AI Foundations · 1.2
📊 How Machines Learn
The training loop that turns data into 'intelligence'⏱ ~2 min
🚲Learning to ride a bike, not reading a manual
Nobody learns to ride a bike from a written manual. You try, you wobble, you fall, you adjust, you try again — until your body just 'knows' it. Machine learning works the same way: the model makes a guess, sees how wrong it was, adjusts slightly, and repeats thousands of times until it's good.
The Core Idea: Learn From Examples
Instead of a human writing every rule, you give the machine a big pile of examples and let it find the pattern. Show it enough photos labeled 'cat' and 'dog', and it figures out — on its own — what makes a cat look like a cat. The 'learning' is the machine slowly adjusting its internal numbers to make fewer mistakes.
The Vocabulary You'll See Everywhere
| Term | Plain-English Meaning |
|---|---|
| Model | The thing that learned — a big set of numbers that turns inputs into predictions |
| Training | The process of feeding examples in and adjusting the model to be more accurate |
| Training data | The examples used to teach the model — quality and quantity both matter enormously |
| Features | The individual pieces of information the model looks at (e.g., pixel colors, word counts) |
| Label | The 'correct answer' attached to a training example (e.g., 'this photo = cat') |
| Inference / prediction | Using the trained model on new, unseen data to get an answer |
| Parameters / weights | The internal numbers the model adjusts during training — big models have billions |
Why Data Quality Is Everything
🔒 Security'Garbage in, garbage out' is the iron law of ML. A model can only learn from what it's shown. If your training data is biased, incomplete, or wrong, the model will confidently reproduce those flaws — and it can't tell you it's wrong. This is why data quality matters more than clever algorithms.
★ FactA famous cautionary tale: a model trained to spot skin cancer in photos learned to flag rulers instead — because in its training photos, cancerous moles were more often photographed next to a ruler for scale. The model found a real pattern in the data, just not the one anyone wanted. This is why understanding your data matters.
🧠Quick Checkfirst try = +5 XP
In machine learning, where does the 'intelligence' come from?
🎮 Practice what you learned
⭐ 0 XP🔥 0 days