How does an AI actually work?

We said AI predicts the next word. Let's see how it learns to do that.

Streams of text and code scrolling on a dark screen — Before you ever chat with it, an AI reads billions of sentences and learns which words tend to follow others.

Step 1: It reads a huge amount of text

Before you ever talk to it, the AI is trained. It reads billions of sentences — books, articles, websites. Nobody hands it a rulebook of grammar. Instead, it slowly notices patterns: which words go together, how questions get answered, how a story flows.

Step 2: It predicts the most likely continuation

Once trained, the AI plays a guessing game every time you use it. You give it some words, and it asks itself: "Given everything I have seen, which word most likely comes next?" It picks a word, adds it, and repeats — word by word — until the answer is complete.

Try it yourself below. Start a sentence and watch how the model suggests what might come next:

[[demo:next-word]]

Notice there is no single "correct" next word — only words that are more or less likely. That is why the same question can give slightly different answers each time.

Because it works by likelihood, AI can sound confident even when it is guessing. Sounding fluent is not the same as being right — something we will come back to.