HackerTube / Video
GradIEEEnt half decent: The hidden power of imprecise lines
suckerpinch
Tom 7 (suckerpinch) delivers a long, comedic digression on gradient descent and floating-point imprecision. He shows how rounding behavior can supply the nonlinearity needed to build a neural network.
Cited as a fun, unconventional example of neural networks built from floating point error.
4 Hacker News mentions · 2023–2024