He’s very good at running a company that designs graphics cards. And I do mean graphics cards: Nvidia hasn’t really been doing AI all that long in the scope of the amount of time Nvidia has been a company.
As an analogy, imagine if Woz had become CEO of Apple in the early 1990s and stayed in that role until now. Sure, he would be picking up many aspects of running a business between then and now. But how much would he know about the inner workings of the M5 chip or macOS compared to what he knew about the inner workings of the Apple II?
I think Jensen Huang is the same way. Nvidia completely unintentionally stumbled into a pot of gold this era of AI. Yes, CUDA has been a big part of ML for the last decade or so, but LLMs are a whole new ballgame.
I don’t think Huang knows a damn thing about AI safety, superintelligence theories, AGI, etc, and he doesn’t need to, either.