AI, ML & Deep Learning
The nested definitions of AI, machine learning, and deep learning — plus the three ways machines learn.
Three nested circles
Artificial Intelligence (AI) is the broad field of making machines perform tasks that normally require human intelligence. Machine Learning (ML) is a subset of AI where systems *learn patterns from data* instead of following hand-written rules. Deep Learning (DL) is a subset of ML that uses multi-layered neural networks — loosely inspired by the brain — to learn very complex patterns from very large datasets.
AI is the whole idea of "smart machines." ML is one way to get there: show the machine thousands of examples and let it figure out the rules. Deep learning is ML with an enormous stack of pattern-detecting layers — like an assembly line where the first station spots edges, the next spots shapes, the next spots faces.
Where deep learning shines vs classic ML
- Classic ML (regression, decision trees) works well on structured/tabular data with engineered features — and is cheaper and easier to explain.
- Deep learning excels at unstructured data: images, audio, video, and natural language.
- Deep learning needs much more data and compute, and its decisions are harder to interpret.
- Generative AI is built on deep learning — specifically giant transformer neural networks (covered in the next module).
The three learning paradigms
| Type | Data | Learns to… | Examples |
|---|---|---|---|
| Supervised learning | Labeled examples (input + correct answer) | Predict labels for new inputs | Spam detection, house price prediction, image classification |
| Unsupervised learning | Unlabeled data | Find hidden structure on its own | Customer segmentation (clustering), anomaly detection, dimensionality reduction |
| Reinforcement learning | No dataset — an agent acts in an environment | Maximize cumulative reward through trial and error | Game playing, robotics, AWS DeepRacer |
Spot the keyword: "labeled data" → supervised. "no labels / find groupings" → unsupervised. "reward / trial and error / agent" → reinforcement learning. Within supervised learning: predicting a category = classification; predicting a number = regression.
Common AI task types
Key points
- Classification — assign a category (fraud / not fraud).
- Regression — predict a continuous number (tomorrow's demand).
- Clustering — group similar items without labels.
- Anomaly detection — flag unusual data points.
- Computer vision — understand images/video.
- Natural language processing (NLP) — understand and generate text.
- Recommendation — suggest relevant items.
- Forecasting — predict future values of time series.
Which statement correctly describes the relationship between AI, ML, and deep learning?