Software trained on a lot of text (and sometimes images or code) so it can continue a pattern. When you type, it predicts what would usually come next.
A model is not a mind. It does not “know” you, and it does not look around the world unless a tool is attached. It is very good at sounding fluent. Fluency is not the same as being right.
Think of it as a machine that has seen a huge pile of examples and learned the shapes in them. Ask for a thank-you note and it can finish a thank-you note. Ask for a fact it was never given, and it may still finish a sentence that looks like a fact.
Whatever you put in — the question, the paste, the instructions. The model continues from that.
Why this matters
- It can be useful without being a person.
- A calm voice is not a guarantee.
- You stay in charge of what leaves the building.