Ask HN: Path to AI GPU Programming

Regarding AI, the thing I’ve heard the most lately is “during a gold rush sell shovels”.<p>It’s a great advice (?) for a semiconductor company ready to get into the gpu market and compete with Nvidia, but what about me? I’m just a machine learning engineer bored to death by data cleaning and writing over and over the same two lines of PyTorch. I guess in that metaphor I’m the one using such shovels.<p>Perhaps building shovels is much more interesting than using them? What if I’d want to jump into CUDA kernel development? Is that a niche hard enough that the competition is low? Say I’m willing to put in the time and effort, where should I start? Best resources on the topic?<p>FYI, I’m well versed in C++ but zero in GPU programming