AI pioneer Sejnowski says it’s all about the gradient

To machine learning pioneer Terry Sejnowski, the mathematical technique called stochastic gradient descent is the “secret sauce” of deep learning, and most people don’t actually grasp its true significance.