There may be a misconception about the practicality of AI/machine learning in 2017. When we say "AI" we're not referring to artificial general intelligence (AGI) or some exotic technology, but rather in the context of training models that are more efficient at performing a particular task than a human designer/programmer can produce. If you want to get started with understanding the practicality of machine learning applications in 2017, I would suggest looking into https://www.tensorflow.org/ (we are not training with TF, but it's the most practical resource available).