Welcome to Connecting the Dots — an intuition-first introduction to modern AI, focused on understanding.
Starting from one question: how can we model a joint distribution over many variables without an exponential number of parameters?
From the chain rule and Bayesian networks through NADE, MADE, and the Transformer — each idea appears as a response to a limitation in the previous one.
Neither ResNets nor Transformers overwrite their representations — they accumulate them.
Depth is accumulation — not replacement.
"Connecting the Dots traces the architecture of ideas that make modern AI work — one honest derivation at a time."