Equivalences:

I love equivalences. One example of an equivalence is how a thousand core graphics card simulating a neural network with a hundred million neurons is equivalent to a hundred million neural circuits specifically designed to simulate that exact neural network at that exact speed. One advantage is that it is less costly to make; due to having less than a ten thousandth of the processing units and not having to be custom built for every neural network. Another advantage is that the same hardware can run other neural networks and even run other applications, like videos and video games.

An example that people without expertise in computer science or deep learning might understand is that a smartphone is like a personal computer in your pocket, an album of photos in your hand, a portable video camera, an office you can hold, a gaming console on the go and, on top of that, a plain, simple phone that can call and text; as well as video call.