01
Neural Networks
4 filmsComing soonOne small neural network reading handwritten digits, opened up until every number in it is on screen.
Machine Learning
The mathematics the AI collection left out, taught in full and shown moving: every number on screen comes from a real run on a real, public dataset, and every idea is a picture before it is a symbol.
02
Mathematics for Machine Learning
6 filmsComing soonThe linear algebra and the calculus a neural network is built from: vectors, the dot product, matrices, derivatives and the chain rule, each shown on a real network.
03
Probability and Statistics
5 filmsComing soonHow a model's scores become probabilities, how evidence updates a belief, how data is summarised and squeezed, and how a classifier is judged.
04
Training
6 filmsComing soonHow a network's weights go from random to useful: the loss, gradient descent, backpropagation, mini-batches, optimisers and regularisation, each on a real training run.
05
Deep Learning
4 filmsComing soonThe architectures that made deep networks work on images: convolution, residual connections, autoencoders and diffusion.
06
Transformers
5 filmsComing soonThe transformer, the architecture behind the word predictor, opened up block by block on GPT-2, whose weights are public.