The Human/AI Language • Chapter 3
AI Tokens and the Role of Randomness
Every conversation with AI is shaped by two key ideas: tokens and randomness. While they might seem technical at first, understanding these concepts unlocks a deeper connection with AI—helping you get more creative, precise, and effective results every time you interact.
Estimated time.
Part 1
Introducing AI Tokens
When you communicate with an AI system, every word you write must be processed in a way the AI can understand. This is where tokens come in—they're the building blocks that AI uses to make sense of human language.
Example
When you type:
"Let's make something amazing today!"
The AI breaks this phrase into meaningful pieces:
"Let's" | "make" | "some" | "thing" | "amazing" | "today" | "!"
These pieces are tokens. Some are whole words, while others might be parts of words or even punctuation marks.
Why Tokens Matter
Tokens represent a careful balance in language processing. If an AI system tried to understand language one character at a time, it would be like trying to understand a story letter by letter—inefficient and requiring immense computing power.
On the other hand, using only whole words might seem more straightforward, but it would limit the AI’s ability to handle:
Variations like plurals and verb tenses.
Complex languages where meaning depends on context.
Completely new words it hasn’t encountered before.
By breaking language into tokens, AI systems gain flexibility to better understand and generate natural responses across diverse contexts.
Part 2
Context Windows
Think of a conversation as a flowing river, with ideas and words streaming by. The AI sits on the riverbank, able to see only a small stretch of that flow at a time—this is what we call the context window.
The context window is the limit of what an AI can "see" at once. It’s like looking through a camera lens: the scene you capture is vivid and detailed, but everything outside the frame is invisible. As new text comes into view, older text flows out of focus.
The size of a model’s context window varies, but even large models are limited. If you’re working on something complex, try breaking it down into manageable chunks or reminding the AI of important details along the way.
Why does this matter?
When you’re chatting with an AI, the context window determines how much it can remember about what’s been said so far. If the conversation becomes too long, earlier details might drift out of view—not because the AI is forgetful, but because they’ve literally moved outside its frame of reference.
Example
Imagine explaining a complex problem to a chatbot. Halfway through, you realise it’s lost track of the original issue. That’s the context window at work—the AI’s vision can’t hold onto everything at once.
Part 3
Randomness
Let’s imagine two musicians performing the same song. One plays each note exactly as written, without variation. The other improvises a little—changing the tempo, adding a flourish, bringing their unique touch. Which performance feels more alive?
AI works much the same way. Without a bit of randomness, its responses would feel mechanical and repetitive, like the first musician. With too much randomness, it might sound chaotic, like a musician playing off-key. But in the sweet spot, the AI becomes a creative collaborator—consistent enough to make sense, yet unpredictable enough to feel dynamic.
Randomness gives the AI a touch of spontaneity—making each interaction feel more human. But it’s not perfect. That’s why your role isn’t just to ask questions but to guide the conversation, steering the AI toward meaningful, creative, and reliable outcomes.
Why does randomness matter?
Every time an AI responds, it chooses words based on probabilities learned from its training data. Randomness introduces variation, allowing the AI to generate responses that are fresh and engaging.
Example
Prompt: "Why do we need sleep?"
Low randomness: "Sleep is essential for physical and mental restoration."
Higher randomness: "Sleep is our body’s reset button, recharging energy and repairing our minds and muscles."
Reflection
Too little randomness, and the AI becomes rigid; too much, and it loses coherence.
In your collaborations with AI, how do you strike the right balance between guiding the process and allowing for unexpected brilliance?
Next • Chapter 4
Ethics, Errors and Explainability
We are standing on the edge of a revolution. For the first time in history, machines can hold meaningful conversations with us, help us solve problems, and even inspire new ideas. But with great power comes great responsibility.
Every time you interact with an AI system, you’re shaping the future. The way you use these tools, the information you share, and the tasks you entrust to them don’t just affect your immediate outcomes—they influence how this technology evolves for society at large. Let’s take a look at how to navigate this new frontier responsibly.