I am currently reading "Data Science from Scratch" and feel this relates here. The book is nicely showing the principles of machine learning by showing how all the blocks are build. (We can program along as we read.)

Also: My kid is too young for this, but I wonder if building a circuit physically beats programming and running it on screen. After all we can't really see the electrons flow so whats the point of building such a clunky model if we can make it in code?

Just wondering if there's a real education advantage of doing it like this.

I started reimplementing the book in Swift to absorb a little more:

https://github.com/melling/data-science-from-scratch-swift

Hey this is pretty cool, thanks for sharing.

I like this just because its cool, but I agree with your point that just the raw parts with basically invisible electrons flowing through probably doesn't help teach. A light up diagram where you can see the signals and their values moving through the network slowly might be a good teaching aid, but I don't think it would help with much other than the absolute basics of "what even is a neuron" type knowledge.

As someone who just started working at an AI startup and with a 7 year old who just got Snap Circuits for his birthday 2 days ago, this is super timely and relevant. Thanks for posting.