FOUNDATIONS • 29 / 397
Start before LLMs. Understand where AI, ML and neural networks fit.
Gradient
A gradient indicates how a small change to a parameter would change the loss.
Think of it like
Think of Gradient as one layer in a very large toolbox: you only need to understand what job that layer performs.
Real life
A phone spam filter, recommendation engine, fraud detector or voice assistant may use Gradient somewhere inside.
SRE lens
Training systems compute gradients; inference systems usually do not.
Remember thisA gradient indicates how a small change to a parameter would change the loss.