Lesson 18 — What Are AI Parameters?

When people talk about AI models, you may hear phrases such as “a model with billions of parameters.”

But what exactly are parameters?

What Is an AI Parameter?

A parameter is an internal value that an AI model adjusts during training.

These values help the model learn patterns from its training data.

You can think of parameters as tiny settings inside the model that influence how it processes information and generates predictions.

During training, an AI model makes predictions and compares them with the expected results.

The model then adjusts its parameters to improve future predictions.

This process happens again and again across huge amounts of training data.

Modern AI models need to represent many different patterns and relationships.

Large models can therefore contain billions of parameters.

More parameters can give a model greater capacity to represent complex patterns, although more parameters do not automatically mean a better model.

Parameter count is often used as a simple way to describe the size of an AI model.

For example, you may see models described as having billions of parameters.

However, parameter count is only one way to compare AI models.

Imagine a huge control panel containing millions or billions of tiny adjustable settings.

Training is the process of adjusting those settings so the system becomes better at its task.

Parameters and tokens are very different things.

So far, we have mainly talked about AI models that work with text.

But modern AI can work with much more than words.

Next, we will explore multimodal AI models that can work with text, images, audio, and video.