How AI Uses Maths: The Simple Explanation Every Kid Deserves

When people say "AI is just maths," they are right — but that is not scary. The maths inside AI is built from ideas most students start learning at age 11-12. Here is the honest explanation.

The Core Idea: Weights and Predictions

Imagine you are predicting whether tomorrow will be rainy based on two things: how cloudy it is today (0-10 scale) and whether there was rain last week (yes/no).

You might reason: "Clouds matter a lot — I'll give that 70% of my prediction. Last week's rain matters less — 30%." Those percentages are weights. AI neural networks do the same thing, but with millions of inputs and millions of weights, all tuned automatically from data.

Probability: The Language of Uncertainty

AI does not say "the answer is X." It says "the probability of X is 94%, Y is 4%, Z is 2%." Everything an AI outputs is a probability distribution over possible answers.

When you ask ChatGPT a question, it is doing probability at every single word: "given the words I have generated so far, what word is most likely to come next?" It does this calculation thousands of times to produce a single sentence.

Linear Algebra: Maths That Handles Many Numbers at Once

AI processes thousands of numbers simultaneously. Instead of doing calculations one number at a time, it uses matrices (grids of numbers) and vectors (lists of numbers). Matrix multiplication lets AI perform millions of calculations in a single step on a GPU (graphics card).

Your image, your text, your data — everything becomes vectors of numbers before AI can work with it. An image might become a vector of 786,432 numbers (1024×768 pixels × 3 colour channels).

Calculus: How AI Learns

When AI makes a wrong prediction during training, it needs to improve. Calculus (specifically gradient descent) is the technique it uses to figure out which direction to adjust each weight to reduce the error.

Think of it like being in a foggy landscape and trying to walk downhill to reach a valley. You can't see the whole landscape, but you can feel which direction is downhill right where you're standing. That local slope is a gradient — and AI uses it to improve step by step.

Statistics: Understanding Data Quality

Before any AI is trained, someone has to understand the data. Are there biases in it? Is it representative? How is it distributed? This requires statistics — mean, median, standard deviation, correlation, outlier detection.

Bad data + good maths = bad AI. Statistics is what catches data problems before they become AI problems.

How to Start Learning AI Maths

• Age 10-12: Khan Academy — Probability and Statistics unit (free, excellent)
• Age 12-14: 3Blue1Brown's "Essence of Linear Algebra" series on YouTube (free, visual, brilliant)
• Age 14+: Fast.ai (free deep learning course that teaches practice before theory)
• Any age: Build things. Curiosity about the maths follows from making AI projects work.