not!
- This is exactly why a confident voice (or confident-sounding AI writing) does not prove something is true. How something is said and whether it is true are two separate things.