But I really wonder how many of these founders can convincingly explain in 5 minutes how gradient descent or transformers work. I know that's the first type of question I would ask any founder of an AI company before investing.
The typical thinking among businesspeople is that you can always "hire a nerd" for those things. The early YC was a departure from that with PG praising technical prowess in his essays so much. It would seem this mindset has much less sway now.
The combination of gradient descent and transformers is a weird choice to ask founders about. One is an almost trivial algorithm, whereas the other is a cutting-edge research result.
