The process of using a trained model to generate predictions or outputs from new inputs (as opposed to training).
Friendly Description: Inference is just "using" a trained AI model, asking it questions and getting answers. Training is the school years where the model learns; inference is the day-to-day job after graduation. Every time you type a prompt and get a response, the AI is doing inference.
Example: When you snap a photo of a flower and an app instantly tells you what species it is, that's inference. The model was trained months ago on millions of plant photos, and now it's putting that knowledge to work on your single new picture in a fraction of a second.