  			STILL NEED FOUCUS ON MATHEMATIC THEORY
  When mentioning the future of AI, the author says that the work on theory proving and model-based approaches will receive reduced attention, due to its strong relies on mathematics assumption that are hard to satisfy the real-world challenge. In comparison, the data-driven approaches will become the mainstream. as one of data-driven algorithms “Deep Learning” has get a remarkable success recently. 
  Numerous notable people work on Deep Learning also states: the critical and practical work on the area is to find the good data representation. May be it can get novel result and make a huge process in the short run. but in the long run if the Deep Learning still lack rigorous and strong mathematic theory to back up what really happens in the black box, the essence of it wii be still mysterious. Is it exit a better start point in the navigating space and can we get more faster convergence, is it the Deep Learning convex-like, which are all need a complete and strong theory to explain. So the rigorous mathematic theory behind the AI algorithms still should be valued.
  However, the people who work on theory tend to be ignored by the National Science Foundation and big company. Because the complexity and abstruse of AI theory results in the slow theory-established and economic benefit lacked. Nevertheless, the complete and strong mathematic theory behind the AI algorithms can become nuclear weapon to destroy any barrier. For example Galois theory not only provides the beautiful answer on the question why is there no formula for the roots of a fifth (or higher) degree polynomial equation in terms of the coefficients of the polynomial, using only the usual algebraic operations (addition, subtraction, multiplication, division) and application of radicals (square roots, cube roots, etc? (-— by wikipedia) but also explains in detail why it is possible to solve equations of degree four or lower in the above manner, and why their solutions take the form that they do. Without Galois theory, Many great mathematicians such as Cauchy, Lagrange can not solve the problem. Moreover the significant and powerful of researching on math structure is revealed by the Galois theory.
  So the LeCun and others suggested focus on rigorous theory of Deep Learning is much needed, as it might to get optimal and faster progress and better intuition.
  And in the Health Care area, the author leaves out the apply of speech recognition. It is reported from Healthcare IT News that speech recognition improve the clinical documentation in many ways, as the demand for documentation is on rise. The patient can speak into the microphone attached to the computer and the words are displays as they are recognized. And dictated, corrected, authenticated are all completed in one sitting, the front-end speech recognition can reduce the transaction cost, the documentation time, and  give the patient more time to narrate himself.Then the condition of patient can be better known. 
  Niklaus Wirth says algorithms + data structures = programs, In my opinion, the algorithms + data structures + (big) data = AI. Creative and effective algorithms are still hidden in the dark, which need scholars to find. Also the engineers needs to design more efficient systems to speed up the algorithms. And how to collect and process scale and diversity of data will be the future  trend in AI research. 
  Above all, as the old saying goes fortune's wheel is ever turning. so the ignored topic in AI area nowadays, may become hot in the future. It tells us that if you want to be scholar, you just do whatever you are interested in and never give up and do not much worried about what are the hot topics in AI in recent.