Andrew Trask is an AI superstar, having come up with the idea for Generative Adverserial Networks (GANS) - one of the most exicting new AI techniques today - in a bar.
In this book he comes down to earth and explains neural networks from scratch to mere mortals. The book is still being written, but whats there so far is a really good easy to understand introduction to neural networks.
While frameworks like tensorflow and pytorch are essential for actually doing useful and practical AI stuff, they abstract away a lot of the inner workings of how data is transformed as it passes through the many layers of a deep learning network.
Grokking Deep Learning has you build a NN from scratch, by hand, propagating neurons down the hill through the snow, first backwards then forwards. Its a great excercise to learn how Neural Networks work internally.
This book will teach you the fundamentals of Deep Learning from an intuitive perspective, so that you can understand how machines learn using Deep Learning. This book is not focused on learning a framework such as Torch, TensorFlow, or Keras. Instead, it is focused on teaching you the Deep Learning methods behind well known frameworks. Everything will be built from scratch using only Python and numpy (a matrix library). In this way, you will understand every detail that goes into training a neural network, not just how to use a code library. You should consider this book a pre-requisite to mastering one of the major frameworks.