Nearby lessons

16 of 30

Why Tokenization is Needed

Concept in One Line

Tokenization is needed because computers work with numbers, and breaking text into small pieces makes language easier for AI to process.

Why This Concept Is Important

  • Explains how GPT reads text.
  • Helps understand token limits and costs.
  • Important for interviews and prompt design.

Key Points to Remember

  • Computers work with numbers, not raw text.
  • Tokenization breaks text into small units.
  • Tokens can be words, parts of words, or symbols.
  • Tokenization makes text machine-readable.
  • Smaller pieces improve prediction accuracy.

Why Tokenization Is Needed

Text by itself is hard for a machine to process directly.

Tokens give the model small, manageable pieces.

Those pieces are easier to convert into numbers.

That is why tokenization is the first practical step before the model can think.

Simple Tokenization Example

Sentence: ChatGPT is powerful

Possible tokens:

  • Chat
  • GPT
  • is
  • powerful

Tokens depend on the tokenizer.

Simple Flow

Raw Text -> Tokenization -> Tokens -> Numbers -> Model

Why Smaller Pieces Help

  • New words appear every day.
  • Splitting words avoids confusion.
  • Smaller pieces improve prediction.
  • Tokenization helps handle unknown words better.

Interview / Exam Points

Q1: Why is tokenization needed?

Tokenization is needed because computers understand numbers, not raw text, so text must be broken into tokens first.

Q2: What does tokenization make text?

Tokenization makes text machine-readable.

Common Confusions to Avoid

  • Tokenization is not the same as token limits.
  • Tokenization is not word meaning.
  • Tokenization is a preprocessing step.
  • Tokenization happens before encoding and model processing.

One Line to Remember

Tokenization makes human text ready for machine processing.

Quick Self-Check

For Students

Computers understand raw text directly - Yes / No?

Tokenization makes text machine-readable - Yes / No?

(Correct answers: No, Yes)

🧠 Test Your Knowledge

3 Questions

Progress: 0 / 3
Keep Going!Types of Tokens in GPT