It does not think. It does not understand. It predicts the next word. Here is how that actually works.
Scroll to begin \u2193
Next-Word Prediction Game
Be the AI.
You are now a language model. For each sentence, pick the word that is most likely to come next. This is exactly what AI does \u2014 millions of times per second.
Training Data Demo
AI learns from statistical patterns, not meaning.
AI reads billions of sentences. It does not understand any of them. It just counts which words appear most often, and which words appear near each other. Try it \u2014 tap words in the corpus below, then reveal the frequency.
The cat sat on the mat. The dog sat on the floor. The cat ate the food. The dog chased the cat. The cat and the dog sat on the mat together.
Tap any word to highlight it. Then click the button below.
Notice: "the" appears far more than any other word. "cat" and "dog" and "sat" appear frequently too. This is all AI sees \u2014 numbers and frequencies. It never "understands" what a cat is. It just knows "cat" often appears near "sat" and "the". When AI generates text, it is reproducing these statistical relationships. That is the entire magic.
How AI Processes Language
Six steps from text to prediction.
Tap each card to reveal what happens behind the scenes. These six steps run every single time AI generates a word.
Scenario Quiz
Is this how AI actually works?
For each statement, decide if it describes how AI really works. Answer YES or NO \u2014 then see the explanation.
Key Takeaways
Six things to remember.
Tap each item to check it off. Come back to this list whenever AI confuses you.