Lesson 3: Data — The Fuel of AI

In the world of AI, there is a golden rule: "Garbage in, garbage out."

You can have the most advanced, sophisticated algorithms in the world, but if the data you feed them is flawed, incomplete, or biased, your AI will produce bad results. Data is the fuel that powers the engine of AI.

Types of Data

Data comes in many forms, and AI can learn from all of them:

Structured Data: Highly organized, like spreadsheets or databases. (e.g., a list of house prices and square footage).
Unstructured Data: Messy and complex. (e.g., images, text documents, audio recordings, or videos).
Semi-structured Data: A mix of both, like JSON or XML files.

Key Vocabulary

Before we start building models, we need to speak the language of data scientists:

Dataset: A collection of examples. (The entire spreadsheet).
Sample / Instance: One specific example in the dataset. (One row in the spreadsheet).
Feature: A measurable property we use as input. (The columns: e.g., bedrooms, sqft, location).
Label / Target: The answer we want the AI to predict. (The output column: e.g., the final house price).

Exercise: Identify Features and Labels

Imagine we are building an AI to predict whether a pet will be adopted quickly.

Animal Type (Feature)	Age in Months (Feature)	Health Status (Feature)	Days to Adopt (Label)
Dog	3	Healthy	4
Cat	24	Needs Meds	21
Dog	84	Healthy	15

Data Bias

If your training data is biased, your AI will be biased. If we train an AI to screen resumes using data from a company that historically only hired men, the AI will "learn" that being male is a desirable feature. Understanding our data isn't just a technical necessity—it's an ethical obligation. We'll explore this more in the next lesson.