Binary System

We normally write numbers using ten digits, from 0 to 9. The binary system gets by with just two digits: 0 and 1. Even so, any number can be represented this way — you just need more digits. The number 5 is written in binary as 101, and the number 100 is written as 1100100. Computers use exactly this notation because their components can only cleanly distinguish between two states: current flows or it doesn’t. A single such digit is called a bit.

Why computers have to make do with two digits

A computer chip consists of billions of tiny switches. Each switch is either open or closed. In theory, one could try to distinguish ten different voltage levels and calculate directly in the decimal system. In practice, this fails because of interference: voltages fluctuate, components heat up, signals blur together. Two clearly separated states, on the other hand, remain unambiguous even under strong noise.

This robustness is the entire reason for the choice. It explains why binary shows up everywhere — from hard drives to working memory to network cables. A DVD works the same way: tiny pits and smooth areas stand for 0 and 1. And because everything ultimately comes down to bits, the same chip can play music, calculate taxes, and run an AI model.

A common misconception is that binary code is a secret language in which programmers write. That’s not true. People write programs in readable text, for example in Python. Only a translation program turns this into zeros and ones that the processor can execute directly.

From bits to text, images, and model weights

Each binary digit has a fixed value, similar to the ones, tens, and hundreds places in the decimal system. Only here, the value doubles from right to left: 1, 2, 4, 8, 16, and so on. The sequence 1011 therefore means 8 plus 0 plus 2 plus 1, totaling 11. With eight bits, one byte, 256 different values can be represented.

For numbers to carry meaning, a convention is needed. For text, a table defines which number corresponds to which character; the value 65, for instance, stands for the letter A. For a photo, three numbers each describe the red, green, and blue components of a pixel. The same sequence of bits can be a letter, a brightness level, or a command, depending on the convention applied.

AI models, too, ultimately consist of numbers, known as weights. How many bits are allotted per weight is a trade-off. Fewer bits means coarser but faster and more memory-efficient. This is exactly what the term quantization refers to, which you often come across in reports about efficient AI models.

Where you notice zeros and ones in everyday life

The binary system is most visible in storage sizes. A kilobyte has traditionally been 1024 bytes, not 1000 — because 1024 is a power of two. That’s why a hard drive sold as 500 gigabytes often shows only about 465 gigabytes on the computer. This also explains why many technical values are 8, 16, 32, 64, or 256.

In news from the tech industry, the term usually appears indirectly. When there’s talk of 8-bit quantization, 64-bit processors, or bandwidths in gigabits per second, it’s always about binary digits. The number of colors a screen can display follows the same pattern: 24 bits per pixel yield about 16.7 million colors.

The binary system should be distinguished from the hexadecimal system, which works with sixteen digits and appears in color codes like #FF0000. It’s not an opposite, but rather a shorter notation for the same bits. Four binary digits correspond exactly to one hexadecimal digit. Programmers use this because long chains of zeros and ones are hard for humans to read.