Lesson 2: Models, Training & Parameters

What You'll Learn

What a "model" really is, in plain language
What "training" means and why it's just guided practice
What "parameters" are and how adjusting them is like tuning an instrument

Three Words, Demystified

You will hear three words over and over in the world of AI: model, training, and parameters. They sound technical, but each one stands for a simple idea. By the end of this lesson, you'll be able to explain all three to a friend.

Let's take them one at a time.

A Model Is the Thing That Makes Predictions

A model is the part of an AI that actually makes a guess or a prediction. You give it an input, and it gives you an output.

Think of a model like a recipe. A recipe takes ingredients (the input) and turns them into a dish (the output). A model takes information in — say, a photo — and produces a result, like the label "cat."

Here are a few everyday examples of what a model does:

Give it an email, it predicts "spam" or "not spam."
Give it a photo, it predicts "dog," "cat," or "neither."
Give it the start of a sentence, it predicts the next likely word.

The model itself is not magic and not alive. It's more like a very elaborate set of instructions for turning inputs into outputs. The interesting question is: how does it get good at this? That's where training comes in.

Training Is Guided Practice

Training is the process of improving a model by letting it practice on examples and correcting its mistakes.

Picture someone learning to throw darts. Their first throw misses to the left. They notice, adjust their aim slightly to the right, and throw again. A little high this time. They adjust down. Throw after throw, the corrections get smaller and the darts land closer to the bull's-eye.

Training a model works the same way:

The model makes a guess on an example.
We compare its guess to the correct answer.
If it was wrong, the model is nudged so that next time it's a little closer.
Repeat this thousands or millions of times.

Crucially, the model is not "studying" the way a person consciously does. It's simply being adjusted, over and over, to make fewer mistakes on the examples it's shown. Training is just guided practice at enormous scale.

Parameters Are the Dials It Tunes

So what exactly gets adjusted during all that practice? The answer is the model's parameters.

Parameters are the internal settings, or "dials," inside a model that get tuned during training. Changing them changes the predictions the model makes.

Two analogies make this click:

Tuning a guitar. A guitar has tuning pegs. Turn them and the strings get tighter or looser until each note sounds right. You don't replace the guitar; you adjust its existing dials until it plays in tune. Training adjusts a model's parameters in the same spirit — turning many tiny dials until the predictions sound right.
Adjusting a recipe. Imagine perfecting a soup. A pinch more salt, a little less garlic, a touch more time on the stove. Each tweak is a small adjustment, and you taste-test after each one. Parameters are those adjustable amounts, and training is the repeated tasting-and-tweaking until the result is good.

One more thing worth knowing: real AI models can have a lot of these dials — sometimes millions or billions of them. No human turns them by hand. The training process adjusts them automatically. But the core idea stays simple: a parameter is just an internal setting that gets tuned to make the model better.

Putting It Together

Here's the whole picture in one breath: a model makes predictions, training is the practice of adjusting it against examples, and parameters are the internal dials that training tunes. A model is a recipe, training is the repeated taste-testing, and parameters are the pinches of salt you adjust along the way.

Key Takeaways

A model is the thing that makes predictions — like a recipe turning inputs into outputs.
Training is guided practice: the model guesses, gets corrected, and improves over many rounds.
Parameters are the internal dials that training tunes, like a guitar's pegs or a recipe's amounts.