‘The Master Algorithm’ Review

A Great Book to Learn Machine Learning

Pedro Domingos’sThe Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our Worldhas really helped me to learn AI and Machine Learning as I get ready to learn AWS’s Sagemaker. Bill Gates recommended the book in 2016 and at the same time ‘Superintelligence’ by Nick Bostrom as the two foremost books to learn about AI.

I started reading it about two years ago and it took some time to digest, and I haven’t yet fully digested all the rich insights into the five tribes of Machine Learning but I understand things on a ground level now well enough to come up with my own theories of AI. I finally came around to neural networks as being preeminent for AI. The fact that is has so many things going on with it that even the people who invented it, don’t completely understand how it works I think means it simply has something of great value  that has to be deciphered more. 

My Burgeoning ML Theory

Yes, I did and I am trying to get to some Master Algorithm and this has made the learning of ML quite interesting and a pleasant experience. My ongoing theory is that AI should be divided up into Learning, Memory and Decision-making. And this related to three forms of reasoning: Analogical, Induction and Deduction. The ML tribe of Analogizers are about Learning because learning is all about making connections. The ML tribes of Symbolists and Connectionists concentrate on induction which deals with memory. The ML tribe of the Bayesians deals with Decision-making because ironically all decision-making is related to probability and chance.

The Evolutionaries, I didn’t quite understand how they are even a tribe of Machine Learning and not just an important element of all Machine Learning. I’m also not thoroughly convinced we need an all-knowing or one final Master Algorithm. Is that the way modern technology ever works? What about the invention of discrete neural networks that cater to specific issues and industries? Yet it all has to come together to make all that complexity have a simple roadmap.