What is machine learning? (Day 1)
One sentence: A computer gets better at a task by learning from data—not by someone coding every rule.
If you've ever wondered what "machine learning" actually means, you're in the right place. No math, no jargon. Just the idea, why it matters, and how it's a lot like the way we learn.
The question everyone asks
You hear "machine learning" and "AI" everywhere. This week we stick to the basics. Today: What is machine learning?
Answer: the machine uses data to get better at something, instead of a human writing every single rule.
It's learning from data—really
Machine learning (ML) = a computer improves at a task by using data, rather than by someone programming every rule by hand.
Think about how you learned to tell a cat from a dog. Nobody handed you a manual: "if whiskers = X and ears = Y, then cat." You saw many examples and your brain found the pattern. Machine learning does the same thing: lots of examples (data) → find patterns → use those patterns on new cases.
So: data in → patterns learned → use those patterns on new situations.
How that shows up in real life
The engine of ML is learning from data. You give the system examples; it adjusts until it does what you care about.
- Spam or not: Thousands of emails labeled "spam" or "not spam." The model learns what spam looks like and labels new ones.
- Cat or dog: Lots of photos with labels. The model learns visual patterns and guesses on new photos.
- Recommendations: What you click or watch. The model learns your taste and suggests what's next.
In every case: data, some kind of feedback (right/wrong or good/bad), and repeated adjustment. Same idea as how we learn from experience and feedback.
One picture: the whole flow
You can say it in four words:
Data → Learning process → Model → Predictions
- Data: The examples (emails, images, clicks, etc.).
- Learning process: The algorithm that looks at the data and updates the model.
- Model: The "trained" result—what was learned.
- Predictions: Using the model on new inputs to get an answer.
You can sketch this in Excalidraw: four boxes and arrows, left to right. Optional: add a second row for the human parallel—Experience → Thinking → Understanding → Decisions—so the analogy is obvious.
Excalidraw examples
- Minimal: Four rectangles in a row: Data → Learning process → Model → Predictions, with arrows between them.
- With analogy: Same row, plus a row below: Experience → Thinking → Understanding → Decisions. Label the rows "Machine learning" and "Human learning."
Why this matters
ML is already everywhere: search, recommendations, voice assistants, fraud detection, medical tools. You don't have to build any of that to benefit from the basics. Once you get "learning from data" and the flow Data → Learning → Model → Predictions, you can read the news, compare products, and judge "AI" claims with a clearer head.
What's next
Tomorrow: supervised vs unsupervised (and a quick nod to reinforcement)—different ways the machine can learn. That's how you choose the right approach for a problem.
See you in Day 2.