AI: Background & Significance

A visual introduction to how modern AI systems represent meaning, learn parameters, and produce predictions. Start with the slidedeck canvas for the whole talk in one view; the interactive demonstrations that accompany it follow below, and each can be explored independently.

• >>
  Slidedeck canvas The talk itself — pan and zoom the complete Excalidraw canvas, every diagram and note in the order it was presented. Start here for the whole picture; the demonstrations below expand on individual pieces of it.
  [EXCALIDRAW]
• ▶ Presentation recording Watch the complete recorded presentation. Expand this entry to play it here without leaving the page. [RECORDING]

INTERACTIVE DEMONSTRATIONS

1. 01
   Semantic vectors Build the familiar man → king, woman → queen relationship in a simplified 3D space.
   [WEBGL]
2. 02
   Embedding space Explore HVAC-language embeddings and see semantic neighborhoods emerge from data.
   [INTERACTIVE]
3. 03
   How a layer learns Follow a real toy calculation through prediction, loss, backpropagation, optimization, and inference.
   [LIVE MATH]