Latent thoughts and specific review regarding an introductory machine learning course taken at the University of Auckland.
My Prior
Since my first introduction to computers and computer science, one of the brightest flashes of insight I’ve experienced is the inherent humanity of it all. Computers, from high-level languages all the way down to basic electrical design, bear the marks of their creators. But machine learning seems different, in that it aims to produce machines which can not just follow specific instructions from people, but impersonate people by making predictions based on some model of past experience. Because of this, and a slight dose of contempt for the domains in which machine learning has been popularised, I have been suspicious.
My Update
I have found myself far less suspicious, and more intrigued by machine learning after a course of study. By studying the fundamentals of machine learning, and peeking inside a few modeling techniques, I feel that I have upgraded my understanding. And after implementing different types of models, like Naive Bayes and Support Vector Machines, I feel significantly more confident in moving forward with the rest of my study program and career. I remember feeling a similar step change in confidence, after building an Android application from scratch as part of an undergraduate course. Based on these matched experiences, I’ve learned that my confidence in any given arena is greatly boosted by completing a significant and properly contextualising project.
Course Format
I am surprised at how appealing the format of the ML course was for me. My previous experience in technical university courses involved a few more low-level implementation projects, and significantly less high-level synthesis of ideas. Working with classmates to produce a tutorial presentation, engaging with other presenting groups, and creating my own questions on the Peerwise platform promoted my understanding of the content more than I expected. As a result, I have enjoyed this first semester of postgraduate study more than any those of undergraduate, though I’m sure the maturity that I’ve gained in the previous few years is a stronger factor in that change.
Application
I still recognise within me some uncertainty about the usefulness of machine learning in those domains which are most important. As interesting as spam classification problems are, they are not all that critical in the grand scheme of things. Issues like global health and poverty, though messy and less tractable, are arenas where data-driven decision-making could potentially do a lot of good. Thanks in part to the ML course, I am starting to believe that machine learning has a place in more consequential domains. I think I better understand what machine learning is: another set of computational tools which, if used carefully, can be used to do humanity’s difficult work. As a burgeoning machine learning practitioner, I am more excited than ever to learn more.