What tokens and context windows mean in AI for Everyone - Mini Project: Build & Apply

Understanding Tokens and Context Windows

📖 Scenario: You are learning how AI language models understand and process text. To do this, you need to understand two key ideas: tokens and context windows.Imagine you are reading a book, but instead of reading whole sentences at once, you read small pieces called tokens. Also, you can only remember a certain number of these tokens at a time, which is called the context window.

🎯 Goal: Build a simple explanation using examples to show what tokens are and how a context window limits what an AI can remember at once.

📋 What You'll Learn

Create a list of example tokens from a sentence

Define a variable for the context window size

Select tokens that fit inside the context window

Explain the final selection as the AI's current memory

💡 Why This Matters

🌍 Real World

Understanding tokens and context windows helps people grasp how AI models read and remember text, which is important for using AI tools effectively.

💼 Career

This knowledge is useful for anyone working with AI, such as developers, data scientists, or content creators, to better design and interact with AI systems.

1

Create a list of tokens from a sentence

Create a list called tokens with these exact string elements: 'I', 'love', 'learning', 'about', 'AI'.

AI for Everyone

# Create the list of tokens
# Your code here

Hint

Think of tokens as small pieces of a sentence, like words.

2

Define the context window size

Create a variable called context_window and set it to the number 3.

AI for Everyone

tokens = ['I', 'love', 'learning', 'about', 'AI']
# Define the context window size
# Your code here

Hint

The context window is how many tokens the AI can remember at once.

3

Select tokens within the context window

Create a new list called current_context that contains the first context_window tokens from the tokens list.

AI for Everyone

tokens = ['I', 'love', 'learning', 'about', 'AI']
context_window = 3
# Select tokens within the context window
# Your code here

Hint

Use slicing to get the first few tokens that fit in the context window.

4

Explain the AI's current memory

Add a comment explaining that current_context represents the tokens the AI can remember at once within its context window.

AI for Everyone

tokens = ['I', 'love', 'learning', 'about', 'AI']
context_window = 3
current_context = tokens[:context_window]
# Explain the AI's current memory
# Your code here

Hint

Write a clear comment about what current_context means for AI memory.

Practice

1. What does the term token mean in the context of AI language models?

easy

A. A type of AI algorithm

B. The total number of words in a document

C. The maximum length of a sentence

D. A small piece of text that AI reads one by one

Solution

1. Step 1: Understand the definition of a token

   A token is a small piece of text, like a word or part of a word, that AI processes individually.

2. Step 2: Compare options with the definition

   Only a small piece of text that AI reads one by one correctly describes a token as a small piece of text AI reads one by one.

3. Final Answer:

   A small piece of text that AI reads one by one -> Option D

4. Quick Check:

   Token = small text piece [OK]

Hint: Tokens are tiny text chunks AI reads step-by-step [OK]

Common Mistakes:

• Confusing tokens with full sentences
• Thinking tokens are AI algorithms
• Assuming tokens mean total words

2. Which of the following best describes a context window in AI language models?

easy

A. The size of the AI's training dataset

B. The maximum number of tokens AI can consider at once

C. The number of AI models running simultaneously

D. The number of layers in the AI model

Solution

1. Step 1: Define context window

   A context window is the limit on how many tokens the AI can look at together when generating or understanding text.

2. Step 2: Match definition to options

   The maximum number of tokens AI can consider at once correctly states it is the maximum number of tokens AI can consider at once.

3. Final Answer:

   The maximum number of tokens AI can consider at once -> Option B

4. Quick Check:

   Context window = max tokens at once [OK]

Hint: Context window limits tokens AI sees simultaneously [OK]

Common Mistakes:

• Mixing context window with training data size
• Confusing it with AI model layers
• Thinking it means number of AI models

3. If an AI model has a context window of 100 tokens and you input a text of 150 tokens, what happens?

medium

A. The AI ignores the first 50 tokens and processes the last 100 tokens

B. The AI processes only the first 100 tokens and ignores the rest

C. The AI processes all 150 tokens at once

D. The AI splits the text into 3 parts and processes them separately

Solution

1. Step 1: Understand context window limit

   The context window limits how many tokens the AI can consider at once, here 100 tokens.

2. Step 2: Apply limit to input length

   Since input is 150 tokens, AI can only process the last 100 tokens at once; the first 50 tokens are ignored or truncated.

3. Final Answer:

   The AI ignores the first 50 tokens and processes the last 100 tokens -> Option A

4. Quick Check:

   Context window limits tokens processed [OK]

Hint: AI processes only tokens within its context window [OK]

Common Mistakes:

• Assuming AI processes all tokens regardless of window
• Thinking AI processes first tokens only
• Believing AI splits text automatically

4. You wrote a prompt with 120 tokens, but your AI model has a context window of 100 tokens. What is a likely problem?

medium

A. The AI will produce an error or truncate the prompt

B. The AI will generate output ignoring the token limit

C. The AI will automatically increase its context window

D. The AI will repeat the prompt twice

Solution

1. Step 1: Recognize context window limits

   The AI cannot process more tokens than its context window allows, here 100 tokens.

2. Step 2: Predict AI behavior with excess tokens

   When input exceeds the limit, the AI either truncates the prompt or returns an error instead of processing all tokens.

3. Final Answer:

   The AI will produce an error or truncate the prompt -> Option A

4. Quick Check:

   Exceeding context window causes truncation or error [OK]

Hint: Too many tokens cause truncation or errors [OK]

Common Mistakes:

• Thinking AI ignores token limits
• Believing AI expands context window automatically
• Assuming AI repeats input when too long

5. You want to design an AI chatbot that remembers the last 500 tokens of conversation. The model has a context window of 300 tokens. What is a good approach?

hard

A. Split the conversation into 300-token chunks and send all at once

B. Send only the last 300 tokens each time to the AI

C. Increase the context window to 500 tokens or more

D. Ignore the context window and send all 500 tokens anyway

Solution

1. Step 1: Understand context window limits and memory needs

   The AI can only consider up to its context window tokens at once, here 300 tokens, but you want to remember 500 tokens.

2. Step 2: Find solution to match memory with context window

   To remember 500 tokens, the context window must be at least 500 tokens; increasing it is the correct approach.

3. Final Answer:

   Increase the context window to 500 tokens or more -> Option C

4. Quick Check:

   Context window must fit tokens to remember [OK]

Hint: Context window must be large enough to hold all tokens [OK]

Common Mistakes:

• Sending more tokens than context window allows
• Trying to send multiple chunks at once
• Ignoring context window limits