I hear ya on the data mining observation. I have been digging into 'applied' ML for the last two years or so and its painfully obvious how many articles are based on science fiction and not the current state of things (which appears to be mostly reinforcement learning)