The Hotel Hallway

People try to explain AI with talk of neural networks and billions of parameters, and your eyes glaze over. You don't need any of it. One simple picture explains how these tools really work—and why they behave the way they do.

Picture a hallway of doors

Imagine the AI standing at the start of a long hotel hallway. Every door is a word it could say next. It has read an enormous amount of text—nearly everything—so it has a strong sense of which door is the likely next step given everything said so far. It opens that door, steps through, and finds a new hallway with a new set of doors. It picks again. And again. One word at a time, all the way down.

That's it. That's the whole thing. An AI is a remarkably good guesser of the next likely word, doing it over and over until it has written a sentence, a paragraph, a page.

What the hallway explains

Why context changes everything. The doors available depend entirely on what's been said. Give it more context up front and you place it in a better hallway, where the likely next words are the ones you actually want.

Why it sometimes makes things up. It's always picking a plausible next word—not a verified true one. When it doesn't know, it doesn't stop; it confidently opens a plausible-looking door anyway. That's a hallucination, and it isn't lying. It's just doing what it always does.

Why it feels creative. Because there's more than one reasonable door at each step, you get fresh phrasings and unexpected combinations. The same prompt can take a slightly different path each time.

What it means for you

If the AI picks its next word based on everything so far, your job is simple: give it a great 'so far.' Set the scene, name the goal, show an example. You're not programming a machine—you're choosing which hallway it starts in.