parameter

Think of a parameter as a knob or dial inside an AI model. When the model learns from data, it's essentially turning these knobs to find the perfect positions. Each knob's position (its value) helps the model recognize patterns, understand language, or identify objects. For instance, in a model trying to predict house prices, a parameter might represent how much weight to give to the number of bedrooms versus the square footage.

The total number of these parameters in an AI model can range from a few dozen to trillions. Larger models, like many popular language models, have billions or even trillions of parameters, which is why they are often called "large" language models (LLMs). The more parameters a model has, the more complex patterns it can potentially learn, but it also requires more data and computational power to train.