You will be able to explain what a model is in plain terms: something that turns inputs into outputs using learned settings.
Auntie Lim runs a chicken rice stall in a Toa Payoh hawker centre. Every evening she decides how many chickens to buy for the next day. She has never written a formula, but she has one in her head. Weekdays are busier than Sundays because of the office crowd. Rain keeps people at home and sends her numbers down. The week before payday is quieter. She weighs all of that, and out comes a number: buy twenty chickens.
That habit of turning a few facts into a decision is a fair picture of what a model does. People who build AI use the word "model" constantly, and it sounds more mysterious than it is. This lesson takes the mystery out.
A model is something that takes an input and produces an output, following settings it learned from examples. In Auntie Lim's head, the inputs are the day of the week, the weather forecast and how close it is to payday, and the output is a number of chickens.
Machine learning models work the same way, with different inputs and outputs. A spam filter takes an email and outputs a label, spam or not spam. Phone face recognition takes a camera image and outputs yes, this is the owner, or no. A language model, the kind behind ChatGPT, Claude and Gemini, takes the text so far and outputs its prediction of the next word or piece of a word. Module 3 looks at that last one in detail.
Mathematicians call anything that turns inputs into outputs a function. You do not need the maths. Just hold on to the shape: something goes in, the model does a calculation, something comes out.
What decides the output? Imagine Auntie Lim's reasoning as a machine with dials on the front. One dial sets how much a weekday adds. Another sets how much rain takes away. A third sets the payday effect. Turn the rain dial up and the machine buys fewer chickens on wet days. Set every dial well and the machine buys about the right number most days.
In a real model these dials are called parameters. Each one is just a number, and the calculation combines the input with all of those numbers to produce the output. A small model, like one predicting chicken sales, might have a handful. The models that recognise faces or write text have millions, and the large language models behind today's assistants have billions. Each dial does very little on its own. The behaviour comes from all of them working together.
The useful idea is that the model's skill lives entirely in those numbers. Two models with the same design but different dial settings will behave differently, in the same way two stallholders with the same stall but different instincts will buy different amounts of chicken.
With three dials, a person could set them by trial and error. With billions, nobody could. So the dials are set by training: the model is shown many examples, and an automatic process adjusts the numbers bit by bit until the outputs come out right more often. Lesson 2.2, Learning means guessing, measuring the miss and adjusting, walks through how that works.
This has a strange consequence. The people who built a large model cannot point to a dial and say what it does. The settings were found by the training process, and no person chose them. Researchers can study a model from the outside, and some work on looking inside it, but nobody can fully explain why it produced one particular output. When a company says it does not know exactly why its assistant gave a strange answer, that is usually the plain truth.
Once training stops, the dials are fixed. The trained model is a file holding all of those numbers, plus the design that says how to combine them. When you use an assistant, that file sits on a company's computers and runs your input through the calculation.
It does not learn as you use it. Ask it the same thing tomorrow and it starts from the same settings. If the world changes, say a new MRT line opens or a rule at your workplace is updated, the model knows nothing about it until someone trains it again, or feeds the new information into the conversation, which you will meet in module 5. Many people assume an assistant is quietly learning from every chat. The underlying model is not. Some products save notes about you and show them to the model next time, but the dials stay where they were.
Back to Auntie Lim. Her instincts did not come from a manual. They came from years of evenings: buy too many, throw some away, adjust; buy too few, sell out by one o'clock, adjust. A model is built the same way, just faster and with far more examples. Being able to say that clearly to someone who has never thought about it is a good test of whether you understand it, and that is what the activity asks you to do.
Write a two-sentence explanation of what a model is that a parent or friend with no tech background would understand, then read it to them.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).