Recently on Twitter, I ran a quick poll to understand if people are fascinated by causality and the response was quite overwhelming.
Results: 44/45 people voted interested, to see what I have to say.
So, I decided to do a full thorough review of Causal Inference in Machine Learningstarting with The Book of Why for many reasons-
- It's one of the latest, beginner level books on the topic thus easily comprehensible and up-to-date.
- Causality is one of the most fascinating topics in the study of artificial intelligence and yet one of the most long ignored ones thus an excellent tool for the imitation of human learning.
- The book has been penned by no other than Judea (Pearl), who I greatly admire and also whom I owe a lot for taking such immense leap of faith in me, as my mentor.
- Causality has not yet been studied widely - in combination with other theories of machine intelligence, namely information theory, deep learning, probabilistic inference, associative memory etc.
- I personally couldn't find any unbiased critical analysis of Judea's work. While there are a few interesting rebuttals of his work by Nancy Kreiger, Rubin and some other academics in general but unfortunately not constructive.
- Thus, the goal of this exercise is - first, to learn more about his work among that of others in causality and secondly, to review his work and dismantle it and pick the true gems and rebuild the rest.
To sum it up, I strongly believe Geoffrey Hinton's statement applies to almost every tool, technique and method we know of in science and esp more so in artificial intelligence. Let us start once again from the whiteboard...