posts

I recently finished watching My Name on Netflix, and was once again amazed at how effective South Korea’s writers are at crafting revenge stories. My Name specifically tells the story of Ji-woo, whose father is killed in front of her by a masked assailant. She is then consumed by the idea of finding the killer, and the show tracks her emotional arc along this multi-year journey. Her life gets very dark, very quickly, yet the show is still constantly asking whether or not she has the conviction to do whatever is necessary to catch her dad’s killer.

During a 2019 trip to LA, I finally had the unique experience of attending a film screening with the director present for a live Q&A afterwards. As someone who makes a habit of watching implicit, purposefully enigmatic movies and then scouring the internet immediately after for answers, being able to hear first-hand from the director of such a film made for an awesome experience. As is also often the case, I learned that I had completely missed the point of the movie, but it was nice to have been told that definitively mere minutes after watching the film and through no additional effort.

One of the increasingly popular subfields of machine learning I started working with is that of generative modeling. A canonical example of the earliest application for modern machine learning is to train a model which classifies whether an image is a cat or a dog. While these classifiers began to work incredibly well, what they couldn’t do was generate new images of cats or dogs. For this task, the field turned to deep generative modeling, which is encapsulated by methods known as Generative Adversarial Networks and Variational Auto-encoders.

Why am I starting this blog?