Deep learning: what it is and how it differs from machine learning

Deep learning is a type of machine learning that uses neural networks with many layers. "Deep" refers to the number of layers, not to deep understanding.

A deep neural network: more layers for harder tasks

In plain words

A small neural network is like a ten-piece building set: you can make a toy car, but not a castle. A deep network is like a huge LEGO box with thousands of pieces and multi-stage instructions. The more stages, the more complex things you can build - from face recognition to writing text.

How it works

You take a neural network with dozens or hundreds of layers.
Each layer learns to see slightly more complex features than the one before.
Such a network is trained on very large datasets.
Training needs powerful graphics cards or special accelerators - weeks or months of computing.
The finished model can do what was once thought impossible for a computer: understand language and see.

A multi-layer network and the graphics cards that train it

Where you'll see it

Almost all modern AI is deep learning: camera night mode, speech recognition, automatic subtitles, translation, image generators and large language models. The deep learning boom started around 2012, when such networks unexpectedly won an image recognition competition.

Does it matter when buying

For buying - indirectly. Deep models are heavy, so the neural processor, RAM and cooling matter. If a feature runs "on device", a flagship will do it faster and without internet, while a budget phone either won't, or will send your data to the cloud.

Myths

Myth: Deep learning means the model understands a topic deeply. Reality: No, "deep" is about the number of layers in the network.
Myth: A deep network is always better than a simple one. Reality: For simple tasks a small model works just as well, but faster and more efficiently.

FAQ

How is deep learning different from machine learning?

It's a subtype of machine learning specifically on multi-layer neural networks.

Why does it need graphics cards?

Because there are billions of identical multiplications, and GPUs do them in parallel.

Can a teenager learn deep learning?

Yes. There are free courses, and a laptop plus online services is enough for first experiments.