When you chat with an AI like ChatGPT, you’re talking to a “large language model.” It works in a surprisingly simple way. Engineers feed it huge piles of human writing - books, websites, articles, code - and train it to play a guessing game. Given a few words, can you guess the most likely next word? Then the word after that? And the next?
After reading billions of pages, the model gets very good at this guessing. It can answer questions, write stories, help with code and even hold a conversation. But it doesn’t really “know” things the way you do. It is incredibly good at finding patterns in language and stringing words together in ways that sound right.
That’s also why it sometimes makes things up. If the model has seen many sentences about dinosaurs, it can generate a new sentence that sounds confident - even if the facts are wrong. ChatGPT, released in November 2022, reached 100 million users in just two months. It was the fastest-growing app in history.
