I wouldn’t worry too much on a lot of these points. I won’t say the math behind AI is simple, but it’s mostly undergrad level. You can get up to speed on it if

I wouldn’t worry too much on a lot of these points. I won’t say the math behind AI is simple, but it’s mostly undergrad level. You can get up to speed on it if you really want to. The hard part is writing fast implementations, but many others are already doing this for us. We do not have a grand theory of AI or a deep understanding, but every year we make improvements in machine understandability, and you can “debug” models if need be. Lastly, the author is right, the best models are closed source, but open source is hot on its tail. There are plenty of good local LLMs and they get better every month. Unfortunately it still is out of reach for a hobbyist to train a good LLM from scratch, but open source pretrained models can mitigate this for now.