Learning path · AI · 16 layers

AI fundamentals: ML, deep learning, domains

Read left to right. The umbrella, then ML’s three signals, then depth — CNN, RNN, transformer — then where it lands: NLP, vision, speech, rank. Pan down for the train loop, with the traps crossed out.

Click a shape or a question

Questions this tree answers

A classic model · data through serve

Artificial intelligence

Three lanes — signal · depth · sensor

Machine learning

Deep learning

AI domains

Classic loop and traps

Classic loop — data → serve

don’t

rootArtificial intelligence

umbrella · pick a branch by data and output

fitMachine learning

examples → function
loss · hold out a test set

labeledSupervised learning

x → y you named
labels are the ceiling

structureUnsupervised learning

no target
you still name the clusters

rewardReinforcement learning

act · delayed reward
sim or preferences

depthDeep learning

stacked layers · learned features
still a split and a loss

unitNeural networks

Wx + b · σ · stack
forward / backward

gridCNN

shared filters on space
vision’s inductive bias

chainRNN / LSTM

hidden state walks time
long context dies — attention won

blockTransformers

QKᵀV over the window
the LLM block, without the API

sensorAI domains

same math · different I/O
metric is part of the domain

textNLP

tokens in
label or generate — still measure

pixelsComputer vision

class · box · mask
mAP, not a blog accuracy

waveSpeech

ASR / TTS
WER and a latency budget

rankRecommendation / prediction

retrieve then rank
offline metric vs A/B

01Data

labels · no leakage
frozen test ids

02Features

train-serve skew
replay the vector

03Train

val · checkpoint
version code + data hash

04Evaluate

baseline + slices
the gate

05Serve

budget · rollback
an API, not a cell

rejectedDeep learning is magic

depth without a split
is a GPU heater

rejectedSkip labels, skip split

train on test
or never gold — you cannot know

72%

Click a box · three arrows off a layer are what it allows

What the fundamentals are