Machine learning (ML) is the idea that instead of programming rules by hand, you let a system learn patterns from data. It's a subset of AI, and deep learning is a subset of it.
The nesting dolls
Artificial Intelligence (any machine doing "smart" things)
└── Machine Learning (systems that learn from data)
└── Deep Learning (ML with many-layered neural networks)
└── Transformers / LLMsThe three flavors of ML
| Type | You give it | It learns |
|---|---|---|
| Supervised | Labeled examples (input→answer) | To predict the answer (spam/not-spam) |
| Unsupervised | Unlabeled data | Structure/clusters (group customers) |
| Reinforcement | An environment + rewards | A strategy (play a game, align an LLM) |
The shift it represents
Traditional code: rules + data → answers. Machine learning: data + answers → rules. You show it examples and it infers the program — which is why it solves problems (vision, language) too fuzzy to write rules for.
LLMs are the deep-learning end of this tree — see A Brief History of AI for how we got here.
Related: A Brief History of AI · How Neural Networks Learn · How Transformers Work