{"id":1935,"date":"2025-04-07T12:13:28","date_gmt":"2025-04-07T10:13:28","guid":{"rendered":"https:\/\/www.dimension-internet.com\/what-happens-when-you-talk-to-an-ai-a-journey-into-the-heart-of-a-prompt\/"},"modified":"2025-06-16T11:23:05","modified_gmt":"2025-06-16T09:23:05","slug":"what-happens-when-you-talk-to-an-ai-a-journey-into-the-heart-of-a-prompt","status":"publish","type":"post","link":"https:\/\/www.dimension-internet.com\/en\/what-happens-when-you-talk-to-an-ai-a-journey-into-the-heart-of-a-prompt\/","title":{"rendered":"What happens when you talk to an AI? A journey into the heart of a prompt"},"content":{"rendered":"\n

You’ve probably already spoken with an artificial intelligence (AI), whether via ChatGPT, a voice assistant, or an AI integrated into an app. You type a sentence, press “Send”… and a response pops up, often impressive, sometimes even mind-blowing. <\/p>\n

But what happens between the time you send your (prompt) message<\/strong> and the time the AI \u200b\u200bresponds<\/strong>?<\/p>\n

In this article, we’ll take you on a simplified and educational journey <\/strong>to the heart of an AI’s brain. We promise, you don’t need a math major to understand. Follow the guide. <\/p>\n

\ud83d\udce5 Step 1: You write a “prompt” (a request)<\/h2>\n

It all starts with you.<\/p>\n

\ud83d\udc49 Example:
“Give me a simple chocolate cake recipe.”<\/em><\/p>\n

This message is called a prompt<\/strong>. It’s an instruction, a question, or a request to the AI. But at this point, it’s just plain text<\/strong>. <\/p>\n

\n

\ud83e\uddec Step 2: The text is encoded into numbers (tokenization)<\/h2>\n

AIs don’t directly understand text like we do. They don’t see letters or words<\/strong>, but numbers<\/strong>. <\/p>\n

How does it work?<\/h3>\n

The AI \u200b\u200buses an encoder<\/strong> (called a tokenizer) to convert the text into \u201ctokens<\/strong>\u201d: units of meaning that can be words, syllables, or word fragments.<\/p>\n

\ud83d\udc49 Example: The prompt
“Give me a simple chocolate cake recipe.”<\/em>
can be transformed into something like:
[2049, 1098, 1325, 5011, 67, 7823, 1250, 32, 9003]<\/code><\/p>\n

Each number represents a piece of a word or a whole word.<\/p>\n

\ud83d\udccc Why numbers?<\/strong>
Because language models<\/strong> are mathematical machines. They only work with numerical vectors. Therefore, words must be translated into numbers before they can be processed. <\/p>\n

\n

\ud83e\udde0 Step 3: The AI \u200b\u200bpredicts the rest word by word<\/h2>\n

Once the prompt is encoded, we enter the heart of the system: the language model<\/strong>, an enormous mathematical machine based on artificial neurons.<\/p>\n

AI doesn’t know how to “answer” in the human sense of the word. It guesses word by word what should come next<\/strong>. This is statistical prediction. <\/p>\n

Concretely:<\/h3>\n
    \n
  1. She receives the prompt tokens.<\/li>\n
  2. It calculates (using millions, even billions of parameters<\/strong>) what the probability<\/strong> is of each possible word to follow.<\/li>\n
  3. She chooses the most likely (or one word among the most likely, depending on the model’s temperature).<\/li>\n
  4. She adds this word to the sequence, and starts again with this new sentence.<\/li>\n<\/ol>\n

    \ud83d\udc49 Example: It might start by generating:
    “Here’s a simple recipe for a delicious chocolate cake:”<\/em>
    Then:
    “Ingredients: 200g dark chocolate, 100g sugar…”<\/em><\/p>\n

    And it continues like this until the answer is deemed complete<\/strong>.<\/p>\n

    \n

    \ud83e\udde0 Zoom in on the model: a “sentence completion machine”<\/h2>\n

    The operation is based on a simple but powerful principle:<\/p>\n

    \n

    Complete the text by guessing what comes next.<\/strong><\/p>\n<\/blockquote>\n

    This mechanism is made possible by a very powerful type of neural network called Transformers<\/strong>.<\/p>\n

    Each artificial neuron<\/strong> doesn’t understand anything about language, but by stacking layers of calculations, AI learns to:<\/p>\n