Thank you for this post. It makes neural networks seem so simple, assuming you have some college-level math.
It seems there is important work here involving looking inside neural networks, understanding the shapes of their weights, and both simplifying these weights to functions and mapping the sets of weights to meaning that is inherent in the data being processed. It seems the concept of 'meaning' and 'how' are what is missing from neural networks, but it must be right there, in the distributions of weights.
I suppose a complex neural network is kind of like an activation function for some larger process.
I'm glad you liked it.. I agree that there is still quite a long way to better understanding (and subsequently also working) with the neural networks and the "functions from weights" that are created inside the models. I personally have still long way in front of me to be able to claim that I understand and know how to build and train a good model.