Mathematics Is the Language of the Universe

Look closely at nature and patterns show up everywhere. Planets travel in ellipses. Snowflakes grow six arms because of the way water molecules lock together as they freeze. The seeds in a sunflower head spiral in two directions, and the spiral counts usually land on neighbouring numbers from the Fibonacci sequence. None of this is decoration. These are structures, and structures can be described with equations.

That is why so many scientists treat mathematics as more than a school subject. It is the language that lets us state exactly what the universe is doing, then check whether we were right. Four centuries ago, Galileo put the idea plainly in his book The Assayer. He described the universe as a great book that we can only read if we learn the language it is written in, and its letters, he said, are triangles, circles and other geometric figures.

"[The universe] is written in the language of mathematics."

Galileo Galilei, The Assayer, 1623 (common English translation)

In 1960 the physicist Eugene Wigner asked the question that still hangs over this idea, in an essay titled The Unreasonable Effectiveness of Mathematics in the Natural Sciences. Why should equations invented on paper, often for purely abstract reasons, keep turning out to describe the real world, sometimes in places nobody imagined? Wigner had no tidy answer. He called the fit a gift that we neither understand nor deserve, and scientists have been working with that gift ever since.

Any language can be coded in math

The idea gets more interesting when it turns from nature to human language. A word looks like a string of letters, but a computer can represent it as a list of numbers, called a vector. The numbers are chosen so that words with similar meanings sit close together, and unrelated words sit far apart.

"hello"   ->  [ 0.21, -0.64, 0.38, ... ]
"hola"    ->  [ 0.19, -0.61, 0.40, ... ]
"stone"   ->  [-0.52,  0.07, 0.83, ... ]

Once words from different languages are placed in the same numerical space, translation becomes a geometry problem. Hello and hola land near each other because they are used in the same situations. A system that has learned this can move from one language to another without a dictionary that lists every pair. That is the sense in which every language can, in principle, be translated into every other.

"You shall know a word by the company it keeps."

John Rupert Firth, linguist, 1957

Firth was talking about meaning, not machines, but his line became the working principle of modern language AI. A model reads enormous amounts of text and learns each word's numbers from the words that appear around it. The results can be surprising. In a well-known 2013 experiment with a system called word2vec, researchers found that taking the vector for king, subtracting man, and adding woman lands very close to queen. Ideas such as gender, tense and geography turned out to be directions in the number space, so meaning itself had picked up a kind of arithmetic.

AI is built on math

Modern AI rests on three branches of mathematics. Linear algebra gives it the vectors and matrices that hold information. Probability lets it weigh possibilities and say how likely one answer is compared with another. Calculus lets it learn: during training, a model compares its output with the right answer, measures the error, and uses calculus to nudge millions or billions of numbers in the direction that shrinks that error. Repeat this enough times and the numbers begin to capture real structure in the data.

A useful picture is a hiker in thick fog trying to reach the bottom of a valley. She cannot see the whole landscape, but she can feel which way the ground slopes under her feet and take a small step downhill. Training does the same thing in a space with billions of dimensions, and the slope is exactly what calculus measures. Step by step, the error falls. Probability then does its part when the model is used. Even a chatbot reply is a probability calculation: at each step the model scores every possible next word and chooses from the likeliest.

AI does not store the world as pictures, sounds, or sentences. It stores it as numbers.

Images, actions and emotions too

Words are only the start. A digital image is already a grid of numbers, one set for each pixel's color. AI models learn patterns across those grids, such as edges, shapes, and eventually faces and objects.

Actions can be written the same way. A movement is a sequence of positions and angles over time: where the shoulder, elbow, and wrist are at each moment. Turn that into numbers and a model can recognize a wave, predict the next step of a walk, or help a robot copy a motion.

Emotion is subtler, and it deserves a careful claim. Researchers can measure signals linked to emotion, such as facial movements, tone of voice, and word choice, and psychologists often place feelings on scales like pleasant to unpleasant and calm to excited. That turns a feeling into a point in space that an AI system can store and compare. It does not mean the system feels anything. It means the signals of emotion can be represented, and anything that can be represented can be processed.

Why a shared code matters

The power of this approach is that everything ends up in the same form. Because pictures, speech, movement, and text all become numbers, one system can connect them: describe a photo in words, turn a spoken sentence into another language, or generate an image from a short description. The same mathematics sits underneath each task, which is why progress in one area so often carries over to the others.

There is a fair caution to add. A numerical description is a model of something, not the thing itself, and it can miss what matters or carry the biases of the data it learned from. Math is a precise language, but people still decide what to say with it.

One code underneath

The universe follows patterns, mathematics describes them, and AI has become very good at learning them. Language, images, actions, and emotions look like different worlds to us. To the math underneath, they are all structure, and structure can be written down, compared, and transformed. That shared code is the thread that ties nature, language, and intelligence together.

The ancient Pythagoreans are often summed up with the phrase all is number. They meant it as philosophy. Today, with AI turning a photograph, a sentence and a gesture into the same kind of object, it sounds closer to engineering.