Backpropagation

can we compositionally do knowledge representation? Here's why this is like directly relevant to machine learning.

Backpropagation, right? The reason backpropagation is so fucking computationally and data hungry, right? Is it's the frame problem.

Right.

Your error signal goes to the entirety.

It's like literally 300 years.

Backprop has been around that long.

Linear regression is one of the simplest types of math you could possibly do, actually.

It's like put tables of numbers together and then multiply and add them together.