How AI actually works

by William Craig

0 of 23 solid · next: Learning from examples

Learning from examplesEverything is numbersTraining vs. usingHow wrong you areWhat a 'neuron' isText as tokensMemorising vs. learningGetting less wrongWhy depth helpsMeaning as positionHolding data backWhat parameters areBlame, spread backwardsAttentionThe transformerWhat the model seesNext-token predictionRetrieval (RAG)SamplingBecoming an assistantHallucinationTool useAgents
FoundationsHow learning worksNeural networksLanguage modelsWhy they behave as they doSystems built on top