Understanding Deep Learning — and why it changed everything
Machine learning learns patterns from data — no explicit rules required.
This deck fills the gap between the two — what deep learning is, and why it's a different beast.
A human expert had to decide what features to feed the model.
Human expertise required at every step.
Doesn't scale. Doesn't generalize.
What if instead of telling the machine what to look for… we let it figure that out itself?
Inspired by the brain — but a very simplified version of it.
Raw data goes in — pixel values, word tokens, numbers. No human hand-crafting required.
The network automatically discovers which patterns matter. You don't program these — the training process figures them out.
A prediction: "cat vs dog," "spam vs not spam," "translate this sentence."
Imagine the network learning to recognize a human face from a photo.
The same principle applies whether the input is images, text, audio, or sensor data — layers build progressively richer representations.
No programmer writes the internal rules — the network finds them by making millions of mistakes.
The numbers on each connection. Training adjusts billions of these tiny dials.
A single number measuring how wrong the guess was. Training tries to make it as small as possible.
One full pass through the training data. A model might train for hundreds of epochs over days or weeks.
Neural networks existed for decades. What changed wasn't the idea — it was the scale.
As you give a neural network more data and more computing power, it keeps getting better — reliably, predictably. This was not obvious, and it changed how everyone thinks about AI.
Each "parameter" is one weight in the network — one dial the training process tuned.
Deep learning is not always better. It depends on what you have to work with.
The next two modules will make a lot more sense now that you know what's powering them.
ChatGPT, Claude, Gemini — all powered by Transformer neural networks with billions of layers and hundreds of billions of weights.
Image recognition, object detection, medical imaging, self-driving cars — all powered by Convolutional Neural Networks (CNNs).
A nested view — every outer ring contains everything inside it.
Deep learning = neural networks with many layers that learn features automatically from large amounts of data.
It is the core technology behind virtually every headline-grabbing AI system of the past decade.
Modules 5 & 6 show you what deep learning looks like in action — for language and for images.