This is really a good read for anyone who is in the field of Artificial Intelligence (AI). Machine learning (an aspect/subset of AI), is the current trend or the buzzword in the technology world at the moment. It has been deployed in wide range of computing tasks e.g. self-driving car, automatic translation etc. It is so powerful that it can beat the world champion of board game Go in 2017. Despite its successful applications, we are still far from creating a true artificial intelligence. Why is it so? Machine learning models do not able to react to new circumstances. Unlike us human being, the models cannot contextualize its situation. Therefore, when a new input is given to the models, frequently it will fail to react accordingly. Another reason is that machine learning cannot understand (or model) the cause -effect relationships which is a necessary to achieve the true artificial intelligence. Humans know and understand that in this world, there are causal relationships between actions and consequences such as the effectiveness of a treatment in preventing a disease. Currently, even the state-of-the-art of machine learning such as Convolutional Neural Network (CNN) does not model such relationships.
Full paper: Theoretical Impediments to Machine Learning with Seven Sparks from the Causal Revolution [link]