I admit I have been abysmally behind learning about machine learning techniques. When TensorFlow was released, I took a bit of time to set it up on my machine and run a few tutorials but left it at that. I never dug deeper, especially in understanding its uses.
Jeffrey Dean gave a talk at YCombinator for the companies using AI all about Google Brain and how they approach AI and ML at Google now. He explains the many applications it has at Google (not including internal cases) and how the Brain team has built models, built & executed experiments, and even built specialized hardware for even faster computations.
When a CPU and GPU fall short, you know you’re doing something interesting.