What exactly is a large language model and how does it 'know' things?

A large language model, or LLM, is a type of AI trained on a massive amount of text to predict the next word in a sequence, and it 'knows' things by building a complex map of how words and concepts relate to each other, not by storing facts like a database. Think of it less like a library and more like a giant, messy web of connections. It doesn't 'know' that Paris is the capital of France in the way you do. Instead, it has seen the words 'Paris' and 'capital' and 'France' appear together so many times in its training data that the path between them is extremely strong. When you ask it a question, it's not retrieving a stored answer. It's starting with your words and then, token by token, predicting what word should come next to form a coherent and relevant response. A good concrete example is asking it to finish the sentence 'The Eiffel Tower is in...' The model's internal math says there's a very high probability the next word is 'Paris.' That's it. There's no thought, just statistical prediction. This is also why they sometimes 'hallucinate' or make things up—the most probable next word isn't always the true one. The useful insight here is that you should treat an LLM as a reasoning and writing engine, not a fact engine. Always verify any important data it gives you against a real source. **Related**: Why do AI chatbots sometimes make up false information? | How is an LLM different from a search engine?