The simplified approach makes it easier to see how neural networks produce the outputs they do. A tweak to the way artificial neurons work in neural networks could make AIs easier to decipher.