Neural networks are a fundamental part of Artificial Intelligence: Software systems that train themselves to make sense of the human world. But if you want to understand how they work at a basic level, a cool new website allows you to get under the hood.
The site is based on Google’s new TensorFlow software library, which is an open source resource for people working with machine learning systems. Daniel Smilkov and Shan Carter, however, have used it to make a simple demonstration of how neural networks work. It may still look a little daunting, with all its options and inputs and layers, but it’s the most straightforward tool I’ve seen for learning about how neural nets work. And, as the pair explain, “don’t worry, you can’t break it.” Which is always nice to know.
It’s all annotated to explain what the various parts of the system do, but it’s perhaps most fun to just dive right in, piece together some neurons, choose a dataset and get your fledging AI to learn some stuff. And if you want a little background, why not read a little about the history and application of the neural networks that you’re already using every day.