As shown on the cover page of the June 2017 issue of Newsweek, Silicon Valley thinks Artificial Intelligence (AI) will cure our sick health care system . Really? As a CIO and MD scientist, I see hypes as well as hopes. AI is becoming the new frontier in healthcare. And here’s why.
It is no longer the AI in the 1980s.
When I read my first AI textbook in the 1980s, I was amazed by the tremendous possibilities. For instance, by knowing the fact that “Toby is a dog” and “dogs have four legs”, the computer is going to infer that Toby has four legs. However, it is very hard to engineer all human knowledge into simple facts and rules. The theoretical beauty of first-order logic from the Prolog programming language could not solve the complexities in the real world.
The AI in 2017, however, is much different. The focus has shifted from knowledge engineering to machine learning. Deep learning now plays a critical role to sift through an extraordinary amount of data. For instance, self-driving cars are making decisions from one gigabyte of data per second generated by dozens of sensors. This data-driven approach lifts the human burden on crafting the decision rules for computers. From a machine learning perspective, the statistical worry of model overfitting is also alleviated by the exceedingly large amount of training data.