For new readers and those who request to be “好友 good friends” please read my 公告栏 first 

The other night, the National Public Radio (NPR) of the US had a program on the title of this blog article. It talks about a current computer algorithm in experimental use on the west coast of the US which using deep learning neural network to predict which patient under care that has a high probability of dying within the next 12 months.

We all know that the current state of medical science is half knowledge-and-science based. The other half are many medical and physiological questions we do not know the answers to. Consequently, experience comes into play. Doctors with many years of experience treating patients often can provide better diagnosis and care than new graduates. The AI algorithm in discussion is nothing but using big data (i.e., a great deal of patient experiences) to construct a computer model which fits and summarizes these data and makes predictions on new data from a particular patient. Thus, it behaves as a super experienced human doctor but with the same bias and failing. For example, since the AI algorithm was trained using patient data mostly from the US and who are insured, its predictive ability may not work well on patients who are from Africa. Furthermore, if the algorithm predicts that a particular patient has high probability of dying in the next 12 months, should the patient be told? How would the patient feel if the prediction came from a computer algorithm? These are just two of the many questions the medical profession will face when AI is employed on a large scale in health care.  Technological progresses these days often outrun our abilities to adapt. Income inequality, gun violence, and other social ethical phenomena are often unintended consequences of good intentions.