He starts with a single perceptron and a simple question: "If the output is wrong, how do we adjust the weights?"
Have you read Tariq Rashid’s book? Did you run into the Python 2.7 trap? Let me know in the comments below!
This is the only pain point. The book was published in 2016. If you download a raw PDF and copy-paste the code, it on Python 3.10+ without modifications.
TL;DR: If you have ever opened a machine learning tutorial, seen a wall of calculus, and closed the tab—this book is your cure.
I recently finished reading Make Your Own Neural Network by Tariq Rashid. I cannot recommend it enough. In fact, I believe it should be required reading before anyone touches TensorFlow, PyTorch, or even Scikit-learn.
You will come out the other side not as a person who uses AI, but as a person who understands AI.