In Lesson 001, we treated Artificial Intelligence (AI) like an experienced store assistant who has learned patterns from many customers.
Now zoom out. Imagine a huge department store. One person is brilliant with heels, another knows handbags, another can match fragrance notes, and another is excellent at finding lipstick shades for different undertones.
They all work in the same store, but you would not ask the shoe specialist which berry lipstick suits you best.
That is a useful way to understand an AI model.
AI is the big category; a model is the specialist doing the work
“AI” is the broad umbrella. A model is a trained system designed to perform a particular kind of task.
A model might be good at:
- recognizing what appears in an image,
- classifying text,
- recommending products,
- generating the next part of a piece of text,
- turning text into an image.
So when you hear “this AI is amazing,” a more precise question is: Which model, and what was it designed to do?
Why can two models behave so differently?
Imagine two handbag specialists.
One mainly serves commuters and knows weight, capacity and work-friendly shapes. Another works with collectors and knows rare leathers, limited colors and resale value.
They both understand handbags, yet the same question may lead to very different recommendations.
Models are similar. Their behavior depends on what data they saw, how they were trained and what goals they were optimized for.
Is ChatGPT a model?
It is helpful to think of ChatGPT as the product or service you interact with. Behind that service, one or more AI models may handle different kinds of work.
It is a little like using a department store’s VIP desk: you see one entrance, but different requests may be sent to different specialists behind the scenes.
Calling ChatGPT “AI” in everyday conversation is fine. But when you want to understand the technology, separating the product from the model makes many confusing discussions much clearer.
A model is not just a giant answer database
A model is not simply a huge table of “question → correct answer.”
It is closer to a specialist who has developed judgment after seeing many examples. She does not need to memorize every customer conversation to recognize that certain needs often go with certain choices.
Likewise, a model learns patterns in data rather than storing every source as a neat searchable page.
One thing to remember today
AI is the broad technology category. A model is the trained system that actually performs a particular task.
Next, we will ask the obvious question: how does a model become good at anything in the first place?