Although I work as a software engineer, I work mostly on graphics these days. (I work on hardware accelerated / GPU graphics, on iOS and macOS, using Metal and OpenGL.) Although I have a broad interest and curiosity of pretty much all of software engineering and computer science, the "deepest dive" I've taken in software has been in graphics and computational geometry.
There are a lot of areas I'm only vaguely familiar with, machine (or deep) learning being one of them. There are a lot of interesting ideas in machine learning and a lot of misconceptions. As there is a focus on "neural" networks and since machine learning is often used interchangeably with artificial intelligence, people sometimes assign magical properties or even sentience to these programs.
Under the hood of course, we're dealing with complex, but very much non-magical, non-sentient math.
The YouTube channel 3Blue1Brown has an excellent introduction to this topic. The first video is here:
The playlist containing all in the series is here.
Although I haven't done so myself yet, I suggested supporting this guy on Patreon. His videos are some of the best introductions to complex topics I've found.