Artificial Intelligence System passes expert exam

Qualified medical practitioners are in severe shortage in many countries of the world and medical training is typically a lengthy procedure. For example, a medical student usually spends more than 5 years of school study and then takes a few years of residency training. Though in recent years, plenty of medical AI systems spring up in both the research and the industry communities, almost all of them are designed to merely solve some pre-specified medical problems, such as classifying skin cancer, detecting pneumonia, and producing treatments for a few pre-defined cancers or diseases. There is still lack of an efficient AI-enabled computer model which, like candidate general practitioners, can automatically learn and master a wide range of medical knowledge from a large medical corpus, and apply medical knowledge, concepts, and principles to solving generic medical problems. The barriers for achieving this goal mainly include

(1) learning such wide range of medical knowledge from text corpus is still an unsolved challenging problem in research communities;

(2) understanding medical problems and making reasoning with medical-views at human-doctor-level is also very difficult for a computer program.


The AI system, called Med3R, was examined by taking a written test of National Medical Licensing Examination in China in 2017. The results officially reported by National Medical Examination Center (NMEC)1show that Med3R has successfully passed the exam and surpassed 96.3% human examinees.


read more at:


Posted by Alan Brown, Tendron Systems Ltd, London, UK

Tendron Systems Ltd, International House,

See Data Science Group on Facebook