AI From Zero - Lesson 3: How AI Learns
The Magic of Machine Learning (ML)
"Tell me and I forget. Teach me and I remember. Involve me and I learn." - Benjamin Franklin
In our last post, we demystified AI and understood the difference between the science-fiction dream of General AI and the real-world power of Narrow AI. Now, let's peek behind the curtain and understand how AI systems learn to perform those incredible tasks: through Machine Learning (ML).
Machine Learning: AI's Learning Engine
Machine Learning is a subset of AI that gives systems the ability to learn from data and identify patterns without being explicitly programmed for every single task. Instead of a human writing millions of specific rules (e.g., "if image has pointy ears AND whiskers AND tail, then it's a cat"), ML models are fed vast amounts of data and learn the patterns themselves.
Think of it like this:
Traditional Programming: You give the computer data and rules, and it gives you an answer. (e.g., Rule: 2+2, Data: numbers, Answer: 4)
Machine Learning: You give the computer data and answers, and it figures out the rules. (e.g., Data: millions of cat/dog pictures, Answers: "cat" or "dog" labels, ML figures out the "rules" to distinguish them).
This ability to "learn from examples" is what makes AI so powerful and adaptable.
The More Data, the Smarter the AI (Usually!)
For most Machine Learning models, the more data they are trained on, the better they become at recognizing patterns and making accurate predictions or generating relevant outputs. This is why companies collect so much data!
Example: An AI trained to recognize human faces would need to see millions of faces (from diverse backgrounds, ages, lighting conditions) to become highly accurate. If it only saw faces of one type, it would struggle with others, leading to bias (a topic we'll cover later!).
Different Ways AI Learns
There are several types of Machine Learning, but two common ones you might hear about are:
Supervised Learning: This is the most common type. The AI learns from "labeled" data. For example, you show it thousands of pictures of cats and label each one "cat." The AI learns to associate the visual patterns with the "cat" label.
Unsupervised Learning: The AI looks for patterns in "unlabeled" data on its own. For example, it might group customer data into different segments without being told what those segments are beforehand.
The "Brains" of Generative AI: Large Language Models (LLMs)
When you interact with AI chatbots like ChatGPT or Gemini, you're using a specific type of Machine Learning model called a Large Language Model (LLM). LLMs are trained on enormous amounts of text data (books, articles, websites, etc) to understand and generate human-like text. They learn the nuances of language, grammar, facts, and even different writing styles.
This is the "magic" that allows them to answer questions, write essays, summarize documents, and even generate creative content.
Practice Exercise
Imagine you want to train an AI to recommend songs. Describe what kind of "data" you would need to feed it using Supervised Learning. What would be the "labels" in this data?
Fun Fact
The amount of data created and consumed globally is doubling every few years. This ever-growing sea of information is precisely what fuels the advancement and capabilities of modern Machine Learning models like ChatGPT!
Learning Reinforcement Questions
What is Machine Learning (ML) in simple terms?
A type of computer hardware.
A way for computers to learn from data without explicit programming.
A new programming language.
Another name for Artificial General Intelligence.
How is Machine Learning different from traditional programming?
Why is "data" so important for Machine Learning models?
What does "Supervised Learning" mean?
What does "LLM" stand for, and what do they specialize in?
Once you've given it a shot, you can find the <guidelines to answering these questions here> to check your understanding.
Next up
In our next lesson, Lesson 4: Decoding AI's Capabilities: What Can AI Really Do?, we will look at the range of tasks AI is capable of performing today as we get closer to a more practical encounter with the technology.
Licensing, Attribution and Commercial use
© 2025 Nacha – AI Activation Hub, a division of Asset Thinking Ltd. All rights reserved.
For commercial licensing, partnerships, adaptations, integrations, usage within an organization or consulting inquiries, please contact the author via email: zack@nacha.life
Done..