What is AI & Machine Learning?
The big picture — types of AI, how machines learn
You used AI before breakfast today. Your phone's autocorrect fixed three typos while you texted a friend. Netflix served up a show it somehow knew you'd love. And your email quietly moved a "Congratulations, you won!!!" message into spam before you ever saw it. All of that? Machine learning at work.
We tend to think of AI as something futuristic — robots, self-driving cars, sentient computers in movies. But the truth is, you're already surrounded by it. It's in your search results, your photo gallery's face recognition, your bank's fraud detection. It's just... quiet about it.
But what's actually happening under the hood? How does a machine "learn" anything? Let's break it down. No math, no code, just the core ideas you need to understand everything else in this course.
What's the difference between AI and ML?
People use "AI" and "machine learning" like they're the same thing. They're not — but they're related. Think of it this way: AI is the dream, ML is the method.
Artificial Intelligence has been an idea since the 1950s. It's the broad goal of building machines that can do things we'd consider "intelligent" — recognizing faces, understanding language, making decisions. For decades, people tried to achieve this by writing rules by hand. "If the email contains 'free money,' mark it as spam." That works... until spammers get creative.
Machine Learning flips the script. Instead of a human writing rules, you give the machine examples and let it figure out the rules on its own. Here's what that flip looks like:
Traditional Programming
See the difference? In traditional programming, you tell the computer exactly what to do. In machine learning, you show it thousands of examples and say "figure it out." The machine discovers patterns you might never have thought to look for. That's the magic — and it's why ML has gotten so powerful as data has gotten so abundant.
Three flavors of machine learning
Not all machine learning works the same way. There are three big approaches, and each one tackles a different kind of problem. Click through to see how they compare.
Step 1: Supervised Learning
This is the most common type. You give the model labeled examples — inputs paired with the correct answers — and it learns to predict answers for new inputs.
Use cases: spam detection, image classification, medical diagnosis, price prediction, language translation.
Think of it like a student learning from a textbook with answer keys. Eventually, they can answer new questions on their own.
In practice, supervised learning is by far the most common. Most of the AI you interact with daily — spam filters, voice assistants, translation apps — is supervised learning under the hood. But the other two are growing fast, especially reinforcement learning, which plays a huge role in making modern language models actually helpful (we'll get to that in the RLHF lesson).
Now try it yourself
Theory is nice, but seeing it in action is better. Below is a simple email classifier. Toggle between supervised and unsupervised mode, then click on different emails to see how each approach handles the same data. Notice the difference: supervised gives you a definitive label, while unsupervised finds groups.
Email Classifier
The model learned from thousands of labeled examples (spam / not spam) and now predicts a label for new emails.
Key Takeaways
- AI is the broad goal of building intelligent machines. Machine learning is the most successful approach to achieving it — learning patterns from data instead of following hand-written rules.
- Supervised learning uses labeled examples to learn predictions. It's the most common type and powers spam filters, image recognition, and translation.
- Unsupervised learning finds hidden patterns in data without labels — great for clustering, segmentation, and anomaly detection.
- Reinforcement learning learns through trial and error with rewards, powering game AI and helping fine-tune language models.
- Deep learning is a subset of ML that uses multi-layered neural networks. It's behind most modern AI breakthroughs, from image generation to conversational AI.
Common Misconceptions
- "AI and ML are the same thing." — ML is one method for achieving AI. There are other approaches (like rule-based systems), but ML has become dominant because it scales with data.
- "AI understands things the way humans do." — Current AI finds statistical patterns. It doesn't "understand" in a human sense — it's very good at prediction, not comprehension.
- "You need a PhD in math to learn this." — The core ideas are intuitive. You'll pick up the math gradually, and most of it is just multiplication and addition at scale.