← Concept IndexDEFINITION WHY IT MATTERS COMMONLY CONFUSED WITH
Training vs. use
Also called: pretraining vs. inference, learning vs. answering
Training is the one-time, expensive process of tuning a model's parameters on large data. Use (inference) is what happens every time you send a prompt afterwards — the parameters don't change.
Your chat does not teach the model. Whether your input is later used to train a future version is a product and policy question, not something the running model does — and it's the question to actually check before pasting sensitive data.
The idea that a model 'learns from our conversation'. Within a session it only reads the context window; it doesn't update itself.