Five guides through neural-network history
The Neuron Before Learning
A fixed threshold neuron turns firing into logic and shows how simple decisions compose into computation.
Explore McCulloch and Pitts
The Neuron That Learned
Rosenblatt adds adjustable weights, letting examples move a decision boundary.
Explore the perceptron
The Error That Traveled Backward
Backpropagation carries credit through hidden layers, making useful internal representations learnable.
Explore backpropagation
The World, Learned in Layers
Depth, data, and computation turn representation learning into a practical research program.
Explore deep learning
Every Token Learned Where to Look
The Transformer lets each token route context through attention, opening the path to modern language models.
Explore attentionThe conceptual throughline
Modern AI is a stack of older ideas.
Logic supplied the parts. Learning adjusted them. Backpropagation assigned credit. Depth built abstractions. Attention routed context.
-
LogicTurn neural firing into a rule.
-
CorrectionMove the weights after a mistake.
-
CreditTell each hidden connection how it contributed.
-
HierarchyBuild useful features across many layers.
-
ContextLet every position gather the information it needs.