Big data plus cloud computing plus the algorithms of the 80s have combined in an explosive mix to result in remarkable achievements. Computers can 'see', they can 'hear', they can tell us what movies we'll like, what clothes we'll like. Pathologic and radiologic diagnosis is now better done by machine. But these are all questions directed to solve the 'problem' of an 'individual'. The real problems of our world though are not usually due to an individual but the emergent behavior of our over populated communities. Can we apply deep learning to solve these issues? For example, can deep learning predict who will become a terrorist? We may have a big enough sample size to train such a neural net if we throw in all the biographies of the Guantanamo inmates and those at other facilities. Or to take a less dramatic problem, can deep learning identify who will become sick with the flu? Using the myriad health parameters available, social parameters, dietary parameters - we could identify those people at risk. Even though we will never know why (because deep learning's a black box) we will at least be able to intervene.
