Google, Stanford build hybrid neural networks that can explain photos

Researchers at both Stanford University and Google have fused state-of-the-art neural network techniques for vision and language in order to create hybrid systems that can analyze images and produce natural-language explanations of what’s happening in them.